0:00
[Music] Welcome Reed. It's great to be here.
Search this video
0:08
Thanks so much for coming on. Yeah.
0:08
So AI is just moving at an incredible speed.
0:12
Every single month AI gets better and better than more and more humans and more and more jobs, skilled jobs.
0:16
Uh you've previously described AI as the cognitive industrial revolution.
0:20
And previously you've talked about AI being a co-pilot for humanity as opposed to a replacement of humanity.
0:27
Do you still think that's true as the pace of acceleration grows and grows and grows?
0:34
Well, um, part of when I make predictions, and predictions are always a great way to seem foolish in the future.
0:38
So, I will engage in such foolishness with some vigor here, uh, is I'm trying to predict the trajectory.
0:46
So, I knew the speed of acceleration.
0:46
So, I knew the speed of acceleration. I knew this could be the acceleration of increasing uh the cognitive capabilities of the machines and what's happening and roughly seeing you know even though we're looking through a glass darkly
0:58
what do we see now will there be a bunch of replacement for human work by AI yes no question for example kinds of things we see line of sight too is any kind of job that that is being done where a human is actually kind of following a script whereas is like being a robot. That's actually in fact like customer
1:18
That's actually in fact like customer service.
1:20
Those jobs are going going to go almost completely away in the shape that they are now.
1:26
And it won't be the oh but I'd really like to talk to a human because you know I just kind of feeling very retro that like humans will be crafting some of the experience and giving feedback but will not actually in fact be driving the vast majority of that.
1:41
And there's actually broader swaths than most people think about when we're trying to have human beings act like robots.
1:47
like robots. Now, what about the fact that when we're doing kind of the co-pilot or the human amplification or or AI as amplification intelligence, what that will play out to be and uh I think that the reason why we know that that will be at least a fairly long
2:05
distance in the future and there is some science fiction over well maybe AI is just better at every single thing that we care about as tasks than we do and maybe it will get there and to say it categorically can't is actually in fact silly. But when you look at it and I
2:19
But when you look at it and I look at and part of the reason I encourage everyone to actually go play with it and I do it myself and I ask myself the question I see where is this thing you know better than me today where it will be better than me tomorrow where it'll be better than me next year.
2:36
I actually see a lot of room for how we are super adaptable and we can bring a bunch of things that GBD2, GBD3, 3. 5, four, 4.
2:46
5, you know, 01, deepseek, etc.
2:49
It's like, no, actually, in fact, there's a bunch of stuff that we still bring that's actually in fact very useful.
2:54
And it's not just, oh, there's 10 human beings who will have co-pilots.
2:59
It's that we will all have these co-pilots.
3:01
And I think part of that is it will require changes from us.
3:05
I don't mean changes like genetics or changes like you know oh we have to do 50 push-ups you know 50 mental push-ups every every day.
3:13
It's changes of oh what I'm special about is not the fact that I have a really good memory.
3:19
What I'm special about is my way of thinking about this in a sideways format asking like today we are much better at asking questions than the AI is.
3:28
By the way, they'll get better at asking questions or but like what are the questions? What are the prompts? And this is not new.
3:37
So, for example, back in the printing press, people said this is going to be the destruction of human cognition because there will be this information that anyone's going to get without the the the value of trained human expertise, right?
3:49
And and now you can read this book and you can formulate your own impression.
3:53
Disaster for human cognition, disaster for wisdom. Clearly not, right?
3:58
not, right? But how we get there and how we change that no longer is memory the most important thing but like question asking or framing is the most important thing and that's the journey we're on which is of course both a wondrous and
4:14
sometimes scary journey and this taps in a little bit to your idea that you recently wrote an article about homochi where it's not always we talk a lot about how us humans create the technology we create the tools but just as we impact them the technology and tools similly impact us. Yes. So perhaps Yes.
4:29
So perhaps you could talk a bit more about that.
4:32
So the first place I really started talking about is I gave a speech uh for Bolognia commencement and I said well look and it's kind of quoting off McLuhan and other things but we shape our tools and our tools shape us and that's what happens.
4:45
happens. So for example before we had glasses you know there were some people who saw the world very clearly and other people said no no no the nature of the world is fuzzy right it's almost like we're impression and it's only when we get the technology that begins to change that we literally can see the world
5:02
because of technology now for example when you get kind of philosophical you say well all right so AI will start changing our sense of epistemology because it starts rendering things and we have aspects of that today you can go to an expert and the expert says, "Well, this is how black holes work." And
5:18
And you're like, "Okay, I trust you.
5:21
You seem to talk to a bunch of other people who are smart about black holes and you all think it works this way with like time compression and other kind like there's a bunch of weird things around black holes."
5:31
You go fine, I don't know, but I take that take I take it as a point of view.
5:36
Now, most of those things tend to be things that are somewhat outside of our experience.
5:41
somewhat outside of our experience. Now sometimes they get odd like germ dis you know germ theories of diseases or bio you know like genetics and other kind of things okay I take the I take the expertise and how this and how this operates
5:54
but now that will get much broader so an AI says well actually in fact you know your theory of European history you know maybe it's the following versus the thing you think or that kind of thing and and all of a sudden you're like well okay what how much do I have to need to cross-check this epistemology. What is a
6:13
What is a set of things that I need to do in order to get there?
6:16
Do I trust this as a artifact the way that I would trust maybe a consensus human expert expertise?
6:23
Do I trust it in this in this vector?
6:26
And what are the things that led to the production the uh the creation of this AI?
6:30
Do I trust the AI on my side?
6:33
What are the things that need to be in order to have that be the case?
6:38
That's like now a new level of what does epistemology mean?
6:41
And that's part of the future that we're already heading towards.
6:45
You say, "Oh, is it totally new?" Well, you trust Google. It's accelerating. It's accelerating. Yeah.
6:52
Do you feel like you know, you understand yourself truly read how the AI works?
6:55
Like how does open AI work?
6:57
Um, anyone who says that they understand it deeply is either deliluding you or themselves, which I try to do neither. Thank you.
7:07
Now, uh, now that being said, we do have some broad theories that are at least partially correct.
7:14
So, the fact that it learns a very complicated intellectual space and then tries to uh through because because by the way, part of how these systems are trained is through a lot of a lot of specific uh reinforcement learning with human feedback.
7:31
they go okay off this kind of training program it's trying to be helpful in these kinds of ways like so you know in simple terms it ingests a trillion and a half words right that's a lot of words more words than any of us are ever going to read even if we had you know dozens or hundreds of lifetimes and then through this a training program both algorithms and then human feedback then says here's how I think I can be most helpful to you.
8:01
And that's the reason why, you know, we have these both amazing things with these systems, but also sometimes you're like, "Oh, that's a real error."
8:11
Because you think it's trying to be helpful to you, but actually, in fact, it's spinning you a story of complete BS while it's trying to do that.
8:20
And by the way, that's a fundamental part of the technology, even as it tries to correct it to have many less quote unquote hallucinations.
8:30
But it's but that's fundamentally the way it works.
8:32
Now, by the way, people say, "Well, that is the same or different how human beings work."
8:35
And the answer is a little of the both.
8:36
Like like for example, we go we think we're complete reasoning machines.
8:40
But actually, in fact, we approximate reasoning machines.
8:45
It approximates reasoning machines too, but just in a different way.
8:49
And that different way gets back to your question which is well we will learn different patterns of what is actually in fact thinking is actually in fact sapiens by what we construct and how we how we understand AI as both similar and different to what we do.
9:10
Reed let's give the listeners some tangible examples of what the future might look like in an AI world.
9:14
I I'd love if you're okay to do it.
9:16
Maybe just to paint one very positive vision of the future created by a one believable but very negative sort of dystopian future. Yes.
9:25
And then perhaps a third version which is when you kind of are whimsical and just imagining some of the more outlandish things that might happen.
9:33
Love to hear some of that. So positive.
9:41
We create something that's between tools and creatures.
9:45
We don't know if it'll be creatures or not, but at least it'll be tools that whenever we engage anything that we're doing in our life, we have one or more agents that are helping us with it.
9:56
And it could be like, oh, I'm I'm engaging with a medical thing.
9:59
And so, for example, a very simple thing for everyone to envision right now is we know how to create a medical assistant that's better than today's average doctor that's 247 available on your smartphone.
10:14
And we and and to run it for under call it 3 to 5 pounds per hour.
10:19
Now, by the way, that doesn't eliminate doctors like that utopian.
10:24
Of course, it's utopian because, by the way, that means that 11:00 p. m.
10:28
at night, you're going, "How important is this?"
10:30
You can take a picture of it. You can talk to it.
10:33
It can say, "Hey, um, you know, also, by the way, you're unclear if you're if you're if you're completely cognitively functioning yourself.
10:40
Call your friend, talk to your friend, see how that's, you know, get another person to look at this and go to the hospital right now or no, no, you're fine.
10:47
You actually should probably go to bed, schedule an appointment, you know, etc."
10:50
and and to be better than the average GP.
10:52
And by the way, part of what we see in this utopia is is when you walk in to see your GP, then then it says, "Hey, would you mind if I talk to your GP in front of you for a moment?"
11:04
It's like, "Okay, so Reed had the following symptoms.
11:07
Here's what I was thinking.
11:07
Here's what I think is most likely.
11:09
Here's the reason why I think these two blood tests might be particularly useful.
11:12
Here's something you might talk to Reed about.
11:15
Reed doesn't seem to be listening to me about maybe he should change his diet a little bit this way.
11:18
Maybe you should talk to him about that, etc. , etc.
11:19
This is all like, hey, it's helping me navigate my life better, and medical is one, but education would be another, legal would be another, work would be another, you know, kind of how do I how do I um, you know, kind of be a better friend and and and interact with my friends? That could be another.
11:36
And these are all, you know, kind of a set of agents that are helping us, each of us individually, and also us together navigate a better life, right? Utopia.
11:50
By the way, I think a utopia is really possible.
11:52
Hence publishing super agency and everything else.
11:54
And what that you get to the future that you're trying to get to by steering towards that future, not by trying to eliminate only the futures that are bad.
12:02
It's the mistake with criticism.
12:04
It's like you have to be going what is good, let's go towards that.
12:08
And people say, well, I'd rather just stay right now.
12:10
I think we're at the best and the most utopic thing of human history in all time, and it can't get better than this.
12:16
That's foolish on lots of levels.
12:18
among the fact that people will be building technology and it will be evolving.
12:21
So you have to go with it. So that's one.
12:22
So dystopia is one of the mistakes that people think about technology is they say well some technology is centralizing and other technology is not centralizing.
12:33
All technology is centralizing.
12:35
Our ability to make it a decentralizing power is what we choose in terms of how we integrate it in our societ society, how we integrate it in our our way of life.
12:49
All technology can empower autotocracies, right? It does. It can.
12:54
They say, "What about encryption?"
12:54
It's like, well, say I'm running a society where I say encryption is illegal.
12:57
When we see your signal of encryption, like the robots come to get you, right?
13:02
Or the police come to get you.
13:05
Like encryption is something we encourage and enable to make a distributed set of society where you say well actually in fact this is a private communication between the two of us or the three of us or between you and your bank or between you and your doctor.
13:19
That's when we we allow that to be an encryption.
13:22
We we we deliberately don't have it in clear text.
13:26
A authoritarian society could say no encryption.
13:29
Everything must be in clear text us.
13:31
clear text us. and and and we are using technology to make the government run as an autocracy against you know call it go all the way back to you know ancient societies where it's like everyone you are all surfs to our small autotocracy
13:50
that's a dystopia that can happen with AI it's one of the reasons why I think it's so important for western democracies to say no no no we want to build this within a society of individual freedoms individ individual rights, a collective health of society together. It's totally possible to say
14:05
It's totally possible to say we're going to have AI propping up North Korea and all societies will be like North Korea, a disaster, a Chernobyl, worse than Chernobyl.
14:17
So that's that's a dystopia.
14:20
Now, some people go, well, there's even a further one of that because it'll be the AI, not the human beings, that are the superpowered, you know, masters in these circumstances.
14:29
I actually think that's more straightforward to avoid than the humans doing it.
14:36
It's like am I wor more more worried about humans with AI or more worried about like Terminators?
14:41
I'm more worried about humans with AI.
14:44
Now again, it's not to say zero worry, but you say which is the worry you're trying to steer against, right?
14:50
And I think one of the things that's frequently kind of uh misleading about the the robots are coming to get us is like no, no, no.
14:59
you're planning on trying to stop the robots coming to get us.
15:01
It's the humans with the robots coming to get us that you need to to worry about that.
15:04
That's what I would say is the dystopic like okay I really want to try to avoid that and trying to steer towards a good future is trying to avoid that.
15:13
So and I do remember the third the outlandish and I'm going to say three things uh each in different vectors of human life and all three of them could be right and could be outlandish.
15:24
be outlandish. So say we say hey actually in fact we do want to to continue to have democracy as the fundamental way that we're governed but the way we're going to do democracy is not like kind of how we do this fully today which is that we have you know like we elect representatives and those
15:44
representatives do the information processing that is you know well should we you know increase you know um you know these taxes or decrease these taxes or allow like this kind of zoning of land use or or or deal with conflicts between our different human interests by the following policy that representatives do. Say for example, we
16:04
Say for example, we do a version of direct democracy, but it's now what my friend who's a professor at at at the media lab says is the avatar state, which each of us has a civic avatar and they all go into agent space and they all argue it through and we get all the policies that come out through that and they're explaining to us how they got there.
16:28
But but we've all proxied our Democratic vote to our avatar who is talking to us about like and we say, "Well, but I actually don't really like that."
16:37
It's like, "Yeah, I tried to argue against that."
16:39
Like, "Well, did you argue this way? Did you argue this way?"
16:41
And the avatar is evolving, but like our democratic governance is through the avatar state. That could be weird.
16:49
could have dystopic elements, could have utopic elements, could get a lot more nuanced and and detailed and all of our voices being there versus the weirdness of having like well we all voted on having that human being or that couple set of human beings making all the proxy decisions for us which is always a much less integration.
17:10
That would be one weird state somewhat utopic or potentially utopic but also weird and unusual and feels very alien from where we are today.
17:20
Another one would be is like well what's the future of human work?
17:22
Future human work might be as opposed to well actually in fact I fill in all this data entry I write these reports.
17:29
It may be that because we have these AI uh tools, creatures that that know a lot of like a lot more broad knowledge than we do.
17:42
They're already today GBD4 has a much broader knowledge set than any human being on the planet.
17:47
He said, well, and it's getting better and it's getting better.
17:51
It's like, well, what's my role?
17:52
It knows how to do venture investing better than me.
17:55
it might know how to to to model like what could be future products better than me.
18:00
future products better than me. So what's my role given that that's what I think I think you know has made me interesting and unique and special in in you know the in the world and it's like well actually in fact my role is to try
18:14
to think about the really odd contrarian cases the like the kind of being in the extreme case like Einstein realizing that there is no ether it's like what if there is no ether what if it's totally different and that's the thing that we
18:31
start doing as human beings is our our role is not to say yeah yeah yeah we know how to predict you know kind of a revenue curve a a scenario planning analysis of what happens if the these products service are working they're not working people are responding positively
18:47
not responding positively all of that happens with AI mine is like all of our jobs are to be more okay I'm trying to think about the weird corner cases my crazy full experiments yes and that's what more and more of our work looks like, right? As opposed to No, no. We're As opposed to No, no.
19:03
We're trying to do like what is an entertaining podcast?
19:07
What is something that would cause someone to both be entertained and think a new big thought what we're doing here? No, no, no.
19:12
It's what would be a complete reinvention of a podcast? Like what would that be?
19:16
What would be the kind of thing?
19:18
Because otherwise it'd be you know the read AI and that giant ideas AI is just having this kind of conversation.
19:24
What what's our shaping of that?
19:25
That would be like another whimsical unusual outcome.
19:32
And then the last one is like what if for example part of the thing that's really fundamental and this is a little bit like inflection and pi and you know this kind of agent startup that Mustafa Sleimman and I did and is still going is
19:48
to say the thing that AI will make its most serious contributions to our well-beings is not necessarily in tasks and performance and enablement but actually in fact having an AI AI that goes with us through our entire life from we from when when we're like one
20:08
and helps us all become much more emotionally mature that for example as we exalt in kind of every religious tradition a certain amount of wisdom a certain amount of compassion say for example like part of the the agent that starts working with you from your
20:25
earliest thing is to say how do how how how do I help you navigate your life in which you're your better self and that part of what we're all suddenly all much better at is being kind, compassionate, love that engaging. Yeah. Like that's a Yeah.
20:39
Like that's a possible outcome that is like most people are think oh machines they're like machines Terminator you know etc.
20:49
As opposed to like Todoro like like it's a little fuzzy rabbit a demon or like a little putter on your shoulder.
20:55
But but like think of it as like say it's a cute fuzzy rabbit that's actually in fact kind of going hey here's how you can have a much warmer wonderful interaction with your friends around you.
21:06
Here's some things to think about.
21:08
Here's like maybe you say it this way versus that way.
21:11
Maybe this is more who you want to be in how you're doing it.
21:14
And maybe that's part of that.
21:16
How would you design design one of those?
21:18
Well, we we did that a little bit with PI, which is all of that human feedback learning where we trained our trainers was to say um help guide PI in how it can be much more uh positively re reinforcing.
21:32
Not by just saying, "Oh, you're so smart.
21:38
Oh, you're so wonderful."
21:38
We don't necessarily respond that well to that, but kind of saying, "Hey, that was an interesting idea.
21:43
When'd you come up with that?"
21:44
Oh, and and and what's the next thing that you thought of after you thought of that?
21:48
Or when you had that that interaction with Sarah or Bob, like how did it make you feel?
21:54
And and and what would you have done that you would have made you made you and them feel better?
22:00
Like now that you're thinking about and maybe that's a thing you should do next time. And pi is all of that.
22:06
And it doesn't try to distract you.
22:07
you. Like for example, if you go to pi p sorry the pun is deliberate like personal intelligence but pi if you go to pi and say you're my best friend he says no no no no no your friends are your other people I'm your AI companion hey let's talk about your friends have you seen them recently is there
22:23
something you do because it's having this theory of humanity elevating humanity but part of elevating humanity is connecting with other human beings and have having like good like like where we're making ourselves better by doing it that's the kind of thing that we're trying to do. And you've talked
22:38
And you've talked there about tools, you talked about agents, but you've also mentioned maybe a few times creatures. Yes.
22:43
Are we talking there about AI robots, something else?
22:48
What are these creatures?
22:48
Well, so part of the thing that's beusing about AI is it's the closest we're getting other than obviously like you know our genetic evolution of various animals that we've been involved with.
23:00
It's the closest we have to to making a new form of creature and how much it's a tool and how much it's a creature is actually still kind of a TBD unknown.
23:14
You can call it like Schroinger's robot.
23:17
Like we don't know the exact measurement of how this plays out.
23:19
And the creature tends to make people a little nervous because you're like creature like what is it like with regard to me?
23:27
Is it friendly and cuddly? Is it on my side? Is it on its own side?
23:34
And and does something dangerous result from that?
23:39
And that's part of what like gets into all the a lot of science fiction speculation about this stuff.
23:44
And by the way, part of the the reason I kind of object to the scientific speculation about this is because you're like, well, it's easy to go like Hollywood tends to always go, you know, man versus machine, you know, and the human versus the evil machine.
24:00
And you're like, well, but I can see a lot of different machines that that are actually making us much like homochnne, like much better human beings, much better human societies.
24:08
And matter of fact, that's what a lot of what machines have been so far.
24:12
And even though this one is a new cognitive technical capabilities, new intelligence capabilities, and also may have certain forms of autonomy, which as they get you get to more of a creature.
24:25
So we like we know that our beloved black Labrador is a creature and not just a tool and we understand that and we know how that goes but that's also partially because we're in control of the entire environment.
24:40
Well, does this creature now have some joint control on the whole environment and how does that work out?
24:47
and and that's that spectrum between tool and creature and it's part of the universe that we're shaping in the future and that's where you begin to get to the science fiction of this is very interesting.
24:58
So, and by the way, we have no idea how to create creatures really.
25:04
We know we would want attributes.
25:04
We know that we would want them to be like like looking out for our best interest as individual in society.
25:11
We know that we'd want that as part of what they're doing.
25:15
Well, what does that look like?
25:17
And that's part of the reason why like when Mustafa and Karen and I were all brainstorming this for inflection and for PI, we're like, well, let's just start with a chatbot and let's just start with a chatbot that's kind and compassionate in its interactions.
25:28
Let's see where that goes. Right, Reed?
25:30
I remember listening to you about two years ago talk about your side project where we're using the OpenAI uh image generator to create art and it was this kind of side hustle I think to try to force yourself to to live in the future and we had Mike Maples the investor on the podcast the other day and he was talking about the best founders living in the future.
25:50
What are some of the habits that you've formed over the years that have forced yourself to continue to live in the future and to stay not just right at the cutting edge but actually creating the cutting edge of technology decade after decade? There's probably three.
26:05
So one is always be focused on where is you know in the kind of the ice hockey thing where is the puck going to not just where the puck is.
26:16
It's always interesting to know where the puck is, where it's come from, but it's always where's it going? Where's it going? Where's it going? Where's it going?
26:22
And so, like, for example, that was part of the reason why like I as soon as I saw GBD2 to GBD3, I was like, okay, I know a bunch of where this is going, right?
26:33
bunch of where this is going, right? and not everything not detail but it's like the the real AI revolution is uh scale compute a tied to learning systems and because it's learning new things like like to your much earlier question about like well how much do we know how much they're operating we don't because
26:50
they're learning systems and they're enormously complicated learning systems and then they say well how can we guarantee that it behaves in certain ways it's like well it's actually a scale complex learning systems it's very unknown how to guarantee things so Okay, we'll go well we can highly predict things and within scope of highly predict how do we make the right things happen? So one where is it going? Two
27:10
So one where is it going?
27:10
Two when you engage with other people and talk with them you're like what do you see that I don't see and what do you see that you think I may be getting wrong and that's how you engage in those conversations.
27:24
That's how I engage in those conversations with Mike.
27:26
That's how I engage in those conversations with other people.
27:28
So when for example you know u couple months ago I was sitting with Ethan Mullik who is a professor at Wharton and he's one of what I think is the leading people for thinking about like what is currently possible to do with these AI systems because he hasn't only taken the red pill he's bathing in the red pill to use the the kind of matrix analysing how these multimodal AI AIS currently work and are already really great.
28:00
Which is to say if you for example today wanted to set up a baby monitor, you could put a camera on, put it on the baby monitor, say here are the things that I would want to be interrupted for for is you know is the kid doing okay, not okay, other kinds of things and what is a set of things and here's why I would want to be notified.
28:21
Call it every 30 minutes like hey everything's fine.
28:25
seems to be sleeping just fine, seems to be playing or burbling along just fine.
28:30
And by the way, if you're noticing anything, I would love to know it.
28:32
Like, are you noticing language development?
28:35
Are you noticing other kinds of things?
28:37
You could just do that today.
28:37
You set up the camera, you pipe it through a multimodal AI, you give it the prompts, instant super intelligent baby monitor, right, for doing these kinds of things.
28:48
and and it was like right I hadn't been thinking about all the different ways that a camera today could be directed to do something that would be really really useful because this is adding intelligence to everything and then that
29:01
adds a whole new suite of things of oh well I could use the camera not just with like for example a baby monitor but a construction project or I'm really curious about you know animals in the following way or I'm really curious about etc. This is all just a question
29:14
This is all just a question of how much kind of creativity do we set up.
29:19
Do you think it's fair to say that the internet added connection to everything and AI will add intelligence to everything? Yes. Right.
29:24
like literally everything like and and not just everything that has a concurrent computational surface, not just phones, not just PCs, but like microphones and clocks and pictures and you know light switches and you know light system, lighting system, ambient intelligence everywhere everywhere.
29:49
How should our listeners not allow themselves to be disrupted in their careers beyond recognition?
29:54
How do people stay on top of this and where should they be hanging out?
29:58
How do they bathe in a red pill?
29:59
Just as you know sometimes when you look at the tech community on on X, it feels like sometimes those people are using tools and using AI 6 to 12 months ahead of the rest of the population who might be left behind.
30:10
Well, part of the kind of central prescription from Super Agency and Impromptu kind of last two books was start using it.
30:19
And frankly, if you're not using it enough that you're finding some that you're not finding something that's seriously useful to you, you're not trying hard enough. Full stop. It's there right now.
30:32
Every single person on the planet.
30:32
And that's that's that's a, you know, kind of addictum.
30:37
Now, frequently happens with very smart people is they try it for something, they go, well, okay, u like, write me an essay on AI in the future.
30:46
And it goes, it's not a very good essay. It's not very good.
30:47
It's like, well, did you think that was the only possible prompt by which that could happen?
30:52
Did you not think about other ways that you could make that prompt happen in really interesting ways?
30:56
And that's the keep trying.
30:58
Now, it doesn't mean it's useful everything.
30:59
So, take a for example a personal thing.
31:01
When I first got access to GB4, I said, well, okay, how good will this be in replacing me as an investor?
31:06
So, I said, okay, how would Reed Hoffman, there's a bunch of information about me, make money by investing in AI?
31:11
And it gave me what I think of as the business school professor who's super smart and super articulate and understands nothing about venture capital, right?
31:19
So it was like you identify the large total addressable markets.
31:24
You'd think about what the product and service substitutions would be.
31:27
Then you think about what the market opportunity would look like against current and future competition of that and then you would identify the teams and you go find those teams.
31:32
You invest in them and you would do that.
31:36
It sounds very rational and that's not how venture capital works for making much money. It's like oh okay.
31:39
much money. It's like oh okay. uh then someone might conclude this is useless for me as an investor nothing to do with me as investor I'm completely safe in one of the things I'm doing as an investor say well but what are the other tasks I'm doing and like one of them
31:52
frequently is due diligence so I said okay like I just fit in the PowerPoint deck fit in the you know fill in the business plan and say what's a due diligence plan and by the way it it gives you a pretty good baseline all of the baseline rational things to think about due diligence doesn't actually get the prioritiz ization, right? Probably
32:10
Probably will eventually like item number three is the really important one, not items number one and two, but it gives you a fairly complete thing.
32:19
Like I'd look at it and go, "Oh yeah, 6 it would have occurred to me two days from now, but is useful to add into the overall plan to execute against right now."
32:29
And so there is a thing where it's useful to me in my day job of being, you know, investor at Greylock to look at this and say, okay, I should seriously consider these things in deciding whether or not to do this investment.
32:45
And so it it is a tool already today that can help me as an investor.
32:50
And that's what I mean for everyone about whatever is the thing you care about.
32:53
You what you say is what I care about is my is how I'm being a parent and relating my ch child.
32:57
It can be useful to that.
32:59
I know of cases with pi and others where literally you put it down, you have it listening and it says, "Oh, read when you said that, did you really mean this or did you mean this other thing?"
33:10
And you go, "Oh, I meant this other thing.
33:12
Oh, that's much more helpful for communication."
33:13
Useful in communications.
33:15
It could be the medical stuff that we've already talked about.
33:20
Like literally anything that you seriously care about, it can be useful for today.
33:24
for today. you wrote a fairly iconic book called the startup of view which which I read in my 20s and was very helpful and kind of the core thesis there was you know we should treat people should treat their lives and careers as startups so sort of take measured risk against hypothesis you know move fast um how would you update the wisdom of that book for today's
33:43
environment or is it still as true as ever well it's definitely still true um eg you have to approach kind of your life and your work life as an entrepreneur uh approaches a company a product a market because entrepreneurs are about the creation of the new when things are changing and things are changing more and more and so therefore all of those attributes are still there. Now part of you know one of the things
34:05
Now part of you know one of the things that Ben and I really emphasized was the importance of a learning clock and that becomes even more so for example when we wrote the startup view it wasn't as important to be thinking about every individual having their own strategy about how they integrate with technology you could decide I'm really close to technology I'm really far from
34:25
technology didn't really matter now being closer to technology really matters so if you're not having a how does AI help me learn and navigate right so we you know one of the metaphors that we use in super agency is it's anformational GPS like navigating information space if you don't have that as a theory you're going to be well behind so that would be like an update and like for example when we were doing
34:50
the the 10-year review of startup review and asking this exact question because you'd always be thinking what's new what should be learning we're like well we didn't cover anything in the gig economy we wouldn't really know the gig economy was fully coming and yet for many people the gig economy whe whether or not you're a student whether or not you know you're you're between gigs whether or
35:10
not that's the way that you balance things in your life the gig economy is an ongoing and important thing so how what does the startup review mean for the gig economy and so those those but it's gig economy but also what does it mean for how you're learning constantly and how you're using technology in order to do that while you're still forming team relationships your network is your team in everything you're doing. Clear. Clear. Clear. Shifting gears a bit.
35:35
Want to shift gears to bit about Reed Hoffman himself.
35:39
So you've had a incredibly successful and interesting life.
35:41
You grew up in the Bay Area. You studied philosophy.
35:44
Well, you went to Stanford, studied symbolic systems, then went on to Oxford for a degree in philosophy.
35:49
From there you were part of the early PayPal crew.
35:51
Um then founded LinkedIn which you sold to Microsoft for I think 26 billion. That's great.
35:56
And now you sit on the board of Microsoft.
35:59
Extraordinarily successful.
35:59
What do you think it is about Reed Hoffman as a person that has allowed you to reach these levels of success in a very, you know, and and and succeed over time?
36:06
So, um, part of the reason I wrote the startup review is I gave the commencement speech to my high school and I was thinking, what would I say to a bunch of 17 year olds, most of whom are not planning on becoming technology entrepreneurs or technology investors?
36:23
They have a wide variety.
36:23
So, a lot of that is that encapsulation of that learning of that approach to life. Yeah.
36:28
that approach life like how do you take a smart risk that could be worth something?
36:32
How do you how do you navigate that smart risk both in changing your plans, mitigating it, dealing with if it doesn't work, uh capitalizing upon it if it does, um how do you approach having people around you as being part of, you know, the team you in terms of how you're doing it?
36:49
And that's part of the startup you was a deliberate pun on the individual and plural, both singular and plural of you, right?
36:57
to the startup of you, you know, broadly.
37:00
And so, so for me, like one of the things that I think I was just fortunate to have gone to Stanford and thought about this is like when I think about where my attributes, I'm like one of the best person to throw at an unknown problem, right?
37:13
Like I'm a really really good person for doing that.
37:16
I'm a less good person for I study it for 10,000 hours and I get it just perfectly right.
37:22
Like for example, the classic kind of what what is endemic within Japanese culture and so amazing is they go whatever we're doing, we're making Hungarian goulash, we're going to work on it again and again and again and get it just right.
37:34
That's other people more than me.
37:37
I I tended the hey, I tried it, I'm moving on to the next thing, right?
37:40
And so and and by the way, understanding those attributes of yourself and and initially I was going, well, I'm I'm good because I can do this academic stuff and maybe I should be a professor of philosophy.
37:49
And I was like, no, no, no.
37:50
What I'm really much better at is the unknowns or the unknown unknowns and the huh well okay how do we sort through this and then it was my reflections on that that caused me to actually write most of my books because they're kind of a sense of here is the world as I found it and here's the things I learned that I hope could be really useful to a large number of other people.
38:14
So did that theory of life sort of take shape originally at Stanford and Oxford and then coalesed during your time at PayPal and and with LinkedIn?
38:21
Well, it it was what do I want to do?
38:21
I want to try to help people think about the world that they're in where the most important thing of for really everyone even introverts and and monks and you know and other folks is the people around you. Right?
38:38
around you. Right? It's very different patterns between an extrovert who wants to go to a party and be the center of a party and someone who wants to like make art mostly in in in quiet in their own private space but still people matter and all that and it's like okay what is that and so when I first thought about it I said okay symbolic systems for artificial intelligence philosophy maybe
39:00
that's the creation of ideas and doing that and then I was like well actually in fact this pattern of academia isn't the kind of idea creation I want Plus part of the thing is I want to be it's almost like for example academics do scholarly work that 20 to 50 people read and and like I would much rather be doing uh you know mass market books that affect lots of people and that would be
39:24
that was the thing that clued me into oh maybe this is the wrong path maybe a different path it was should I be an author or actually in fact should I conceptualize new software because software is like the homoche thesis is part of how we like we build into our registration of the world like epistemology and AI like what is the epistemology of AI what is the epistemology of a human using AI what is
39:46
epistemology of a group of humans using AI what is the epistemology of a group of human of humans using a group of AIs like that kind of thing like oh that changes the way we understand each other and we understand the world that's what I want that's the medium that I want to be working in that's where I that's that that's where everything kind of comes together yeah you're also part of the PayPal mafia. Arguably, no other group
40:09
Arguably, no other group of people has affected the world more in the last couple years.
40:13
Um, especially the last one. Yes.
40:15
And what what do you think uh were the traits that made that group so extraordinarily successful?
40:20
And do you think this group of people now has too much power?
40:26
So um the the first question is much easier which is it was a group of people who believe very strongly in the individual's ability to make a massive difference in the world to be just incredibly ambitious about what you do.
40:43
how you assemble the tools of of a group of people working together in a company.
40:48
Um, of technology that could make a difference and and and kind of uh make such a difference that it that that it makes a massive difference not just to individuals lives or company's lives but industries and societies and that everybody like it was this kind of this this whole group of people who came together with that kind of intensity with different points of view and different values and doing that.
41:10
different values and doing that. But but that as a mechanism and so then when it sold to eBay relatively early, everyone went off and did different things like they started YouTube and they started LinkedIn and and like just a set of different companies and they invested in you know kind of different companies that they thought were really
41:28
fundamental and in not just like the the kind of networks in the world but also things like Palunteer and other kinds of areas for doing this and that was I think and by the way I think that's a really important lesson for thinking about like the importance of networks, the importance of being bold, the importance of technology, the importance of of being just crazy ambitious in what things you could possibly do. And part
41:51
And part of the challenge I think we've had is that companies and governments are not the same thing.
41:57
Like for example, with a company taking a very bold bet where the company might just blow up and not work as long as it doesn't blow up in a way that leaves thousands of casualties.
42:10
You know, companies could do that.
42:10
You have to be, you know, but most of the time within these companies that's not the case.
42:15
Like even if PayPal grew up and then blew up and went away, like okay, there's a bunch of small businesses and people who are depending on it for payment services and they're scrambling.
42:24
it could really hurt a whole bunch of different businesses and commerce transactions.
42:27
But by the way, we could probably fill in, you know, things from other banks and other services in order to do that.
42:33
So, it's not it's not like society collapses because of it.
42:37
But part of the reason why governments are different than companies is if they blow up, the level of suffering and damage is huge.
42:43
This is the problem with, you know, Mao's cultural revolution.
42:48
This is the problem with the French Revolution.
42:49
This is the problem with, you know, the year zero.
42:53
There's a problem with it's like no no no it's very important to have stable societies and to be keeping key institutions and renovating them versus blowing them up because if it blows up it can create you know thousands to millions of deaths.
43:08
It can create the nonfunctioning of society.
43:10
It can create wars between societies.
43:13
So a stable social infrastructure is one of the things that we really need.
43:18
And so it's not so much that any particular individual say Elon has too much power as much as the way the power is being used is not understanding this difference between corporations and governments both domestically and internationally and the risks that being taken that could cause like frankly millions of deaths that were unnecessary that weren't just kind of the natural portion. That's awful, right?
43:45
And so that's an awful problem, but it's not a per se problem because an individual has too much power because human society works through individuals having power.
43:58
Whether for example, if there's too much power, there's no check on it.
43:59
Well, look, I think we we always try to have checks and balances in society.
44:05
I come from a a the US form of government is checks and balances are what should be there.
44:11
We are we have systematically broken some of those checks and balances. That's a bad thing.
44:16
the the executive branch and Doge is working ways that are illegal, right?
44:21
And sure, they may seem like small illegalnesses like, "Oh, you're supposed to give 60 days of notice and talk about it."
44:29
Well, actually, in fact, a process of 60 days of notice allows you to study what might be going wrong before you do something like firing all the nuclear safety inspectors, right?
44:40
and not figuring out not being able to rehire them suddenly because you're being you're being you know like dangerous in the way that you're you're being bold and it's like we like be more careful about that kind of thing.
44:54
That's that's the reason why you have these legal checks and balances and why it's important to do those kinds of legal checks and balances.
45:00
of legal checks and balances. So it's not but by the way we we like like ultimately when you live in a democracy you allow a vote to happen to put a president in power where that president could decide well um actually I'm going
45:15
to declare retroactively that January 6th people were not insurrectionists who were storming a building who who assaulted police officers enough that caused them to die that were explicitly talking about how they were hunting the vice president to try to kill the vice president. Like if we looked at that and
45:32
Like if we looked at that and saw that happening at Westminster, we'd be like, "Oh my god, things are really broken."
45:38
And no, no, we're going to say they were peaceful protesters and we're going to pardon all of them and we're not going to specifically carve out the ones who were actually doing also explicit physical violence.
45:47
That's a power that is granted to the president.
45:52
That is a completely legal thing to be doing.
45:54
That's what a enough Americans voted for.
45:58
You always have that form of power no matter what your checks and balances when you do.
46:02
It doesn't mean that power can't be misapplied.
46:06
And you know, I'm sad that that our society is doing that to itself and to the world.
46:11
But that is actually in fact a functioning of power and so therefore it's not too much power.
46:16
It's like well let let's get the the the democracy functioning in a healthy way. Read more positively.
46:24
We're going to end talking about the future again. Great.
46:26
I want to ask you, what was the dumbest idea anyone's ever pitched you that turned out in fact to be a giant idea? Oh.
46:33
Oh, there's a number of those.
46:33
Um, and they're not per se dumb as I just didn't see them.
46:39
So, um, I got pitched on what it was called odd blog, audio blogging, and obviously we had podcasting and so forth, but I was like, no, that's way too early for this. Not doesn't really work.
46:51
That turned into Twitter, right?
46:53
and is actually, you know, like, okay, a set of different like status updates and communicating what you're seeing in the world.
47:02
world. I got pitched on Pinterest several times which I didn't understand that there was a medium of of um you know kind of uh like like it is a it is a it is another functional medium just like photo sharing and everything else
47:18
to have a pin board of different things and that's actually a medium for executing that obviously got large and in terms of how it functioned another one which is probably maybe my deepest regret is stripe because I knew from PayPal how hard payment systems were. So
47:32
So when Patrick and John were saying, "Well, here's this new way we're doing it."
47:38
I'm like, "Oh, payments are really hard."
47:40
The the the path to building a global payment system just has tons of, you know, pitfalls and landmines and everything else.
47:53
And of course, part of the thing of the lesson there is always bet on just extraordinary talent, right? Just do that.
48:00
And so, you know, I could go on and on and on, many, many uh mistakes as an investor, but part of how I try to teach investment is if you're not making both so-called type one and type two errors, both false positives and false negatives, you're not playing aggressively enough.
48:17
So, the fact that I have errors doesn't make me go, "Oh, I suck." Right?
48:22
obviously not just LinkedIn and PayPal and Facebook and Airbnb and Zingga like like lots and lots of things in the success category too.
48:31
Um I don't go oh I'm a genius because I did all those or I'm a complete failure because I missed all of these.
48:37
It's the how am I evolving my theory of the game about which things you invest in and what things you do and how you make it work for creating great new businesses that help elevate humanity. Last question.
48:49
And this is something that Sam Orman asked you in 2016, so almost a decade on.
48:52
Is there something that you believe fundamentally that others just totally disagree with you on about the future?
49:02
Well, so the asterisk tends to be about which community, not necess because you say what does the entire world, 8 billion people disagree with you?
49:10
It's more likely to be where you're crazy, right?
49:14
But you get to an expert community and you say what is this expert community disagree with you on?
49:22
And so I guess what you know and part of the reason like publishing super agency there's a bunch of people who think it's just obvious that we're going to get to super intelligence and to me that's not that obvious.
49:39
I think we'll get to amazing savant intelligence for sure.
49:42
We already have it with GBD4 with 01 and so forth.
49:45
It's a it's a savant and I think the savant will get better.
49:49
Will it get to the no it's just more intelligent cognitively capable as in every vector it's not clear that we aren't at least some major inventions which may or may not happen away from that even though I believe we will have amazing increase of everything being cognitive.
50:08
And so there are people who think, "No, no, no.
50:10
You just don't get exponential curves.
50:11
You don't understand the math of exponential.
50:14
You look at this exponential curve.
50:15
That's an IQ exponential curve.
50:16
You just don't understand."
50:18
I'm like, "My bet is I do understand and you don't.
50:21
Even though you're super smart and the things you're making are amazing." Fantastic. Love that.
50:28
Reed, what a wonderful note to end.
50:29
Thank you so much for joining us on Giant Ideas. Always a pleasure.
50:31
I look forward to our next. Thanks so much, Reed. [Music]