0:00
Founders, you're getting in a masterclass here because the circumstances have changed.
Search this video
0:04
If a business is operating in October of 2022, the same way, the same strategy as it was in October of 2021.
0:14
I don't know if you were wrong on the 2021 end or if you were wrong on the 2022, but you must be wrong.
0:19
You can't run this business the way it was run in 2021.
0:21
You gotta have a new plan for 2022.
0:25
If you go out and all you do is make the same mistakes, then I could have gotten in some other company then there's no ROI on this investment.
0:32
My ability to be sitting in Greenwich, Connecticut and yet find somebody in Bangladesh who is like amazingly talented and give that person money.
0:41
That historically never happened.
0:44
This like pursuit of original mistakes is sort of what one is doing as a startup founder.
0:50
Life is movement, stasis is death.
0:54
Hi, I'm Jim O'Shaughnessy and welcome to Infinite Loops.
1:01
Sometimes we get caught up in what feel like infinite loops when trying to figure things out.
1:06
Markets go up and down, research is presented and then refuted, and we find ourselves right back where we started.
1:13
The goal of this podcast is to learn how we can reset our thinking on issues that hopefully leaves us with a better understanding as to why we think the way we think and how we might be able to change that to avoid going in infinite loops of thought.
1:31
We hope to offer our listeners a fresh perspective on a variety of issues and look at them through a multifaceted lens — including history, philosophy, art, science, linguistics, and yes, also through quantitative analysis.
1:47
And through these discussions help you not only become a better investor, but also become a more nuanced thinker.
1:57
With each episode we hope to bring you along with us as we learn together.
2:02
Thanks for joining us, now please enjoy this episode of Infinite Loops.
2:06
Disclaimer: Jim O'Shaughnessy is chairman and Co-Chief Investment Officer of O'Shaughnessy Asset Management, where Jamie Catherwood is an associate.
2:14
All opinions expressed by Jim, Jamie and podcast guests are solely their own opinions and do not reflect the opinions of O'Shaughnessy Asset Management.
2:22
This podcast is for informational purposes only, and should not be relied upon as a basis for investment decisions.
2:29
Clients of O'Shaughnessy Asset Management may maintain positions in the securities discussed in this podcast.
2:35
Jim O'Shaughnessy: Well, hello everyone.
2:37
It's Jim O'Shaughnessy with another episode of Infinite Loops.
2:41
And my guest, many people don't really know because they see our relationship only on Twitter where Jeremiah appears to be my nemesis by taunting me with horrible, horrible excretable puns every week with great consistency.
2:58
Which most people don't know that I actually love because the real backstory here is I have known Jeremiah since he was knee high to a grasshopper.
3:08
Because he went to school and often came to my house because he was a friend of my son Patrick.
3:13
So I've known Jeremiah all of his life, essentially.
3:16
And I want to tell you honestly, he is one of the smartest people I've ever met.
3:22
Full disclosure, I'm an investor in his company, Prefect, because when he outlined it for me.
3:28
And we're going to talk about this as our first question, but he had done maybe the most thorough job of any founder that had come asking me for money, which I thought was great.
3:37
But in addition to being such a tech genius, he can also speak human, which is so rare that I did...
3:46
We were talking beforehand and I joked if we were going to do a Venn diagram of tech geniuses, speak human, you're going to see a little intersection.
3:57
And in that little tiny intersection you're going to see Mr.
3:57
Jeremiah Lowin, founder of Prefect and an all around super amazing guy. Welcome.
4:08
Jeremiah Lowin: Well, thanks Jim.
4:09
I think with an introduction like that, we should just end the episode right now.
4:11
I have nothing else to say.
4:14
Thank you for saying all that, I appreciate it.
4:17
I'm glad someone took the bait.
4:19
I don't think folks know this, but I was a little worried recently and I was like, "I go after you every Friday," and I don't honestly remember how this started or why this started or anything.
4:30
And now I just keep doing it.
4:31
And I forgot a few weeks ago I was in the car, I think it was taking my boys up to see my parents, and I just forgot.
4:37
And I got this text message from you and you were like, "Hey man, what the hell? Where's the pun?"
4:42
And first of all, I was so surprised that you were actually counting on it.
4:45
I thought maybe you've used a reprieve, but I appreciated that.
4:49
I was glad that we found this idiotic way to connect through the ether that is Twitter.
4:54
And I don't know if one person gets a small smile out of it, I guess it makes it worthwhile.
4:58
Jim O'Shaughnessy: As usual, you are demur and so I'm going to blow your horn here for you.
5:04
It's not one person, man.
5:06
It is the number of DMs that I get on Twitter after your Friday pun.
5:13
One guy actually, and I don't know that person will be nameless, but he DM'ed me.
5:19
And he goes, "I feel really weird about telling you this, but I look forward to that pun."
5:27
He goes, "I start looking forward to it like Wednesday."
5:30
And I'm like, "Okay, why?"
5:33
And he goes, "Because it's like..." What's his name...
5:36
"Daniel Craig on Saturday Night Live going ladies and gentlemen, the weekend." Jeremiah Lowin: Yeah.
5:43
Do you know if he could achieve...
5:45
First of all, we're a few zeros short of the engagement on the weekend tweets.
5:47
But I think if we can just achieve any little piece of that.
5:50
You and I have talked for a long time since we've been sort of cooperating as business partners in this about memetic behavior.
5:57
And following some sort of cultural whatever it is. It could be nothing.
6:01
It could be idiotic, it could be a pun, but it's funny to see how that actually plays out.
6:06
And we've created this expectation and we're going to satisfy.
6:07
And I do my damndest, my wife comes down and I'm sitting at the kitchen counter having my coffee and just trying to come up with something, anything. It'll be Tuesday.
6:16
And I'm like, "I don't understand.
6:17
Jim's counting on this now. I know that he is."
6:20
I have lost so many hours now to these dumb jokes.
6:22
But we keep doing them because somehow they are connecting and it's just fun.
6:28
And frankly, it's so different from everything else that we do.
6:30
And the market's crashing and we got to do this and we've got business, and this is this little island of nothingness that is so much fun to get to revisit.
6:38
I do have people who DM me now, by the way, especially with your recent announcements.
6:39
And they're like, "Do you know Jim, is it okay?
6:42
Do you know who that is?"
6:45
And I don't know what to say to them.
6:47
I'm like, "Yeah, I know him. I think it's okay. I think he's cool.
6:50
And I honestly don't know."
6:52
Jim O'Shaughnessy: No, actually you'd know the actual secret that I am truly an evil genius.
6:58
I am way, I am the Emperor Dark in if we're doing Star Wars.
7:04
Jeremiah Lowin: I know this.
7:06
I just didn't realize we were telling a few tens of thousands of our closest friends on this podcast.
7:13
Jim O'Shaughnessy: Now let's... Guys, joking. I'm joking.
7:16
Jeremiah Lowin: Just a JK, LOL.
7:18
Jim O'Shaughnessy: Ha, ha, ha, ha.
7:21
Well, a serious observation actually about the whole pun thing and everything.
7:27
What that buys you is you are such an intense thinker on what you do.
7:31
And that's going to be my next question because I want an update on what we talked about a couple years ago.
7:37
But so you are really able to focus, which is now becoming an asymmetric advantage.
7:44
And to everybody who can be like that, they need slack time, they need downtime.
7:51
If you're going to be creative and you're going to be effective, you got to have something in your case, maybe it's thinking about the pun or whatever.
8:01
But you got to build that into your schedule because if you don't have it, you just go to burnout, in my opinion. Jeremiah Lowin: Totally.
8:08
And for me, I speak with someone like yourself.
8:10
I speak Patrick and he's telling me about the 2000 books he read last month and this, and he's learned about this.
8:17
And it's this nonfiction and it's this autobiography.
8:18
And I'm thinking about the books that I read and they're way out there like science fiction.
8:23
I love this, Hitchhiker's Guide and some Rick and Morty and all this stuff where for me, when the world is a very real place, a lot of the times.
8:33
And a very intense place, especially when we're making decisions that have consequence and increasingly decisions that will impact the careers of the folks that I work with.
8:42
And when I'm doing something that's not that I want it to be far from that, I want it to be something, as I've said to folks, "It's got space, it's got superheroes, it's got explosions, or it's something to laugh at."
8:52
That's my criteria for entertainment.
8:54
If it's not that, I'm probably not that interested.
8:56
Jim O'Shaughnessy: I love it.
8:57
Jeremiah Lowin: Dragons, dragons too.
8:59
Jim O'Shaughnessy: Yeah, dragons. Oh, come on.
9:01
If you don't like dragons, you can't be my friend.
9:03
Jeremiah Lowin: Exactly, it's a hard one. Jim O'Shaughnessy: ...
9:05
Because, it is, it's a hard line.
9:07
And I wish I could have a hard line with Adams because as I've shared with you, if you stop thinking about him as a sci-fi writer and start thinking about him as a philosopher like, "Oh my God, does it make so much more sense?"
9:21
And so I've been having some experiences recently where all I can think of, the perfect metaphor or analogy is the Vogons.
9:29
I'm like, "You don't understand.
9:32
All these people are Vogons and young..."
9:36
I will admit it's mostly younger people.
9:39
They'll be like, "What's a Vogon?"
9:41
And I'm like, "Oh my God."
9:43
Because he created that perfect image of complete incompetence that was institutionalized in a species that fucked everything it touched up in the galaxy.
9:56
Jeremiah Lowin: Pure bureaucracy. Pure bureaucracy.
9:59
Jim O'Shaughnessy: Totally. Nothing but red tape.
10:01
Jeremiah Lowin: Now, I can't remember if we talked about this last, and this is the dangerous thing.
10:04
We're going to find out if anybody is really, really listening to these episodes or as I volunteered to you before, if we're just scraping the bottom of the barrel for guests at this point.
10:12
Because I'm probably going to say some things that I said two years ago, last time you had me on, and we're just going to deal with it.
10:16
And this is one of those things, I don't remember if I said this or not, but Hitchhiker's Guide, Douglas Adams in particular is the reason that I like pun so much.
10:25
Because that book is just line after line after line of this incredible, very British wordplay, which when I was a kid, fascinated me.
10:34
The way that he says things, there's a very famous line, which is, that's about this liquid, this drink.
10:39
And he says, "It's almost, but not quite entirely unlike tea."
10:43
There's this line about when the Vogons arrive and he says, "They hung in space in exactly the way that bricks don't."
10:51
And it's this inversion of expectation.
10:53
And I've remembered these phrases, these lines for decades now since the first time I read them.
10:58
And I just love them and there's one which I actually couldn't understand for a really long time and I'll probably mess it up in trying to retell it right now, but there's a moment when the two characters first go through hyperspace or whatever and one of them says, Ford says to Arthur, he says, "That was so unpleasant."
11:14
And he says, "Yeah, it's a lot being drunk."
11:17
And the other one says, "What's so unpleasant about being drunk?"
11:19
And he says, "I don't know, ask a glass of water."
11:21
And when I was young, I never understood the joke.
11:26
I was like drunk with alcohol.
11:27
I never connected it to actually the act of drinking and that little tiny little thing...
11:32
I mean this must sound preposterous to everybody listening to us.
11:34
But I read this book probably once a year, I get to this line and I wouldn't understand it because the little, some switch in my brain just hadn't quite gone to the on position to understand that particular piece of wordplay.
11:45
And it's not that there's complicated about this, it is literally an idiotic pun.
11:49
But there's something there about finding that connection and understanding this dual meaning that I just got obsessed with and I've just carried now with me.
11:58
And so when we do these jokes and they are idiotic, but there is a piece of me that absolutely loves each and every one of them.
12:05
And we've never repeated one.
12:06
We've never ever repeated one.
12:08
They're always different because there's always something new to explore.
12:10
Some of them are terrible, some of them are terrible, some of them are really bad.
12:14
Those are the ones I wake up on Friday morning.
12:15
I'm like, oh goddammit, I've got to do the pun.
12:18
But there's a few of them that are really, really clever.
12:20
Jim O'Shaughnessy: That's quite true. That's quite true.
12:23
Jeremiah Lowin: Those are the ones that I didn't probably come up with.
12:25
I walk around, I'll see a coffee shop with one of those chalkboards in front of it where they've said coffee is great, or some pun that I can't think of right now.
12:33
The daily grind or just brew it. And I'll write it down.
12:36
I'll be like, oh, that's good.
12:37
I can work that into my little narrative structure with Jim somehow.
12:40
Jim O'Shaughnessy: Listen, the mark of a great artist is they steal broadly and- Jeremiah Lowin: I steal all the time.
12:46
Jim O'Shaughnessy: Yeah, me too.
12:47
And anyone who says that they don't is lying or incompetent in my opinion.
12:52
On the last thing, Prefect, right?
12:55
Ford Prefect, do people make that connection or no? Is that too deep?
12:59
Jeremiah Lowin: Sometimes they do, but usually they don't. Usually they don't. I don't know.
13:05
I think it's a little just outside the norm for knowledge.
13:08
But the nice thing about the reason we call the company Prefect in fact, is not to honor this character, but because what we do.
13:15
We're on the spectrum of Vogon-ity in that we are bureaucrats in some sense, we help companies make sure their data infrastructure is running properly.
13:25
We keep an eye on it, we monitor it.
13:27
And that is the role of a prefect.
13:30
Now interestingly, the place that most people think ‘Prefect’ comes from is Harry Potter because the Weasley brothers were prefects, they were like RAs, right? Within Hogwarts.
13:42
And so today it's actually, it's not Hitchhiker's Guide, which is in fact where our culture is borrowed from, but it's Harry Potter that's actually provided a really familiar place for most people.
13:51
At least in America, where we don't have prefect as either a government or school position.
13:55
And so that's been really interesting is people will ask about that all the time.
13:58
They're like, is this a Harry Potter reference?
13:59
And that took me a moment to figure out where that was coming from. That's the most common.
14:03
Jim O'Shaughnessy: That's brilliant because you're just, you're essentially mapping people's neural real estate and you're seeing where are the bright, warm, sunny spots where they really like something.
14:17
Prefect from Harry Potter would be one for people of my vintage, Ford Prefect from the Adams series.
14:24
But the other thing you did, which I think is brilliant, and understanding that you can colonize already well-trodden neural mind real estate, what a skill that is.
14:36
Is the other thing that people silently see when they see this, they see perfect.
14:42
Jeremiah Lowin: So I actually need to tweet this.
14:46
Just last week we had sort of a whole bunch of internal promotions and folks taking new roles, which is always wonderful.
14:52
And so the folks in our Chicago office had a cake made with the faces of two of our colleagues who just moved into new roles and they asked Costco to print their faces on the cake and write ‘Prefect’ under it.
15:03
And very clearly Costco decided that we were idiots and didn't know how to spell and so the cake came back ‘Perfect’.
15:11
Which works just as well, right?
15:13
Because they are perfect employees and they are also Prefect employees.
15:17
I mean we get offers all the time addressed to Perfect Technologies.
15:22
And it's one of those things where we can always make a little joke about it like, huh, we wish. And then we move on.
15:28
Jim O'Shaughnessy: It's great, but I want to get to the second question because obviously, interesting year so far in financial markets.
15:37
In startup companies and in their ability or inability to get funded. And one of the things...
15:44
I put up a quote by Steve Case a few days ago that was along the lines of, “No matter how good your idea is, if you can't execute against, it's a hallucination.
15:51
” And you and I have talked about execution a lot.
16:01
So many people focus on, I want to do the zero to one, I want to be the zero to one guy. Well, okay great.
16:09
I just made an investment that I think was zero to one.
16:13
We're going to talk about that later.
16:15
But I think the real big winners are the ones who can get one to ten right.
16:19
And so before we started recording, you were telling me about what you put in place, which I think is genius because you did it in January and obviously you've hooked into my human OS and I'm like, oh, what a genius this guy is. But talk a bit about...
16:32
Because last time we talked a lot about building the company, how you were doing that, et cetera.
16:38
And now we're two years on, I'm pretty happy with as an investor, seeing where you've gotten to.
16:44
But talk a bit about the business building aspect and what you've learned because we have a lot of people who are founders who listen to us.
16:51
Not nearly as many as my son, the podcast king, but we have a few.
16:55
And more importantly, we probably have a lot of the funders of those founders that listen to this particular podcast.
17:02
So talk a bit about the business building, what you've learned, and then specifically I'd like you to talk about the risk group you put together and why.
17:10
Jeremiah Lowin: I think when we look back at that episode we did two years ago, there's a lot of things there that I strongly believe, but one could reasonably also hold my feet to the fire in a sense and say, so what?
17:24
How does this actually play out?
17:25
Why do you waste your time worrying about all these outcomes?
17:29
The startup game is, I think people obsess over the zero to one and the easiest way to go from zero to one is to forget you're at zero.
17:35
To just run for, just get there.
17:38
But then you get there and you're right, all of a sudden execution really matters.
17:42
And by the way, I really appreciate this quote.
17:43
I keep a sticky right here on my screen.
17:45
It says, "Ideas are always wrong and execution is always right." Jim O'Shaughnessy: Yep. Jeremiah Lowin: Right.
17:50
You can't stop halfway through it.
17:51
You have to actually get it done.
17:52
And I think the important thing is to understand what is execution?
17:54
What does it mean to perfectly execute something?
17:58
If you think that it means that you announce something exactly how it was going to go and then you did it, you live in such a fantasy land.
18:05
Yeah, I'd love if that were possible.
18:06
I'd love to declare right now we're going to do this, we're going to do this, we're going to have a hundred million in AR, it's not how it works.
18:13
The world is a probabilistic and wonderfully intricate place and it is so much easier to destroy order and introduce chaos than to contain it.
18:22
And so the idea that we're going to declare precise outcomes and then have not our own efforts, but also the entire world conspire to materialize them is insane.
18:34
Especially because if I canvas the other companies of similar size to myself in my space and out of it, the way that we speak about those goals are all in conflict with each other.
18:45
They're wildly inconsistent.
18:46
We're not describing a world that can even exist.
18:49
And so reconciling our ability to speak the future and achieve it in a way where all of these companies, all of these actors, all of these players are going to do it in the same way is sort of what I obsess over a lot.
19:03
And when we spoke two years ago, we talked a lot about ways of thinking about the future.
19:08
We talked about not pinning to precise outcomes.
19:11
We talked about sort of embracing a distribution of outcomes and how one can think about that.
19:16
And honestly, maybe we even have that conversation again, cause I think even my own thinking has evolved from there.
19:21
But we had an opportunity in January of this year, of 2022, to really, really, really put a lot of that into practice in a very, very, very tangible way, which started very simply with me getting scared, which is how a lot of things at Prefect start.
19:35
I got a little nervous about just the world, the market, the inflationary pressures, everything.
19:40
And so we formed this group, which we called it the R2-D2 group, which was a pun with an adjustment that was sort of in the wrong cultural bucket.
19:51
But we didn't expect this group to go anywhere, so we didn't think too hard about its name.
19:55
All its name had to do was not be what it was actually was, which was a recession planning group.
20:01
Because if this name were viewed by anyone, it would frighten them.
20:06
And the point of this group was not to be a frightening group.
20:09
It was supposed to be a confidence inspiring group, but in order to be a confidence inspiring group, it had to become a safe place to talk about very, very frightening outcomes.
20:18
And the reason I decided to form this group was because I remember going into COVID, so what, two and a half years ago, and this feeling of simultaneously needing to make plans and execute those plans.
20:31
Which is to say evaluate those plans in the same moment and never really getting that opportunity to take a deep breath and actually think holistically about a strategy that we were putting forward.
20:43
That was very much just a moment of like, can we do this? Can we get through this?
20:47
And I think I have never felt as off guard, as caught off guard in the history of the business as I did in that moment and sort of resolved that we would never find ourselves in that position again.
20:55
And so in January this year, all this group together said, I don't think what I'm about to describe is actually going to happen, but I think the probability that it happens is high enough that if it does, we need to be ready for it.
21:09
And very much what is playing out, not in a way that I went out and bought puts on the S&P 500 or anything like that.
21:17
I wasn't predicting with such precision.
21:18
But a lot of what we're seeing now were baked into the assumptions that that group outlined and then began to make plans for.
21:25
Capital markets would shut down, stock market would collapse, inflationary pressures would rise, there would be massive layoffs in our industry.
21:33
And so the question becomes what do we do?
21:36
And the instruction to this group was to find a way for us not just to survive in such an environment but to thrive.
21:42
And this was something that Patrick, who is on our board, and I had some very, very, very fascinating conversations about.
21:47
What does that actually mean and how much prep work needs to be done in advance versus just demonstrating we were ready to do something and then we just run forward and do it?
21:55
But the most important thing that we did with this group, we'll talk about the outcomes in a moment, but the most important thing that I think we did is we created a space where it was possible with no caveat and with no sugar coating to talk about these draconian outcomes.
22:09
What would happen if there was a recession so severe that our largest customers called us up one day and simply...
22:14
and they just went bankrupt, right?
22:17
Let's just go, let's live in an extreme world. What would we do?
22:21
How would we pay our bills?
22:23
How would we plan to move forward? Who would we ally with?
22:26
What would we say to our investors?
22:28
Would our investors even be there for us?
22:30
Have we earned credibility subset in the absence of if a customer disappeared like that, could I go to our investors and raise capital?
22:38
Maybe I could maybe, maybe I couldn't.
22:39
Maybe relationships are strong enough, maybe the evidence is too powerful.
22:42
There's a lot of unpleasant things that I needed a space for us to be able to talk about.
22:48
And these are the sorts of things where I think every startup is a default dead business.
22:53
Every young startup, if it just stopped caring would fail.
22:58
And so the exercise of building a startup is about not acknowledging that reality.
23:02
It's about looking to the future and to positive outcomes and living there and manifesting that and getting there.
23:09
And so dealing with these things as real threats feels a little odd for a startup.
23:16
It feels a little counter-cultural in a sense.
23:19
And that's why creating this group was so important because it gave us that space.
23:25
So we did this in January and kept having these meetings.
23:29
We met regularly, we made plans, we made budgets.
23:31
I'll talk to some of the sort of actionable things that we did that I think other companies could take advantage of.
23:38
And then as we got into May, we became violently aware of a change that started in capital markets and the rising costs of capital.
23:49
And you could see it in my inbox even, you could see the outreach from VCs, which is constant.
23:56
The drum beat stopped one day, it just ended.
23:58
And the conversations that we were having got more drawn out and you know I've always said, I've always had just a very, very, very open door relationship with the finance community having come from it.
24:06
And so it gave us a very real pulse of what was going on.
24:09
And so one day, sure enough, we said, it's time.
24:12
Something is happening, we don't know what.
24:15
This was the week before, this is in middle of May, I think.
24:18
This is a week before the market did that first big sell, did that first big sell.
24:24
But a week before that we decided to go to our, well, what you and I can today call a recession plan.
24:31
But it wasn't a recession plan that time.
24:32
We called it our one scoop plan.
24:35
And maybe here I'll just talk about how we actually, what does it mean that we got a group together?
24:39
What does it mean that we've made budgets?
24:40
What does it mean that we've made plans for a recession?
24:42
So there's a few things that go into that.
24:44
The first is understanding what is our budget?
24:45
We've always had a great budget at Prefect, we've always cared deeply.
24:48
And even as a young company, we tried to keep at least 24 months of board cash flows. Were they nonsensical?
24:55
Probably in all honesty, they were probably nonsense.
24:58
But they meant that we understood exactly what the dynamics of the business were.
25:03
And they meant that if we had to think about investing cash here or there or hiring this person, we actually could see what it would do to a very young business.
25:12
And I've learned most founders of seed and series A businesses not only don't think this is a good idea, but they're advised not to do this because it can't possibly be real.
25:22
And I have exactly the opposite.
25:23
In fact, I'll confess something right now.
25:26
This is a mistake I made once, and I don't think I've talked about publicly unless of course I talked about two years ago and don't remember.
25:32
One night I was looking at this...
25:34
The company is a few months old at this point.
25:36
It must have been one now, but it wasn't a year old.
25:39
One night I'm looking at this spreadsheet and I'm summing up expenses and I accidentally include in my sum the existing sum, the total.
25:47
And so I'm double counting everything.
25:49
And I don't know, it was late at night, I was doing some analysis and I conclude that the company will run out of money in three or four months.
25:57
Which shocked me because I thought we had a year account or whatever it was.
26:00
And I had this panic moment.
26:04
I let it in for one sec, right?
26:05
And I'm like, this can't be.
26:06
I literally, I remember thinking, it was like, how did I screw up this much? How did it come to this?
26:13
And sort of just went into action.
26:16
And I brought up all these bad outcome plans that we had made, that we were ready for.
26:22
I was like, okay, so we're going to do this, we're not going to hire this person, we're going to do this, we're going to get this product to market, we're going to do this.
26:28
And it wasn't until about a day later, I think, or that night I was standing in the shower and just, how did I do this? How did I screw up?
26:36
And all of a sudden I have this spreadsheet in mind and all of a sudden I realize my idiotic and very simple mistake.
26:41
Go out, huge sigh of relief.
26:43
The company is not dying overnight.
26:46
But I kept the plan that I had made and that feeling of, oh, thank God as a consequence of this plan, I am guaranteeing the extension of the life of this business.
26:58
That's something that I chase now. I love that feeling.
27:01
I hate the feeling of, oh God, I don't know.
27:03
This seems like a good investment, but I don't know.
27:05
That margin of safety that comes with being overly conservative, but still being able to achieve the goals of the business is my new favorite thing in the entire world.
27:13
And so this was a group designed to manufacture that feeling of safety.
27:19
But how, what does that mean? Okay.
27:23
So we did a couple of things.
27:24
The first thing that we did is we looked at our budget.
27:27
I just wanted to know where are we versus our budget?
27:29
Because the budget at the end of the day tells us how long the company will live.
27:32
So I need to know where we are against that budget.
27:34
We can plot that budget over time and it grows because our expenses grow, so do our revenues.
27:38
So plot that budget over time.
27:40
And then we plotted two more lines.
27:42
We plotted one line, which was, let's say someone walks up and just wants to buy a huge chunk of our business and make this investment and we're going to grow the business in a really accelerated basis. What would we do?
27:52
If someone just showed up today and was like, here's a check. What do we do? We have to know.
27:56
So that became this great outcome line that we plotted.
27:59
And then we did the recession line, basically.
28:02
If capital markets are closed and we become extremely conservative with our capital [inaudible] capitals higher, what would we do in that case?
28:08
And so we plotted these three lines over time, and they make sort of ice cream cone shape because they'll start from the same place and then they expand over time.
28:16
So because of that, we started calling them the one scoop, the two scoop and the three scoop budgets.
28:21
And again, it became this euphemism for the thing you didn't want to actually talk about, recessions, opportunistic finances.
28:28
We don't want to talk about those because then we're talking about stories and outcomes that aren't real.
28:32
One scoop, two scoop, three scoop.
28:33
This we can talk about all day long and it doesn't have any meaning. It's just a balance.
28:37
But the important thing is it wasn't just cute.
28:40
We could plot our expenses at any moment in time on this and know, quantify even, where we were versus our expectations and adjust them over time.
28:49
So I could say we are 20% of the way between the one scoop plan and the two scoop plan, or we are 30% of the way from the two scoop plan to the three scoop plan. And that's too far.
29:00
We could actually talk about what it meant to be too far from our budget, too far ahead of ourselves.
29:05
One of our objectives was that if a recession showed up one day, we wanted to be able to transition to the one scoop plan and not have to deal what a lot of companies have done, which is lay people off.
29:14
That's an extraordinarily damaging thing to do to a company.
29:16
And it sort of reflects that a company got too close to its own three scoop without really realizing it.
29:21
And so think about what that means on a chart like this.
29:23
It means I'm holding up my hands, even though nobody of course can see it except for you.
29:27
It means that the line from the two scoop plan to the one scoop plan should stay flat.
29:33
That's what it means to not have to do a layoff.
29:35
If the only way for us to get to that one scoop plan in time before we run out of money is to go vertically down, that's a really, really, really bad and unpleasant scenario.
29:43
And it became the goal of this group to guarantee that at any moment in time we could transition to the one scoop plan without having a negative slope in that line, so to speak.
29:51
So that became this group's mission, was to keep us...
29:54
We can't just move to the one scoop because in a world where the recession doesn't show up, that puts us as an extraordinary disadvantage.
30:00
But we also can't stay at just the two scoop because there's increasing probability that we'll need to go to the one.
30:05
And so we came up with this strategy for quantifying that and more importantly for talking about it, forget the quantification.
30:13
The quantification is an output of this process.
30:15
But it gave us a way to talk about it that led to really constructive conversations and debates, in fact, about the right way to think about what these budgets should be.
30:24
These budgets aren't handed to us by God, we made these budgets, were they too high? Were they too low?
30:29
And so it was in the middle of May when we decided we needed to go to our one scoop plan as a protective measure, as the world basically decided it was time to end.
30:40
And this I think, will for some time be the most tangible output of a lot of the processes and thoughts that you and I discussed two years ago, which do in all honesty, sometimes seem like an enormous waste of energy.
30:54
But every now and then, I guess this is the risk persona speaking, every now and then they're super useful and really let us operate in this constructive way.
31:03
Jim O'Shaughnessy: So I mean, there's just so much to unpack here that I'm just going to kind of review the way what I heard so that I'm getting it right.
31:16
The understanding of semantics is so incredibly important. And I have on...
31:24
I'm usually pretty good at reframing people and getting them to understand the old joke, death insurance, zero sales, life insurance, trillions in sales.
31:33
Jeremiah Lowin: I've never heard that. But that's very... Yeah, I like that.
31:37
Jim O'Shaughnessy: That's one word.
31:38
And what we don't understand is because people haven't really done deep dives on human OS.
31:42
We were talking before we started to record about the work of Will Storr.
31:47
You really need to do this deep dive because we are story creatures.
31:53
So if you understand the way that homo homo sapiens, smart human, right?
31:59
Very few people know that it's our actual name is homo homo sapiens. Really smart humans.
32:05
Well, there's a line and everyone thinks this little part, and what I'm doing is an exponential curve here.
32:13
Listen, we are still guided by our hunter gatherer genes, which are still very much here.
32:20
And there's one theory that I love actually, and that was language was invented to gossip.
32:26
Because essentially you, if you're in a tribe and you think that Grog over there is a freeloader, you're not going to just come right out and point at him if you don't have no language or go beat on him with a bone.
32:42
You want to make sure that the group kind of thinks what you're thinking.
32:46
So we invented language so we could gossip about Grog and I could come up and kind of go, "Hey Jeremiah?" "Yeah Jim?"
32:54
"Do you think that Grog is a fucking freeloader?" "You know, I do."
32:58
And then it would go through the rest of the tribe.
33:02
And this is like Sun Tzu, “The battle is won before you have the army start to march.
33:02
” So because we acquired language, we could tell stories of gossip about Grog and then he's already dead and he doesn't even know it because you and I started talking about him and the rest of the tribe agreed.
33:21
So the incredible importance of semantic and story, and this is coming from a quant, okay?
33:30
I used to give speeches that were, I want to tell you a story about why paying attention to stories in public investing is a bad idea because he who has the best story wins in many respects.
33:45
And now it leads to the second observation is creating a place where people know that they can be frank and be open.
33:55
Because listen man, if you're sitting in a room and you don't know how to read a room, a lot of people don't.
34:02
Luckily it's, I've been lucky that I'm not horrible at reading a room.
34:06
And you'd know who's sitting over there in the meeting and thinking, yeah, Jim, yeah. Just keep talking. You are so full of shit.
34:14
You have no idea what's coming, you fucking idiot. Right?
34:18
And listen, the third thing is we all fail.
34:23
And if you are able to reframe failure as an opportunity, as opposed to as some horrible thing that we're going to be deeply ashamed about. Everyone that I've...
34:34
We were talking about David Senra earlier, the Founders Podcast.
34:37
He and I were talking, we click because I so agree with him on so many things.
34:43
And it's like, show me a founder that doesn't have a failure in his or her past, and I'll show you a fraud. Because people who...
34:54
It's like, I love this idea, just keep going.
34:57
The easiest way to get from zero to one is just go.
35:00
Jeremiah Lowin: It's easier than people think to go from zero to one.
35:05
Jim O'Shaughnessy: Exactly.
35:06
But the third observation here is, this is my little hobby horse and soapbox, we are deterministic thinkers living in a probabilistic world and usually hilarity or tragedy ensue, right?
35:19
And this whole zero, one, black, white, all, nothing, is you're doomed.
35:25
If that's the way you think you are doomed, because most of life is gray and maybe, right?
35:33
So when you open a space for people to, and I love again under the deep understanding of semantics here.
35:41
So I studied neurolinguistic programming, which has the unfortunate thing of having the same initials in an acronym as natural language programming. Jeremiah Lowin: NLP. Sure, sure. Yeah. But it's very different.
35:54
Jim O'Shaughnessy: Very, very different.
35:55
We're going to talk about that in a bit too.
35:56
So neurolinguistic programming is, it really looks at understanding that how people communicate with one another is like everything. And if you... Who was it? I can't remember.
36:10
I'm going to just steal the quote without attributing it to the author.
36:13
But he basically said, and I paraphrased his quote, that “We are these islands of misunderstanding shouting at each other across sea of uncertainty.
36:21
” Kipling, I think it was.
36:27
So, I mean, I love that quote because it's so true.
36:32
And so if you open up the experience, and this is why I love you did not use the term recession because again, this is one of the things that I go out of my way to avoid.
36:43
But we are label thinkers too.
36:46
And if you say recession, you've implanted, it's like, don't think of a gray elephant.
36:50
What are you going to think about? Right?
36:53
All you're going to think about and your reactions because, and then I'll get to my next question, the other understanding about we domesticated primates is that we are primarily emotionally based.
37:06
And the emotion that drives us is fear.
37:09
And the minute you understand that and...
37:14
Novel threats, we humans have been trained to go snap all of our attention on novel threats.
37:22
What the fuck is that that's moving behind that bush over there? So by not...
37:28
I love the ice cream, the one scoop, two scoop, three scoop, because you have semantically made it okay to talk about this.
37:37
So my question now is a follow up is what came out of those group meetings that like beforehand, you would've been surprised? I. e.
37:50
you had the foresight to think about setting it up, but what didn't you as the CEO and founder, what didn't you get that afterwards you're like, "Holy shit, I'm glad I put this committee together."
38:00
Jeremiah Lowin: Well it's fascinating because, let me skip to the punchline of that actually, because I think it's how I came to view the framing of this and aligns really closely with what you were just talking about.
38:10
In fact, my team heard me say many times that I believe that naming things is a superpower.
38:16
And when we name things at Prefect, we care deeply about them.
38:19
And this resonates throughout the organization.
38:21
With engineers, it's about what variables you might use.
38:24
At that scale it matters because it's a way of communicating to a future developer or even yourself.
38:28
When it comes to marketing, obviously naming things matters.
38:30
When it comes to missions that we have as a company, we agonize over the precise words that we use.
38:35
We even added four letters to our mission at the last minute, the 11th hour before we announced to the team most recently.
38:40
And I messed up one of these.
38:41
There was a mission that we announced to the team where I used a word that was a synonym, but not the exact right word for what I wanted.
38:48
And I won't go into detail cause it's probably boring for everyone who's not me, but I came to view that as one of the bigger mistakes of that.
38:55
This is about what, six months ago?
38:57
Because the word that we chose was correlated, but not the same as the outcome we wanted to drive.
39:05
And it led to a series of actions that were almost, but not entirely unlike what we were trying to do.
39:11
That naming things is the greatest superpower we have and it extends into communicating and all the semantic ideas that you're describing.
39:18
What ended up coming out of this was, at some point having decided to enact this, we need to share this with a team.
39:25
We need to explain to Prefect that we are going to operate differently.
39:28
In retrospect, it's very easy to talk about because frankly it's never been easier to run and have confidence in our business than now running on a much tighter budget with hiring for only strategic roles, as opposed to more opportunistically and revenue continues to grow with our beginning of objectives.
39:47
The business is very easy to run right now.
39:51
That was actually one of the most surprising things and it was why I remember that night sitting at the counter thinking, "Oh shit, I screwed everything up."
39:57
But kept the plan that I came up with at the time because it creates that margin of safety but it's a hard thing to communicate to folks, right?
40:06
Because there's a slippery slope there.
40:07
The ultimate margin of safety is to shut the business down.
40:10
How do we find the right balance?
40:13
Ultimately, what came with this was communicating this as something related to the cost of capital.
40:18
I remember having a conversation with Sam Cates, who works with Patrick at Positive Sum, and I was sort of giving him a little preview of my thinking on this.
40:24
I was saying, "Look, cost of capital is higher.
40:26
We're going to adjust how our company operates.
40:28
We're not going to sacrifice our top line.
40:29
We're going to adjust our expenses though, and recognition the cost of capital."
40:32
And Sam, I hope he doesn't mind me saying this, he literally goes, "You can't do that.
40:37
If companies could respond to the cost of capital like that, then they all just would do that and this wouldn't be a problem."
40:43
He's like, "These companies can't just do that."
40:45
I was like, "Well, we're not most companies.
40:46
This is what we're going to do."
40:47
That's what we went off and did.
40:49
That realization that we actually, even when we discussed this with our team, so I learned from my team, half of them are intimately familiar with what's happening in the markets and they're watching the market fall and they're watching interest rates go up and a large
41:00
number of them are not familiar with that and don't come from finance like I do and obsess over this stuff and maybe they just don't interact with it or it's not interesting to them and that's totally fine. First we need to explain what's happening in the world and then once we've explained
41:11
First we need to explain what's happening in the world and then once we've explained what's happening the world, we need to explain how it affects us and ultimately this lever, this cost of capital and our response to it by saying that what capital buys us is time.
41:25
What Prefect uses with time is we use it to experiment.
41:28
If the cost of capital has gone up, the cost of time has gone up, the cost of experimentation has gone up, we're going to experiment less and we're going to double down on our core competencies.
41:37
Our revenue is not an experiment for us, which is why we're not going to compromise there.
41:41
But there are initiatives that we have that are very clear and that we'd like to pursue that are very clearly experimental and very clearly exploratory.
41:46
Those we're going to tighten up and that's how we're going to pursue our top line without, we're going to tighten our bottom without compromising our top line.
41:55
That was one of discovering that there was a narrative here, which wasn't frightening.
42:00
In turn what we can now say is the pursuit of growth in all costs, which was such a hallmark of especially VC back companies in the last two years.
42:11
Not only was it a bad idea, it turns out you can run a great business without it.
42:16
I think that was, that's sort of the great lie if you will, of the investing world last years was that you had to do this to grow and with the opportunity to step back, which is not an experiment I think I would've been brave enough to try, right?
42:31
In two years ago, if we're having this conversation or a year ago, if we're having this conversation, there's a row, there's bootstrap companies and there's investor back companies and we pursue growth in different ways and we don't cross over and we don't try to hold back.
42:43
To be clear, Prefect is not and does not manner like a bootstrap company.
42:46
I have such ultimate respect for that category of business, which I used to.
42:51
Once upon a time I was, and that was a very different experience.
42:53
But it was the realization that we don't need to grow and to operate in the manner that a bunch of VCs seem to be insisting that we are and we can still operate in a perfectly fine business.
43:04
That was the light bulb, that was what I didn't expect.
43:08
What I went into this group seeking to understand was, "Where do we need to make compromises in our business?"
43:14
I expected those to be universal compromises.
43:16
I expected to end up with a business that was harder to run, slower to scale, more difficult to motivate, more difficult to pitch.
43:24
What I've ended up with, there's a business that's easier to run and scales faster and is more nimble and is more profitable than at any previous time in its life.
43:36
That is that how we came to that mission of not survive but thrive.
43:43
Well, whatever this is, we're going to be fine on the other side of it, but we're not just going to be fine.
43:49
We're going to be infinitely preferable company to how we went into it because we had this opportunity to do the thing that frankly we probably would've been too scared to do otherwise and experiment, so to speak.
43:58
I just did air quotes, of course, forgetting we're on a podcast but to experiment with operating in a very different manner.
44:07
That's what I couldn't have expected.
44:08
Jim O'Shaughnessy: I will remind you that we also release on YouTube now, so we have a fairly large... Jeremiah Lowin: This? The whole time?
44:17
Jim O'Shaughnessy: Yeah, yeah, yeah. We're both doing this.
44:19
As you were talking I was thinking, you were talking about you have framed something just a little wrong and I was thinking of that wonderful attributed to Mark Twain quote, which is, "The difference between the perfect word and the almost perfect word is the difference between lightning and lightning bug."
44:37
Jeremiah Lowin: That's great, love that.
44:40
Jim O'Shaughnessy: That is so true. It's just so true.
44:44
The other thing, I think, is that you're giving a masterclass here for other founders because a lot of this isn't intuitive, right? I love the idea.
44:58
Again, I'm a finance guy, so I think about in terms of cost of capital all the time, "Okay, what's my cost of capital here?"
45:06
I look at that as a hurdle rate, what's my hurdle rate?
45:10
Now that hurdle rate has changed dramatically. Guess what?
45:14
It changes the decision vectors about what I'm going to invest in and what I'm not. Right? Jeremiah Lowin: Exactly.
45:24
Jim O'Shaughnessy: It's hard to convey to people who are not finance people that I used to say, "If the math gets any higher than algebra, it's pretty certain you're going to lose your money, all of your money."
45:38
Because honestly, not to belittle the super brains in my field but, come on man, this is basic stuff.
45:48
Cost of capital, you don't have to be a genius.
45:51
I could have my grandson Pierce doing cost of capital analysis.
45:56
Well, now he is a genius but let's just put that to the side that I could have eight year old Pierce O'Shaughnessy figuring this out really, really easily because it is that simple.
46:09
But people don't, I am in awe often, of like, "Did you think about this thing over here?"
46:15
Jeremiah Lowin: Well, with the cost of capital, what it had been, there are massive numbers of companies that have come into existence as a bet essentially that cost of capital say the same because it enables extraordinary experimentation and tiny things that couldn't possibly generate an ROI with any real hurdle rate.
46:35
When your stakes are zero, extrapolate into the future and become incredible things.
46:40
I think the amazing thing is it's very easy for us now to say like, but that's an illusion.
46:43
But the fact of the matter is if interest rates stay low for long enough to incubate, build and then sell a company, then who care if it was an illusion because you actually did it.
46:52
The number of folks who I think pursued that recently is extraordinary.
46:55
I remember when COVID first took place, so back in March 2020, I remember we were talking and saying, if your business is running in April of 2020 the same way it did in February of 2020, you are wrong.
47:10
You are either wrong before or you're wrong now, but the world has changed and if your business hasn't then you're wrong.
47:15
I think that is still true today.
47:16
If a business is operating in October of 2022, the same way, the same strategy as it was in October of 2021, I don't know if you were wrong on the 2021 end or if you were wrong on the 2022, but you must be wrong.
47:31
Or you don't know what you are doing.
47:33
That is your job, is to adapt in the face of massive economic change to situate yourself sort of optimally.
47:41
Interestingly, this was, I recently as I was in California recently and took advantage of being there to sort go on a listening tour from some of the greatest investors in VCs that I've built relationships with and just find out what are they seeing?
47:54
For many of them, interestingly because I've been such a COVID hermit with my family here, it's the first time I've met people that I actually have these great relationships with, which is a lot of fun.
48:03
Overwhelmingly, what I heard that they are looking for right now as a proxy for execution, which we discussed earlier as this critical element is, to what degree did a business change how it operates in 2022?
48:16
Now interestingly, the reason that they framed it that way is because they said there are a lot of businesses that they see that didn't change say in 2021, and take advantage of what was happening in capital markets.
48:26
In their eyes it's the adaptation to the 2021 market and then the adaptation to the 2022 market that they're really, really looking for to see where operational excellence, so to speak, is in place.
48:40
A company that's flattened both and hasn't changed how it's operated is, I mean, it might be the best, most boring company in the world.
48:50
It might be, but it's probably not.
48:52
It's probably a company that doesn't really know how to operate in a world that's anything other than the blindness that it is on.
48:58
I thought that was a fascinating insights.
49:01
It's not growth, it's not AR. It's, "Did you change?
49:04
And how well did you change?"
49:05
Jim O'Shaughnessy: You know what's so interesting, is like you are absolutely right.
49:10
But as one who was for most of my career, which is changing rather radically now as with this new O'Shaughnessy Ventures that I've started and it's not going to be just an investment company, we have different verticals.
49:22
We have one for film, everything's branded Infinite by the way, because of Infinite Loops.
49:26
So we have Infinite Films, Infinite Stories, Infinite Adventures, that's venture infinite.
49:33
We're going to like seed the most talented people we can find.
49:36
Now because we have the variance amplifier that is the internet, we'll find them because absolutely geography no longer is our foe.
49:45
What we're looking for is that bright shiny penny.
49:50
People don't quite understand how revolutionary this actually is.
49:54
My ability to be sitting in Greenwich, Connecticut and yet find somebody in Bangladesh who is amazingly talented and give that person money, that historically never happened.
50:09
I recorded with David Senra of the Founder's Podcast and one of the insights that I can't remember which of us did it, because we're so both so pithy, but the idea was the internet is removing all of your excuses for not trying. It's removing them all.
50:27
Like, "Oh, I'm stuck here in failed country that most people don't know about."
50:33
"I'm stuck here in Bangladesh and I can't do anything." Wrong. "I'm afraid I'll fail." Wrong.
50:42
That's an opportunity and the ability to pivot.
50:47
You now have all these tools man, and you are running out of excuses to not try.
50:55
But one of the things that I was going to say, so most of my life was as a strategic public market investor.
51:06
For a strategic public market investor, we have to sell strategy as opposed to tactics, which you're talking about, which is what on the private side I would demand of any of the companies that I invest in. Hello.
51:22
You can't run this business the way it was run in 2021.
51:25
You got to have a new plan for 2022.
51:29
Essentially with "This too shall pass" as our guideline. Strategically, right?
51:36
Strategically, if you are able to arbitrage human nature using a series of empirically guided algorithmic selection models, most of this comes out in the wash, right? Because guess what?
51:51
Those companies that fail to adapt also have, they have quant markers and you can see, you don't even have to say they didn't adapt.
52:00
You don't even have to offer a reason because they all have an underlying...
52:06
Let's apply it to an x-ray or an MRI.
52:09
Like if you take an MRI of a very sick person, there's a pattern there, right?
52:16
You can see it and you can understand that.
52:19
Same thing in strategic, all of our algorithms basically, we don't use just one, we use them in concert with one another.
52:27
I can tell you there are algorithms there that are identifying strictly which companies we do not want to buy under any circumstances because they all have these quant markers of what we're talking about.
52:43
They're baked into them is, how did they get that way?
52:47
Their fucking CEO or their board or whoever was decision making didn't adapt and so they died.
52:54
That would be my first kind of remark. It's interesting.
52:58
It depends on where you are on the spectrum, right? Mr. and Mrs. and Ms.
53:01
and whomever who is listening to this right now, if you are 40 and you are investing in public markets for your retirement and you anticipate you're going to retire when you're 65, pay no attention to this.
53:16
Keep putting your money into the S and P 500 or whatever index or hopefully O'Shaughnessy asset management, although you can now customize those portfolios too. Talking about...
53:28
Jeremiah Lowin: I've heard about that. I've heard about that.
53:30
Jim, why don't you tell me us a little bit about custom indexing.
53:31
Jim O'Shaughnessy: I very rarely use this platform to talk my book and I'm going to use it today to talk my book a little bit because it's leading to my next question, but the other observation I wanted to say here is absolutely, if you are at the level of trying to run a successful startup, if you can't pivot, you're dead.
53:49
Life is movement, stasis is death. Jeremiah Lowin: Yeah.
53:53
Jim O'Shaughnessy: You mentioned chaos and entropy earlier.
53:59
Entropy's always going to win at least probably while I'm still alive.
54:02
Hopefully, when you and your kids are running things, you'll figure out a way to reverse entropy.
54:08
But my group is going to figure that out.
54:12
But the point is, you've got stasis is death.
54:16
If you don't understand that and you don't flow with what's actually happening, you're dead. You're dead.
54:24
Jeremiah Lowin: It's true.
54:25
One of the things that did surprise me, except from what we were just talking about, but sort of surprised me in my own education, running a company was not just a degree to which what you just said is true and of makes sense to me philosophically, there's always things always changing, there's this and that.
54:40
But I think we have literally rebuilt the company more or less corresponding to each funding ground that we've done.
54:49
The company has to be rebuilt, it operates differently, it serves its customers in a different way.
54:54
It has new employees and a much wider employee base.
54:56
It has to be a different company than it was before.
54:59
Jim O'Shaughnessy: Totally. I just have to jump in.
55:02
The reason I think you're succeeding for all these great reasons that you've just been giving and founders listen up man, because if you're not doing this, you're going to have a problem.
55:12
But also the way you just said that, what did you invoke?
55:16
You evoke the Ship of Theseus problem, right? Jeremiah Lowin: Yeah.
55:19
Jim O'Shaughnessy: Which is right, you're continually rebuilding.
55:23
It's still Prefect but it's the different. It's different.
55:28
You've rebuilt it, right?
55:29
The other thing you're invoking is one of the pre-Socratic philosophers, Heraclitus said, "The man cannot step in the same river twice because the river is different as is the man." Jeremiah Lowin: Yeah.
55:43
I only know about this of course because it was a major plot point in WandaVision, which features both superheroes and explosives.
55:48
But you're absolutely right, we have to reinvent the company because the company is not the thing it does on any day.
55:54
The company is the value that it delivers to its customers.
55:58
We have to continuously refine and reimagine what that is. It's true.
56:02
We swap out many, many things along the way.
56:05
Our company today is almost no resemblance to the company two years ago. It's the same mission.
56:09
Actually, funny enough, I keep this on my desk as a reminder.
56:12
I tore out the staple the last time when I held it up.
56:14
This is the original deck.
56:15
This is the piece of paper that I...
56:16
Jim O'Shaughnessy: I remember that. Jeremiah Lowin: ...
56:17
I probably sat with you and showed you this?
56:19
Jim O'Shaughnessy: You did.
56:20
Jeremiah Lowin: I did print this out. Yeah.
56:22
This is the actual paper from five years ago.
56:23
I could pitch the company from here. Most of it is the same.
56:25
The mission, the software, everything.
56:27
But we've gone through such wild, wild changes of how we do it and what we bring to bear and the internal product.
56:33
I mean, something we want to talk about of things I've learned also is the degree to which the company itself is a product.
56:41
Prefect makes two products.
56:42
Prefect makes software, which is for our customers and has a customer base of thousands and thousands of people, and we make Prefect Technologies Inc.
56:50
Which has a customer base of about 85 people who are its employees.
56:54
It has to be an amazing product that they choose every single day to partake in.
57:00
That's where all of my energy is now.
57:01
Those are real product meetings.
57:03
This isn't just semantics.
57:04
There are real product meetings, real objectives, real features we ship, so to speak, to make sure that, that's the best product it can be and it changes. It goes to versions.
57:13
Jim O'Shaughnessy: One of my recent obsessions has been, so most of the personality driven tests are bullshit. They don't replicate.
57:22
The one that does replicate is the Big Five.
57:26
I started looking at my own Big Five because somebody was talking about it and I talked to David about it as well, Senra.
57:36
You know me, I can't help myself.
57:39
I create an algorithm of everything, right?
57:43
It's like, because I'm so fucking lazy.
57:46
It's like the lazy person, the German general, right?
57:50
His perfect combination was lazy and clever.
57:53
He said, appoint to the highest levels of the staff because lazy, clever people are going to find the easiest way to do anything.
58:03
My mom, so when I was a kid, think 1960s kitchen, right?
58:10
I was born in 1960, so I must have been nine or 10.
58:13
It would've been 1970, but it was still the shade carpet, all of the horrors of that period.
58:19
Anyway, my mom said, “Hey Jim, would you get me something out of that cabinet?
58:19
” That cabinet was missing a knob, right?
58:29
My mom and dad would always bend down and open the cabinet with their hand from the bottom, right?
58:38
She tells me as a kid, "Go get that thing”.
58:40
She goes, “you walk over there and you do it with your told.
58:40
” She goes, "You are going to make everything the easiest way to do anything because..."
58:51
She said, because I was clever.
58:53
I later admitted it was because I was lazy that that laziness and slothful nature had really been inspired me to figure that out.
59:02
But the idea that if you can't pivot the algorithm that I brought up the Big Five, it's like, wouldn't it be really interesting if I could find algorithmically a way to score founders who are coming to me and asking me to fund their company?
59:24
I haven't started down this path because I've been busy with a bunch of other stuff, but I still think I'm probably wrong.
59:32
That's the other thing by the way, that I love hearing from you quoting Jed McKenna here.
59:40
"The smartest thing I ever did was stop being smart." Right?
59:43
I am the first to say I'm probably wrong because that frees you to explore various things, right?
59:52
Because if you are wrong, you are much quicker to say, "Oh, I'm wrong."
59:57
And pivot over to something else.
59:59
If you're not wrong, you're still on this creative obsessive journey trying to figure things out and back to the Big Five, right?
1:00:08
I think, and you're the tech genius here, although I do have a competitor for you that we're going to talk about next, that might be one up on you in this particular thing. I don't know.
1:00:21
Anyway, openness to new experience is got to be like it's unless, which I could be totally wrong, but that's got to be somewhere on the spectrum of founders.
1:00:36
Because to be open to new experiences is to be understand how failable we all are, how failable ideas are.
1:00:44
You mentioned, I love that, right?
1:00:46
An idea, Yung said two things that really made what your point was really nice.
1:00:52
The first was, "People don't have ideas. Ideas have people."
1:00:57
In my mind, the ideas exist outside of me.
1:01:01
I realize that they have me.
1:01:03
When you put it in that regard, you don't say, "This is a hill I'm going to die on.
1:01:08
I take the George Patton thing.
1:01:10
No, I'm going to let the other poor dumb bastards die on there.
1:01:14
Yung said that, and then Yung said, "You are not what you say you will do. You are what you do." Hello?
1:01:24
I mean, that this is even a conversation kind of baffles me, honestly.
1:01:30
Because again, my grandchildren, now I'm going to go down to Maeve and Langdon, my two youngest grandchildren right now, who are five and two.
1:01:40
I bet even from my two year old Langdon, when she acquires the ability to spar conversationally with me, she'd go, "Well, duh, papa." You know? "Duh, hello."
1:01:53
Maeve was becoming such a beauty, was Patrick's youngest and she could pull it off in the mean girl fight.
1:01:59
Although she's not a mean girl, she's a very nice girl.
1:02:01
I'm referring to the movie Mean Girls.
1:02:04
She could like, "As if, papa." Right? It's like, "Hello?"
1:02:09
That theme that we're still trying to get people to rock these very simple ideas amazes me.
1:02:16
Jeremiah Lowin: Well, you have to be a little bit insane to want to actually go out and start a company and necessarily be saying that, "I don't quite agree with the way things are done and I'm going to do it this other way."
1:02:31
And go against what is by definition all conventional wisdom.
1:02:36
I mean we were joking earlier, but it's worth repeating that as a front of a company, I'm the only person that didn't have to actually interview. There was no vetting.
1:02:44
Jim O'Shaughnessy: I love that. I love that.
1:02:46
Jeremiah Lowin: I just decide on a piece of paper.
1:02:47
We should never, ever presume that just because someone has started a company that they actually...
1:02:52
I mean the evidence is actually that someone is more likely to start a company because they couldn't work somewhere else than because they could.
1:03:02
We have to be incredibly careful about this.
1:03:04
One of the things that I worry about is, look, it's all well and good for me to come on.
1:03:07
I spoke to you about this a couple years ago, and I say, "This is how I think about things.
1:03:11
I don't know if it's right or wrong, I have to believe it's right." Right?
1:03:13
because I'm going to make this.
1:03:14
Jim O'Shaughnessy: Exactly Jeremiah Lowin: In line with this, so I spent a lot of time worrying about that, thinking about it.
1:03:18
I don't know that it's right.
1:03:19
I don't necessarily know how to translate it into something that someone else could pick up and follow.
1:03:24
I know that if we go out and we get a line of folks who have started companies and when we collect all of our opinions, I'll be shocked if any of us agree on anything, right?
1:03:35
It's very difficult to think about what would those commonalities be that you could pick up on.
1:03:40
In fact, I'm reminded, one of our earliest investors gave me this line, which I've never forgotten, and I think about a lot actually.
1:03:48
He said, "Look, we're going to invest in Prefect.
1:03:51
We all know that you're going to make mistakes.
1:03:55
What I'm investing in are original mistakes."
1:03:58
Jim O'Shaughnessy: I love that. Jeremiah Lowin: Right?
1:04:01
If you go out and all you do is make the same mistakes that I could have gotten in some other company, then there's no ROI on this investment.
1:04:07
I can go, I can buy those mistakes somewhere else.
1:04:10
I think that in some ways this pursuit of original mistakes is sort what one is doing as a startup founder in a constructive way, right?
1:04:19
They can't stay mistakes. We got to fix them. But we never go out.
1:04:23
Or at least I think it's an incredibly, it's a huge red flag if someone goes out believing that they're going to do things right.
1:04:31
It is one of the most dangerous attitudes.
1:04:34
We instead our skill set and here using our to mean my collective company, right?
1:04:41
Our skillset is to go out encounter a mistake, which we must, because we're trying to do something new.
1:04:48
We must not do it perfectly at first.
1:04:50
It's ludicrous to think we would do it perfectly.
1:04:52
Second, that is how we do it and encounter that mistake and fix it and move on to the next one.
1:04:59
That's the skillset over and over and over that I think we need to optimize for.
1:05:02
Jim O'Shaughnessy: Again, like I said it earlier, founders, you're getting a masterclass here because the circumstances have changed and you're watching in real time how, obviously I'm talking my book here so let's be very clear.
1:05:20
I have a rather large investment in Prefect and obviously I wanted to succeed and I'm going to have blind spots because of that investment, so let you know. I know.
1:05:32
Jeremiah Lowin: Disclaimer, disclaimer.
1:05:33
Jim O'Shaughnessy: Disclaimer.
1:05:34
But these are real disclaimers, right? They're true.
1:05:36
I bought in, I still believe.
1:05:38
Jeremiah Lowin: Well, the fact that we have a joint financial incentive doesn't preclude the fact that we are very interested or it may in fact enhance the fact that we are interested in maximizing the outcome.
1:05:50
Jim O'Shaughnessy: Totally.
1:05:51
Jeremiah Lowin: What I'm speaking here is the strategy that we have put in place to achieve that. Is it wrong?
1:05:57
Of course, it's going to be wrong because we're making a mistake somewhere in what I've said, right? I guarantee it.
1:06:02
Something I've said here is very wrong. Maybe all of it.
1:06:04
Jim O'Shaughnessy: No, that's my job. That's my job. That's my job.
1:06:08
Jeremiah Lowin: That's fine.
1:06:09
You tell me that it's all wrong.
1:06:10
Jim O'Shaughnessy: No, no, no. It's my job to be wrong.
1:06:12
Jeremiah Lowin: Oh, that's fine. You'd be wrong.
1:06:14
Jim O'Shaughnessy: Don't try to take that from me.
1:06:15
Jeremiah Lowin: You be wrong as much as you want.
1:06:17
I'll be wrong as much as I want, as long as we don't make the same mistakes. Everybody can be happy.
1:06:21
Jim O'Shaughnessy: Hello.
1:06:22
Wisdom of crowds, cognitive diversity. Of course.
1:06:26
If we got to be wrong in different ways or we're fucked, right? Jeremiah Lowin: Yeah.
1:06:30
Jim O'Shaughnessy: This the...
1:06:31
Jeremiah Lowin: Sorry, sorry, go ahead. Go ahead.
1:06:34
Jim O'Shaughnessy: Yeah, no, no, no.
1:06:35
This is such a great point and it's so important, which is, again, I'm going to hack the book that I keep hounding everyone to read The Beginning of Infinity by David Deutsch.
1:06:49
If you haven't read it yet, you got to read it.
1:06:52
Because it is a blueprint for, it sounds like you've read Deutsch and you've downloaded his way of looking at the world, which is being a rational optimist is actually pretty optimal because it doesn't mean that you're, "Oh, there's never ever going to be a mistake."
1:07:12
Quite the opposite in fact.
1:07:14
Fallibilism is the idea that, "Oh, we're going to be really wrong."
1:07:20
There's a great quote, I can't remember whose it is, people can look it up, but it's like, Fire was the greatest discovery of humans. Right?
1:07:30
We have the legends of Prometheus bringing fire and being punished severely for it.
1:07:36
Obviously that didn't happen, but we did discover fire.
1:07:39
But guess what, any profound innovation...
1:07:42
This is a great segue to what we're going to talk about next.
1:07:45
Any profound innovation can be used by we humans for good or bad.
1:07:52
Fire created modern humans because it allowed us to cook our food, which allowed our brains to get much, much bigger. Like "Yay, fire."
1:08:02
But the George Carlin joke about flamethrowers, he's like, "I always thought, I always had it about flamethrowers, that somebody somewhere at some time had the idea that I'd really like to set those people over there on fire but I don't have any means to do that."
1:08:21
But the point is, fire is also incredibly destructive. What did we invent?
1:08:27
We invented the fire department, the fire alarm, the fire drill, the fire extinguisher, right?
1:08:35
What you're describing is perfectly aligned with what Deutsche's propounding in The Beginning of Infinity, which is of course, we're going to make mistakes.
1:08:46
That's how you make progress.
1:08:48
It's the other side that is in favor of stasis that uses the precautionary principle to say, "Oh, no, we better not try that because you know better the devil you know."
1:08:59
If you're looking for an example of why there was no progress for a thousand years in human history, it's because we were ruled by the precautionary principle where people like you and me were burned at the stake because “what are those crazy people saying? ” You mentioned crazy.
1:09:16
I've started four companies, well now five and so I'm batshit crazy. But like, okay. I'll own that. That's fine.
1:09:25
I like being that way, which leads me into getting your opinion on my newest investment.
1:09:31
I announced a few days ago that I had invested in August in a company called Stability AI, which is a true or open source AI, artificial intelligence company.
1:09:47
Imagine placid Lake Michigan and then imagine a thermonuclear bomb being lobbed into the center of Lake Michigan and the waves and tsunami that would actually happen from that happening.
1:10:00
Well, the founder, Emad Mostaque, decided in August that he was going to release the entire Stable Diffusion model with weights.
1:10:12
People who don't know about AI don't understand the importance of weights because if you have the weights, super creative people can use it in all sorts of different ways.
1:10:22
And what we saw after he released it, we saw a tsunami of creativity that came through use cases that he himself told me I wouldn't have thought about.
1:10:37
That's part of my thing that to be so presumptuous to think that you know what the best use cases are going to be around anything, it's like the Spanish Inquisition, man. Are you an idiot?
1:10:51
You're just demonstrating how stupid you are if you think you know everything that humanity is capable of. So I saw this.
1:11:01
I'd already been doing due diligence on him for months and I saw this and then met him. He was the real deal.
1:11:08
So I made an investment and he asked me to be executive chair of the board of Stability AI.
1:11:14
You were one of the first people I called because a lot of people might not know, but it's one of your passions.
1:11:21
You have a deeply abiding passion in AI, which you proved to me when you told me that you read Disney's AI team wanting to get the hair to flow more naturally on their mermaids, right?
1:11:34
Jeremiah Lowin: Oh, absolutely.
1:11:35
We have to footnote that.
1:11:36
The Disney research team, even their YouTube channel, forget the papers, just watch the stuff they put out. It's incredible. Jim O'Shaughnessy: Yeah.
1:11:45
So I came to you and I'm like, "Am I just batshit crazy here?"
1:11:49
You had some great feedback for me, which I really appreciate.
1:11:56
I don't want to ask the question in a leading way.
1:11:58
I've declared what side I'm on here.
1:12:01
I have grandchildren, soon to be more than three, which delights me.
1:12:11
If something that I can do that makes certain that they don't grow up in a world that is controlled by a panopticon with only a few people controlling what people can and can't have access to, then I'm going to do that thing even if I fail, which chances are I will.
1:12:33
So first off, I'd love your take because this is all I've been doing for the last month, doing this deep dive on AI, and yeah, generative art is really cool, but that isn't where AI is going to change things.
1:12:49
AI is, in my opinion, going to change things most profoundly in two things that I'll just say right now, education and medical research and the innovations in vaccines all the way through.
1:13:04
But tell me why I'm wrong.
1:13:07
Jeremiah Lowin: It's hard to tell you why you're wrong when I agree with you.
1:13:10
So instead, how about this.
1:13:11
First of all, because, I might be wrong, I'm pretty sure this is the first recording you've done since you went public with your [inaudible].
1:13:19
Jim O'Shaughnessy: It is.
1:13:20
Jeremiah Lowin: On behalf of everybody, let me congratulate you because it's- Jim O'Shaughnessy: Thank you. Jeremiah Lowin: ... incredibly exciting.
1:13:26
It's really extraordinary and you're going to sit in this front row seat at the helm of what I strongly believe will be one of the most important AI companies, period, just because of the impact here.
1:13:36
When we look at not just the model, the stability, excuse me, the Stable Diffusion model and the image generating, when we look at actually what happened here, the bomb that was dropped on Lake Michigan, so to speak, was arguably not so much the open sourcing, though that was necessary, but the fact that the model fits on consumer hardware.
1:13:57
I think that that accessibility of this model is what has led to this explosion of interest.
1:14:04
You shared with me some adoption numbers, which I don't know if I can share or not, but they're extraordinary.
1:14:09
It's like nothing I've ever heard of.
1:14:11
Jim O'Shaughnessy: I'll use one that we can use and we went from zero users to our DreamWorks API to a million in 27 days.
1:14:19
Jeremiah Lowin: It's incredible.
1:14:20
It's an incredible adoption story. And why is it there?
1:14:26
Interestingly, because that's through your API, that's not on the consumer accessible version of this, which is the fully open source one, but I think it's a proxy for how this has captured that mindshare so successfully.
1:14:38
The way that it's doing it, and it's something that is very near to two intersecting interests of mine, which is the AI technology, but also the open source strategy, and that's something where I spend an awful lot of my energy is thinking about open source strategy for a variety of different companies, but it's that intersection and tapping into a true flywheel there.
1:14:57
That is so extraordinary.
1:14:58
Some of the things that have been done, I want to talk about them, but since I'm don't want to presume that everyone knows how the image generating systems work in general, it's going to be a little bit tricky, but for example, there's one called inversion, which instead of the normal pipeline of type words get image, it lets you provide representative images and then use those to bias them off.
1:15:18
For example, I could take a couple pictures of you and all of a sudden I could say, "Jim in the style of van Gogh."
1:15:22
Even if you're not in the data set, it doesn't know who you are, although I guess it probably does, I'm going to try that, it will start outputting based on what I fed it, which it will infer the style.
1:15:33
So I think that the magic in this, and I do think it's magic.
1:15:36
Once upon a time, the very first time I was on Patrick's podcast, that was by the way, when someone really was scraping the bottom of the barrel just trying to get anyone to show up.
1:15:49
I think I'm on episode 18 or 19 on Invest Like the Best.
1:15:50
Jim O'Shaughnessy: No, no.
1:15:51
The scraping of that barrel, young Jeremiah, was when he had his father on.
1:15:55
Jeremiah Lowin: Oh, well, you're right. You're right. What was that?
1:15:58
That was probably episode 17 and then I was on episode 18.
1:16:00
Jim O'Shaughnessy: You were 18.
1:16:02
Jeremiah Lowin: Right after that, yeah.
1:16:03
Jim O'Shaughnessy: Yeah, I think so. You were on the upcurve.
1:16:07
Jeremiah Lowin: Yeah, sort of rebound. We're talking about AI.
1:16:10
We're talking about the role of AI in finance as a matter of fact.
1:16:12
He asked me what's the most magical AI that I'm aware of.
1:16:17
At the time, honestly, it caught me off guard and I was grasping for it and the definition that I came to was, what's the technology that lets me interact with a machine solving a problem for me? That was what I went to.
1:16:27
So my answer, which I stand by, but it sounds ridiculous in the context of [inaudible], my answer was Google Maps because it was the most magic encapsulation of a technology.
1:16:38
This was what, in 2016 or '17 we were having this conversation.
1:16:41
So the most magic encapsulation of technology.
1:16:43
Since then, I got a car that drives itself.
1:16:45
That was the top of my list for a while.
1:16:47
And today, it's unquestionably the ability to take words and see a cohesive image.
1:16:55
That magic is very, very, very real because well, what have we been talking about the entire podcast? We are storytellers. We are people. We use words.
1:17:05
We use language to tell stories.
1:17:06
Our whole history is sitting around a fire and trying to communicate our ideas and our dreams to each other.
1:17:12
This is the first time that anyone can go and do that and say, "I want to see Jim O'Shaughnessy not in the style of van Gogh," which I think will become the cliche “Hello World” of any of this technology, but I want to see.
1:17:24
I can't be in the room with you when you're at the head of the conference room table in the boardroom making these decisions.
1:17:31
I can imagine it, but now I have a way to actually maybe show somebody what that might be.
1:17:37
And of course, then we start to get into the dangerous side of this where I could show that to somebody and they might believe it and I could show them something else and they might believe it.
1:17:46
We get into with great power comes great responsibility situation very, very quickly.
1:17:53
If we believe that with great power, we automatically can assume no responsibility, we end up in a world without a lot of amazing things.
1:18:03
Jim O'Shaughnessy: Let me just interject there because that's very important.
1:18:07
I could not agree with you more.
1:18:08
Two of our initiatives that will probably have already been announced when this drops because I already have you planned to drop at the last episode of October, the first one is we are going to be announcing on October 17th in the Bay area, our opening “Hello World”.
1:18:25
We're announcing a competition where we're putting a bounty, $100,000 from Stability AI and because I believe so passionately in this, $100, 000 is going to come from O'Shaughnessy Ventures.
1:18:37
So we're going to put a $200,000 bounty on whoever can give us the best deep fake detection software.
1:18:47
What we're going to do with that software if it passes our own internal hurdle rate, which is pretty high, we don't want people thinking...
1:18:56
You know how I feel about value at risk models. I hate them.
1:18:59
I hate them because they lull people into this complacency about, but the model said it was only a 13% loss. They're crazy.
1:19:08
So we have a pretty high hurdle rate, which is like say if it can detect currently and it's going to iteratively get better and better.
1:19:17
But let's say if right now it can identify 90% of deep fakes, not only are we going to release it, we're going to give it away for free.
1:19:25
We are going to make sure it's on every platform.
1:19:28
We are going to send it to every media outlet.
1:19:30
We are going to make certain that everybody...
1:19:34
You talk about consumer, right?
1:19:36
Emad, the founder, told me when we were discussing this, the thing that he's most proud of is the compression algorithm.
1:19:44
I said, "But did you model that on the HBO show Silicon Valley?" and he started laughing.
1:19:52
He's like, "No, but I think I might start saying that I did." Jeremiah Lowin: Yeah.
1:19:56
Jim O'Shaughnessy: The compression algorithm is the magic here because we took like 200 terabytes and it's now two gigabytes and we can put it here.
1:20:05
I'm holding up an iPhone, guys.
1:20:08
So that's the first initiative and I want that released everywhere, because can you imagine when the first person not related to Stability AI in any way runs that software and finds that something that they had loved is a deep fake? Jeremiah Lowin: Is fake.
1:20:26
Jim O'Shaughnessy: That is going to go so viral that...
1:20:29
By the way, again, my soapbox here, but we've had deep fakes forever, guys.
1:20:35
There are deep fakes hanging in the Louvre, in the Metropolitan Museum of Art, everywhere.
1:20:41
There's a wonderful story about a guy who wanted to prove that the so-called art world authorities were idiots.
1:20:48
He was a master sculptor and he created a sculpture in the style of the ancient Romans and he buried it in a farmer's turnip patch.
1:20:58
So the farmer was pulling up the turnips and up comes this sculpture of a Venus and the sculpture is missing its hands, its feet, and its head while all of France decides that this sculpture was from when the Romans tried to invade Gaul.
1:21:17
The president of France at the time declares it a national treasure.
1:21:22
All of the art historians say this is the authentic thing and this guy comes forward.
1:21:25
No one's ever heard of him.
1:21:27
He's like, "No, you're wrong. I did that."
1:21:29
And they're all like, "You did not."
1:21:31
Then he produces the hand, the head, and the feet.
1:21:36
When asked about why he did it, he said, "I wanted to demonstrate to the world that all of these officious people here in my country, France, they assert things like they're true. They're not true. Let's be clear."
1:21:52
That's the first initiative.
1:21:53
The second initiative is we want to be the leading voice on the ethics of open AI, because you're absolutely right, it can be used by what I think is a minority of humanity, but they exist and trying to say they don't exist is to blind yourself voluntarily.
1:22:13
There's a lot of people who want to do really sad, horrible shit with AI and we're going to do everything we can to make sure they can't do that.
1:22:24
Will we miss a few things?
1:22:25
I'm sure we will, but we are going to have in place, we want to have a council that is open to everyone, experts in AI, academics, religious scholars. We want everyone there.
1:22:39
In fact, I'm advocating for a wiki that anyone can come and add their two cents.
1:22:45
It'll be moderated, of course, because people are going to put stupid things on there and we want this to be a cumulatively good wiki.
1:22:54
So completely agree with you and we not only understand that, we want to be the leaders.
1:23:01
Of course, we'll fall down.
1:23:02
We'll make mistakes, but ours will be out in the open.
1:23:06
Jeremiah Lowin: Yeah, original mistakes.
1:23:08
I would love to talk with you more about this idea of AI forensics, which it's going to be a field unto itself.
1:23:17
There's no question of digging in and understanding and tracing the lineage, if you will, of these things.
1:23:24
One of the things that's so fascinating to me as someone who's been in this field now for coming on two decades is to see the degree to which progress is slight and iterative, but brilliant in delivering these step functions where...
1:23:37
In a different forum, we'll actually explore the model.
1:23:40
We'll explain how it differs from others, and the differences are small, but the outcome can be profound.
1:23:45
But one of the things that comes to mind as you're describing the battle between generating the artwork and detecting the fact that it was generated is that there was an entire class of models that were actually built on this premise.
1:23:58
They were called GANs, or I used to call them GANs until I actually met someone else more attached to it than me, G-A-N, generative adversarial network.
1:24:04
Jim O'Shaughnessy: Yeah, no.
1:24:05
I've learned all about GANs. Jeremiah Lowin: Yeah.
1:24:07
It's fascinating because the premise of those models was to take the mechanism you just described and capitalize on it.
1:24:15
The model consists of two components, one that generates an image or whatever it's trying to generate, and another one that guesses if an AI generated that image.
1:24:25
As the guesser becomes better, the generator needs to in turn become better at degenerating.
1:24:30
They basically have this little race to ultimately produce high-quality samples from this probability disfunction.
1:24:38
It certainly didn't fail.
1:24:41
It produced incredible, incredible results, but the reason it was ultimately superseded by diffusion models and other analogous models was because ultimately, that closed window, the expressiveness of the generator, the expressiveness of what's called the discriminator led to a very collapsed probability distribution. Jim O'Shaughnessy: Yes.
1:25:00
Jeremiah Lowin: The thing that is magical is that I could type pretty much anything into a Stable Diffusion model and get it out, whereas with these GANs, they were very, very, very tied into the center of a distribution of whatever they were trained on and generalization became very difficult because the whole strategy to train them was about fooling each other on a very specific outcome.
1:25:22
Whereas the diffusion models are by definition, designed to explore a state space.
1:25:27
I don't know if this is the form we want to go into.
1:25:30
There's extensions there of ask it to generate something that's symmetrical and it usually doesn't do a great job because symmetry is difficult to express within the form of this model.
1:25:40
So that's an obvious extension that I'm sure has a lot of folks looking at it.
1:25:43
But for me, it's fascinating to see, okay, so we have this little insight that's led to this class of models and the next insight is around the corner.
1:25:51
And because it's an open model, that next insight might come from anywhere and anyone. Jim O'Shaughnessy: Yeah.
1:25:59
Jeremiah Lowin: This is based on just seeing open source software in general.
1:26:04
I have a belief that just being open will not be sufficient to automatically elevate someone to really get in and understand.
1:26:11
I think that there's a history of what's been done or maybe better said, what's failed that folks who have not studied how these models have been historically built and developed will just end up repeating other's mistakes.
1:26:24
So I think the goal of making original mistakes here is more accessible through the openness, but I don't think it's an automatic win, in other words.
1:26:30
I think it gives more [inaudible] for humanity to advance this thing, but it's still going to take a lot of respect for the nature of the model to actually achieve better.
1:26:39
Jim O'Shaughnessy: Totally agree.
1:26:41
We could do just an entire episode on just AI and maybe will someday, that would be cool, but I took some of the feedback that you gave me to our offsite that I attended in London and I was happy to see that everybody was already thinking that way.
1:26:59
The idea that it's not sufficient, you're absolutely correct.
1:27:04
So we have to build out by making a lot of mistakes, which we will, but there's a trade- off in having access to what is now, I think, 40 plus thousand coders working on the open source model on our behalf, we think of it that way, and saying that that's sufficient.
1:27:25
Well, it's sufficient to build a really cool model when you open source it because again, we're back to cognitive diversity, different skill sets.
1:27:34
That's the only way you build the human colossus or the AI colossus is to have a whole bunch of different points of view, but not sufficient. You still need more.
1:27:47
I think we have a pretty good business model and I think that it has the chance of really doing well.
1:27:55
But like you, it's like that's for right now.
1:27:59
If we are going out and making sales calls a year from now, and God forbid we're in a depression, we got to have a different plan for getting people interested.
1:28:08
I love this idea of AI forensics though.
1:28:12
I was talking to a friend who I made through Twitter, by the way.
1:28:17
Let me do a pitch here, guys.
1:28:19
Again, being on Twitter, not sufficient.
1:28:22
What you need to do is use that to meet interesting, fascinating, brilliant people in the real world.
1:28:30
That's the value of that platform.
1:28:33
It is not for shit-posting as much as I do.
1:28:37
It's not for that, it's for actually meeting clever, interesting people.
1:28:42
I was talking to one who I'm a friend of now and he's deep domain knowledge on AI.
1:28:49
So he was sending me ideas and the first one he sent me was having to do with medical research.
1:28:54
He goes, "You got to set this AI loose on every published study."
1:29:01
He goes, "because I'm going to make a guess here that at least a third of them are either phony data, which the researchers actually faked to make their point."
1:29:12
One thinks of Ancel Keys at the University of Minnesota who basically faked all of the data saying that fat was our heart disease.
1:29:20
It was going to kill you if you ate fat and was fake. The data was all faked.
1:29:26
So what my friend said was, because he's also deeply into the science and scientific research as well, really smart guy, but he's like, "I can't wait until you guys go and audit all of the earlier scientific...
1:29:41
” The way I just think about that is how cool is that idea? Jeremiah Lowin: Yeah.
1:29:52
Possibilities are endless.
1:29:55
When I think of a medical opportunity that harnesses that's something that most people can now see and touch through Stable Diffusion is giving medical students examples of what they're looking for, especially in the case of more rare...
1:30:10
Well, I was going to say, "How many times have you gone to your doctor?"
1:30:10
and they say like, "Oh, this looks like something, but I need to call a specialist because I haven't seen this before."
1:30:18
Specialists exist for a reason because they pursue the very specific knowledge that can be difficult to acquire because we don't see all diseases and we don't see all MRIs and all brain scans.
1:30:29
One of the most interesting things that I've seen recently, just to take what you said and turn it on its head about a positive, like an advancement of scientific research is the ability to generate fake, but representative lung scans or brain scans or any scan that doctor or a medical student could use to learn and test themselves on, for example, identifying things.
1:30:53
Then of course, we can start talking about AIs that can do it themselves, but that's the second stage.
1:30:57
We don't need to jump to the end to still deliver this value.
1:31:01
That's the sort of stuff where I start to get really, really excited when people are like, "What's the point of typing in a box and getting an image?"
1:31:08
The image is useless, it's what do you do with the image.
1:31:11
There are so many things that we do in the world, which are ultimately dependent on examples and being able to generate those examples, it's very expensive and time-consuming.
1:31:20
And in medicine, there are ethical boundaries.
1:31:24
We can't just go around cutting people open to take pictures of them so that we can learn what their lungs look like. We can't do that.
1:31:32
Jim O'Shaughnessy: Exactly. Yeah.
1:31:34
We're very simpatico on this because you highlighted two things that I deeply believe in and I think that people are legitimately afraid of.
1:31:43
The first thing is we are envisioning AI at least for the coming decade, maybe two decades.
1:31:50
It's just a tool for humans to use and interact with.
1:31:56
Humans are not going to be replaced here.
1:31:58
If anything, they're going be made 10X more effective.
1:32:03
Let's continue with your example of the doctors. Exactly right.
1:32:07
Ethics prevent us from doing...
1:32:10
Unfortunately, some incredible medical findings came out of the monsters and horrors of World War II where they did unspeakable things to human beings, and yet some advances came out of that.
1:32:26
Let's hope that we have gotten past that horrible portion of humanity, but that we still recognize we don't have to do horrible things to living creatures.
1:32:39
We can create them here in AI as examples, and so I'm totally on board with you there.
1:32:45
And want to underline, guys, whether you're an artist or whether you're a doctor or whether you're in my field, this is a tool and it's not here to replace you.
1:32:57
It's here to make you vastly more effective and this is just one of the places that it will.
1:33:05
I could talk because I did a very deep dive in a very short period of time and so I'm still really dangerous because if I know- Jeremiah Lowin: You're drinking from the fire hose.
1:33:13
Jim O'Shaughnessy: Exactly, but also, I know just enough about the adversarial models and the history and how GPT-3 actually got so smart or appearing to be smart to humans.
1:33:29
I could talk about it all day, but I can't because I've got to go.
1:33:34
We've already done two hours. Can you believe it? I can't believe it.
1:33:37
As always, I love talking to you and I bet I'm going to make you incept two new things into the world.
1:33:44
I now have actually a microphone that you can speak into that I will make available to you. I'm kidding. I'm kidding.
1:33:53
We're imagining that you can incept the entire world's population and you can speak two prompts into the... Look at me.
1:34:03
I'm even using the language of it now.
1:34:04
You've got two ideas that you can incept in every human.
1:34:07
They're going to wake up the next day and think it was their idea and they're going to do it. Go.
1:34:12
Jeremiah Lowin: Jim, I was thinking about this question all last night because I knew you were going to ask it to me. You know what?
1:34:19
I couldn't think of a single thing that I would say because every single thing that came to mind, I was like, "That's too stupid.
1:34:24
There's got to be something better.
1:34:26
There's got to be something better.
1:34:27
There's got to be something better."
1:34:28
Last time you asked me this question, I talked about folks just giving other people a chance to speak, so I'm going to tell you what.
1:34:33
All I'm going to do is double down on what we just talked about.
1:34:36
Number one, give AI a chance.
1:34:38
It's not artificial, it's not intelligence, just an equation. It's going to be fine. That's the number one.
1:34:42
And number two is to make original mistakes.
1:34:45
Jim O'Shaughnessy: I love that. I love both of those. As always, thank you.
1:34:50
Jeremiah Lowin: Thank you.