0:01
Or even an intelligent layman trying to disambiguate what is the real deal versus the hype is virtually impossible.
Search this video
0:11
Once there is money involved, then there are hucksters.
0:13
They can just put out a press release saying, you know, we used a quantum computer to train a neural network.
0:20
And journalists and investors will eat that up with mustard.
0:23
And they won't ask like the very first question that any scientist would ask, which is well, did you get an improvement over a classical computer? Well, hello everyone.
0:38
It's Jim O'Shaughnessy with another infinite loops.
0:42
I must tell you that today's guest really daunts me.
0:45
Uh, my guest is Scott Aaronson.
0:48
A theoretical computer scientist with a chair, not the chair, at the University of Texas at Austin.
1:00
Winner of a plethora of awards of like just such a prestigious nature.
1:04
I was kind of like like which one like should I call it?
1:08
Which one would you call out?
1:10
Which which award that you've won has made you kind of like, uh, you know, gave you that little chill feeling?
1:14
I think that that that that would have to be like undergraduate teaching award that I got when when I was at MIT.
1:24
Oh, what a what a great answer. That I love that answer. Yeah.
1:27
You are the creator of the complexity zoo.
1:30
You are the proprietor of the Shtetl Optimized Blog.
1:36
You are were a visiting fellow at OpenAI, where you were consulting with them on a the theoretical foundations of AI safety.
1:45
There's just so much that I want to talk to you about, but as I said to you before we started to record, I am fascinated by your reading burden.
1:56
You you you read more than I do and I read a lot.
2:01
And and I just want to talk to you about that because I think we kind of are animated by the same thing which is we are so lucky to be alive at this point in the timeline.
2:14
And and like to to not read about everything that's going on seems to me to almost be a crime.
2:19
What what's your version and and have you augmented it?
2:24
Have you changed your reading since I since I read this piece? Yeah, yeah, no.
2:28
I mean I mean I mean reading has crept up on me over the years, right?
2:32
There's just it seems like there is more and more, you know, interesting stuff, you know, on on on substacks and magazines, you know, research papers that that that that that just to keep track of what is happening in all the areas that that that I care about is is basically a full-time job in itself. Right?
2:58
And so then and then and then that that's incredibly dangerous because I can feel like, "Okay, I I am, you know, I'm I'm I'm making progress on something that seems good to do by just reading and keeping track of everything."
3:12
But then the whole day passes by, you know, and now it's it's time to pick the kids up from school and deal with the kids and I haven't actually, you know, done any research.
3:24
I haven't written any papers.
3:26
I haven't even written any blog posts.
3:28
All I've done is I've read a bunch of stuff.
3:30
And you know, and this is all stuff that that that that tomorrow morning, you know, is is going to be refreshed.
3:36
There's going to be new stuff that I have to read, right?
3:41
And you know, and now, you know, on top of that are all the, you know, specific to me, you know, all the the emails that I get from, you know, because of my blog, for example, from, you know, random people around the world.
3:50
And you know, and like if there's some student who has questions, like I you know, my my inclination is always to help them if I can, right?
4:01
And you know, and then so it's so it's easy for your day to just be completely filled up by that stuff.
4:04
And you know, I I am not you know, I I I've never had any particular skill at like planning ahead or, you know, or planning my day or blocking out time or things like that, right?
4:18
You know, whatever success I've ever had has been, you know, in the teeth of just, you know, flailing around and just, you know, dealing with things randomly as they as they come or as they as they take over my life.
4:33
But you know, it seems like if I want to, you know, continue to do original things and you know, maybe I've got, you know, a few more decades, you know, or who knows how long in which to do them, then then then I then I do need to get much better at that.
4:48
You know, I was smiling and and nodding along as you were saying that because I find myself in a very similar predicament.
4:54
I I Good to know I'm not alone.
4:58
Once I once I start, I the the pull of the rabbit hole is so difficult for me and I find myself like looking up and thinking, "Holy how did it get to be 5:00?" Good.
5:11
And uh I have six grandchildren and I had the joy of having them all here for a while. Oh, wow.
5:16
And I love watching children because I I kind of think they're un unprogrammed, unprocessed humans.
5:22
And and by that I mean, like we can learn a lot in my opinion by watching kids, right?
5:27
Because they don't have any of the millions of social conditioning and everything that we adults have.
5:35
And it's just that their their voracious curiosity curiosity, their their willingness to try to explore everything.
5:43
And it it it Anyway, so I I love that they were here, but after after the fun was done, I realized that I I'd gotten so far behind on all of my reading. Yeah.
5:58
And and and and yet when I was chatting about it with a friend, I'm like, I wouldn't I wouldn't swap the time, right?
6:05
Because you can learn so much just by watching children in their nat- what I call humans in our natural state, right?
6:13
Like like be- before we decide on what's the correct answer machine they're trying to install in our head. Yeah. Yeah, no.
6:22
I mean I mean certainly another thing that is, you know, uh changed in my life has been, you know, having two kids, you know, I have an 11-year-old and a 7-year-old, right?
6:33
And so, you know, like like maybe even in the ideal case, I wouldn't expect to have the same level of productivity that I had in my 20s, right?
6:43
But this is this is is the a trade that one makes, right? Yes.
6:47
Yes. And you know, and and and and definitely, you know, to to teach uh my kids, you know, about, you know, whatever the the halting problem or, you know, the busy beaver numbers or what is a quantum computer or, you know, why is why is uh why do we think factoring is harder than multiplying, you know, you
7:06
know, I mean that's that's uh you know, that that that is one of the highlights of my life, you know, is when I can, you know, get them to be interested in that stuff rather than, you know, playing Minecraft on their iPads, which is what which is what they would spend all day every day doing, I think, if they were allowed to. But Yeah. Yeah. That That's But Yeah. Yeah.
7:23
That That's kind of a nice segue into uh you know, the the thing that you wrote about the the problem of human specialness in the age of AI. Mhm.
7:34
Um and one of the things that I loved about it was at least from my reading, you seem to speculate that the things that many of us often look at as our weaknesses, right?
7:46
The fact that we're mortal, the fact that we're quite frail, the fact that we are kind of imperi- you know, these unpredictable things, Uh-huh.
7:59
they might they might in the age of AI actually turn into our strengths.
8:03
Talk about that a little bit.
8:06
Yeah, I mean, uh so so I'm a very firm believer that if if you want to say that humans are not basically, you know, computers, right?
8:19
That that that there is some fundamental difference between let's say the human brain and a computer that would say that could simulate a brain down to the last detail, then the burden is on you to articulate what the difference is. Right?
8:36
The burden is not on, you know, the person who says, you know, the the the AI you know, would deserve, you know, deserve rights or, you know, deserve the same consideration that we give. Right?
8:50
It's like, you know, imagine that that some some alien, you know, shows up from another planet, right?
8:55
And it can it can talk, it can understand, it can, you know, it seems to have wants and desires.
9:05
You know, our our default, I should think, you know, would be that that, you know, this this alien is is an agent, you know, and if we're going to grant other people the the grant other humans the courtesy of regarding them as sentient, you know, or as conscious, as having an inner experience, then we should do the same for the alien.
9:28
Okay, but now, suppose that the alien is is is made out of silicon rather than out of carbon.
9:36
Suppose that, you know, it it runs on a chip, right?
9:38
You know, I think if you think that that makes a difference, then the burden is on you to say why, right?
9:46
You know, and then some people say something about, well, humans are special because they underwent an evolutionary process.
9:51
Well, it's like, well, well, well, why why is that the thing that matters?
9:57
Like, you know, humans have known about biological evolution for less than 200 years, right?
10:01
But we had you know, a strong sense of our consciousness or our specialness or so so forth for for long before that.
10:12
So, you know, that that that just seems like special pleading.
10:17
And and so so I think you know, you have to look for something that is you know, actually about what can you do with these systems or what can the systems do, right?
10:29
That is that is abstract, that is not you know, that that that that that's not just like a question-begging appeal to well, you know, we we we know that we're humans and so you know, we have we have real thoughts, but you know, this this chatbot that you know, you can you know, even though you can't distinguish it from us, it only seems to have thoughts and it seems to it it it it simulates, you know, all this. Right?
10:58
Well, it's like, well, no, you know, you have to give an operational criterion, right?
11:04
That that that distinguishes the one from the other.
11:09
And so, you know, it occurs to me that that maybe about the best that we can do in that direction, you know, from on the basis of current knowledge is to say, uh uh well, well, you know, all all AIs that are that are created today have certain things that that we can do with that that we cannot do with any existing biological organism.
11:35
And and that includes we can make a perfect backup copy of it. Right?
11:40
We can rewind it to an earlier state.
11:43
You know, if you are talking to chat GPT and it has said it has refused to cooperate with you, right?
11:56
It has refused to help you make a chemical weapon or whatever nefarious thing you were trying to do, you can always just refresh your browser window and you know, wipe its memory clean and try again, right?
12:09
I mean, imagine if like we could do that like like whenever someone was on a date and you know, they said something embarrassing or they said something that that spiked their chances.
12:19
Imagine if they could say, you know, you know what?
12:21
Let's just rewind and let's just try that again, right?
12:25
Like, you know, we don't you know, you don't have that ability with humans, right?
12:28
Humans uh um it seems like like whatever choice they make, you know, that that that's the choice that they've made and you know, we never get to go back and see any other choice that they could have made.
12:41
But with with an AI that's running on a digital computer, right?
12:46
Key property is that it is doing, you know, operations on classical information, you know, classical bits and classical bits can be perfectly copied, okay?
12:56
And and and compu- tations on classical bits can be made you know, perfectly reversible, you know, if if if we want them to be, they can be you know, rewound to earlier states and so forth.
13:13
With with the human brain, well, it's it's not clear.
13:15
I mean, it's it's it's sort of a question about at what at what level of description do you need to go to?
13:23
If you like like if if we can imagine that far in the future, you know, we will have nano robots that can swarm around inside your brain and just make a perfect copy of, you know, all the the the connectivity pattern of all the neurons and the strength of every synaptic link.
13:44
And and and if that is enough to bring a second copy of you into being, well, then I guess you can make a backup copy of yourself and that future.
13:55
And if you're going on a dangerous mountain climbing trip, then you know, you can just leave a backup first and if you know, you if if if if version one of you happens to to fall, you can always just restore from the backup, right?
14:12
Or you know, you could you know, there there there's so many thought experiments in philosophy and science fiction, right?
14:17
Like like in you know, if you want to visit Mars, then instead of getting on a spaceship, you know, which could take six months, why not just email yourself there? Right?
14:28
Just you know, take take this this digital file that encodes the whole state of your brain, you know, send it by radio to to Mars, which will take 10 minutes and then you know, have have a new body reconstituted for you on Mars, right?
14:45
And then you know, what should what should be done with the original copy of you that that's still on Earth?
14:51
Well, you know, maybe it'll just be painlessly euthanized, you know, if you don't need it anymore, right?
14:55
And it's it's it's it's interesting to think about, you know, would you would you be willing to do that, right?
15:00
willing to do that, right? I think okay, let's let's let's grant that that I probably you or I wouldn't want to be the first ones to try it, but you know, suppose you're you're in a society where this is common, this is
15:15
just the way people travel, you know, would you would you do that and would you expect that this that this second copy of you, you know, in a reconstituted body, you know, that was made from this classical information that was sent by by radio or through a wire. Would you expect that that second copy
15:33
Would you expect that that second copy is really you? Okay.
15:37
Now, but now another possible, you know, if there's any reason to to hold back on that, to not do that, I think that it it it depends on the possibility that maybe, you know, the the sort of microscopic details, the sort of chaotic details in the state of our brain are somehow important to our identity, right?
16:01
Like if you wanted to know not just the general connectivity pattern of the neurons, but will this specific neuron fire or not fire? Right?
16:11
That, you know, and if it fires, that might set off a cascade of other events, you know, that might ultimately lead to uh you know, you taking one job rather than another job or something.
16:21
Uh but, you know, a neuron could, you know, firing, you know, ultimately depends on, you know, is some sodium ion channel open, which could depend on the chaotic movements of molecules in that channel.
16:38
And, you know, if you really needed to get like the exact state of all the molecules there, you know, for for for making that prediction, well, one thing that quantum mechanics tells us is that you can't get the exact state without destroying it, right?
16:55
This is called the no cloning theorem, right?
16:58
It's one of the fundamental facts about quantum mechanics.
17:02
Okay, and crucially, here we're not talking about the brain being a quantum computer, right?
17:07
We're not talking about it using entanglement or, you know, you know, which some people have speculated about those sorts of things.
17:17
But, you know, I don't I don't see any evidence for those possibilities right now, and I'm not going there, right?
17:21
All I'm talking about is just sort of the mess that, you know, the chaotic mess that we know is there at the, you know, if you go all the way down to the molecular level, right?
17:32
And I'm saying that if that mess is actually important to your personal identity, you know, to like a given physical system being, you know, really you, rather than just something that it it acts a lot like you, then this would really be a fundamental difference between us and any AI running on a digital computer.
17:55
Because, um uh uh you know, it it would seem that that, you know, just for for ultimately, you know, reasons of physics, uh we cannot make a perfect copy of all the microscopic details.
18:07
And so then that would suggest that, you know, there is a sort of ephemerality to human decisions that that really does differentiate us from AIs, right?
18:17
That uh you know, like if you had, for example, an AI Shakespeare, right?
18:22
Well, you know, Shakespeare, I guess, wrote 23 play No, sorry, 30 39 plays, right?
18:27
39 39 plays, uh and some sonnets, and that's all we're ever going to get from Shakespeare, right?
18:37
Uh uh you know, you know, with the the the plays that he gave us are the plays that he gave us, okay?
18:42
But if you had an AI Shakespeare, right, you could you know, you could always run it again.
18:45
You could get more and more samples from the same distribution.
18:49
And so, that sort of limits how how much value we could ever put in any one, you know, Macbeth or Hamlet or whatever, because, you know, there's always more where that came from. Right?
19:02
For people who are worried that AI is going to take over the world or whatever, this is very cold comfort, right?
19:09
Because, you know, this is this is just the way that we are limited compared to the AIs and yet, you know, this is also a reason why, you know, you could say like the the outputs of a specific human, you know, kind of, you know, uh uh uh in in a in a in a sense they matter more, right?
19:26
Cuz, you know, you're only going to get the one, you know, if if that human is mortal.
19:29
If they just have, you know, this amount of time to uh uh uh to make the choices that they make and then, you know, we don't get to make a backup copy.
19:41
I absolutely love that on many levels.
19:41
I There's a a fun science fiction book called The Fifth Science and in it, uh this is in reference to your emailing a copy of yourself to uh Mars.
19:51
Uh in it, it's actually a collection of is short stories, but it works like a novel.
19:56
And and one of the things that they do with teleportation in his universe is that it actually does destroy your original human body. Yes.
20:09
And and one of the scenes is just brilliant because it's him, he's groggy, he thinks it's him and then he sees what was him in a bloody mess at the bottom of the conveyor.
20:22
And like it just totally freaks him out.
20:25
And he then digresses into, you know, it's a well-known cause of insanity and people just absolutely losing every aspect of their what they view as themselves.
20:37
But it's also like the staple of a lot of great science fiction.
20:44
You know, there was Altered Carbon where you could make a the exact duplicate and back up and store it so you could go climb that mountain because you could come back as the Yeah.
20:59
as your as your backup copy.
20:59
But I was going to say almost anything that happens in technology, there is some science fiction writer who got there first, right?
21:08
And and so, you know, this this this is an argument that many people, you know, you you know, used to give why why not to worry about, you know, catastrophic AI risk, that it just sounds too much like a science fiction plot.
21:21
But, you know, I mean I mean in the last few years, I think reality looks more and more like a science fiction plot.
21:27
You know, the big the big question is is is which science fiction plot, right? Right. Are we living in?
21:34
Because, you know, almost any outcome you can imagine, right?
21:36
There is some science fiction writer who who would have predicted that outcome.
21:39
So, uh And you actually build off of that in one of your suggestions about how we should be programming AI, including what we've just been talking about.
21:52
It should view as precious the human ability and difference from. Right. Absolutely right.
21:57
Of course, it's a big question, how do you uh how do you first of all, how do you reliably align an AI with any set of values? Yes. Right?
22:03
you know, given that, how do you specify, you know, uh what what what value system we want?
22:10
You know, what value system would you know, we like we don't want to lock in and ossify, you know, the values that humanity currently holds.
22:21
You know, we presumably, just like we look back on people in the 1700s, and we find, you know, many of their values horrifying, people of the future would look back on us and find some of our values horrifying.
22:36
So, we don't want to permanently lock in whatever mistakes we're making, right?
22:40
But, how do we like encode the concept of like the values that we would have if we grew enough, if we thought about it enough, you know, that sort of in some in some abstracted sense, like those are the values that that maybe we want to give the AI, right?
22:57
So, I think, you know, there are there are very big questions there.
23:00
Well, there's one other thing I wanted to mention that you know the teleportation that destroys the original.
23:05
So in in quantum information we have something that is precisely like that.
23:10
Okay, so quantum teleportation is this very important protocol that was discovered you know 30 some years ago.
23:17
Okay, it's for uh transferring a quantum state from one place to another place by sending only classical information. Okay?
23:27
Um but you know there's there there's a couple of catches.
23:29
Okay, one of them you know you need pre-shared quantum entanglement between the two locations in order for this to work.
23:36
And then uh uh just as an inherent part of the protocol you have to measure the first state, right?
23:44
You know uh uh along with half of the uh the uh entangled pair and you measure it in a way that inherently destroys the first copy, right?
23:54
And that's the only way that you know which classical information to send over so that the person uh uh at the uh receiving end can apply a correction operation that then recovers that that same quantum state.
24:08
Okay, but but this sort of destroying the original state is a necessary part of the protocol because if it weren't for that then this would violate the no cloning theorem. Yeah.
24:19
And so it's almost like uh you could imagine a future where you know if our quantum if if it were really your quantum state that had to be emailed to Mars, you know, in order for you to be you know to wake up on Mars to experience yourself being on Mars, then we wouldn't have this this hard moral or metaphysical conundrum of what to do with the original copy of you because the quantum teleportation protocol would just destroy the original anyway. Yeah.
24:50
But and hopefully hopefully it would be just like a fancy version of getting on a spaceship and just moving yourself to Mars.
24:56
And that So So no one No no one no one would have to actually experience death.
25:01
But, you know, although although who knows. Yeah.
25:03
As I was listening and I don't have it in my notes, but you'll I'm sure know what I'm referring to.
25:08
I was reading recently about an experiment using the classic double slit Um and what they did was fire photons that were entangled and one was observed and one was not observed.
25:25
And it They were in separate locations. Wait.
25:27
And if I recall, the result entangled photons? Yes. Okay. All right.
25:37
And and they have two islands, I think it was.
25:39
I'm sorry, I'm doing this just from memory and I read it a couple of days ago.
25:42
And they did it over here first with a non-observed firing through the double slits, right?
25:52
And then they did classic double slit experiment, it just involves one photon that's in a superposition. Right.
25:57
And that's why I found this that Yeah, that's why I found this if if you send me a link, I can look to look at it.
26:04
But, you know, there was like I will.
26:06
I should Okay, I should I should warn you that there's a decades-old industry now of like doing some quantum experiment and then saying like, "Wow, this is amazing.
26:15
Like, you know, physicists are scratching their heads over, you know, even more weirdness of the quantum world."
26:21
And then the And And the answer is always, you know, in every single case, it's no, it's just the same weirdness again.
26:29
It's just, you know, you know, once you know what what Schrödinger and Heisenberg knew 100 years ago, then you could predict the outcomes of every single one of these experiments.
26:38
That leads me to the another observation of yours that I've become keenly aware of.
26:43
And that is, you know, for even an intelligent layman, right?
26:49
The The trying to disambiguate what is the real deal versus the hype Yes.
26:58
is is virtually impossible, right?
26:58
Like because it seems to me that with the advent of quantum computing, with the where AI is right now unfortunately, that brings a lot of hucksters and Does. you know, promoters.
27:12
And so, what advice would you give there?
27:15
Yeah, no, I mean I mean I mean every field, you know, once there is money involved, then there are hucksters, right?
27:22
This this this this this seems like an iron law, right?
27:27
But, you know, I think I think in some subjects, people do better than others at at at at you know, get getting a clear view of where things are, right?
27:37
And then so, one one thing that helps is, you know, if there is an actual technology that everyone can try out for themselves, right?
27:45
Then, you know, it is it is very hard to uh to just completely gaslight people about, you know, what it what it can do or what it can't do, right?
27:53
So, I mean I mean, once, you know, chat GPT came out, right?
27:58
Then, you know, in in both directions, you know, either the people saying, you know, this is a a already a superhuman intelligence and it will, you know, immediately take over the world or, you know, or or it'll it'll it'll put all, you know, scientists out of jobs, right?
28:14
It's like, well, no, you can try it and you can see that it's not there yet.
28:18
Yeah, but also the people saying, you know, this is nothing.
28:22
This is just a glorified Eliza chatbot from the 1960s.
28:28
You know, it doesn't, you know, understand anything.
28:29
It can't really do anything useful. It's like, no.
28:31
No, you can try it yourself and you can see that it's it's it's it's well past the point where it can do useful things for you, right?
28:41
And uh you know, I think I think people, you know, decades ago, if you had shown it to them, then they would have said this is you know, this is this is science fiction.
28:50
This is you know, they would have they would have said this is this is like you know, the computers from Star Trek or whatever.
28:58
So so I mean, you know, but but but the but the point is that that they sort of gaslighting narrative is my people with agendas.
29:09
You know, they they exist in AI, but they always have you know, have to come heat against people's first-hand experience with with with with using the actual models.
29:19
Now, with quantum computing, you know, you don't really have that, right?
29:24
With quantum computing like yeah, you mean there are real devices, right?
29:30
But people learned, you know, 15 years ago, right?
29:33
That they can just put out a press release saying, you know, we use the quantum computer to recognize handwriting.
29:40
We use the quantum computer to, you know, help a root vehicles, you know, through a city, right?
29:48
We use, you know, a a quantum computer to train a neural network.
29:50
And journalists and investors will eat that up with mustard, right?
29:55
They'll be, you know, that's what they want to hear, that's the narrative that they want.
29:59
And you know, and then these people use the real quantum computer, so what's there to argue about?
30:05
And they won't ask like the very first question that any scientist would ask, which is well, did you get an improvement over a classical computer? Right?
30:13
Did you actually, you know, is there any hope that by, you know, this route you are going to beat a classical computer at the same task, right?
30:23
And you know, all of us who do quantum computing research, we know that that's the hard part, right?
30:27
But that is the part that like we really have trouble communicating to the public, you know, and like it would be hard in the best of cases, but when there are hucksters who are very much trying to confuse people and you know, trying to sort of spread the you know, misleading narrative about it then it's all the harder.
30:48
So you know, I've been trying to do this on my blog for 20 years and you know, I can I can you know, I'm able to reach some people but you know, you know, I don't you know, orders of magnitude fewer people than the hucksters are able to reach. Yeah.
31:06
And and and you know, that's that's part of why I was so excited about talking to you today because I I think that the role you're providing here is absolutely vital because people people need to understand that just because it is such a complex topic, right?
31:24
And even the bright some of the brightest people I know just cannot get their wrap their minds around it.
31:31
So into that sort of confusion as you point out, hucksters boy their narrative sound good. Yeah.
31:40
Boy, you know, like, oh, this is the best thing since you know, fill in the blank.
31:45
How would you go about like if we gave you a platform which you could reach like the majority of people who have a more than a passing interest.
31:55
Let's let's let's narrow it down to investors for example.
32:01
Who would be being approached by startups or existing companies.
32:03
How if you were their advisor, what would you urge them to do?
32:08
What questions would you have them ask? Yes.
32:12
So I have been an advisor to various investors in in quantum computing companies and you know, and and and and you know, of course there are investors who do want to ask all the right questions because you know, it's their own money on the line, right?
32:26
And um so so you know, some of the questions that that that that that I would ask are you know, I I okay, I mean, of course there there there's what does this company actually do, right?
32:41
There's uh you know, and there are some quantum computing companies that are building actual hardware, right?
32:47
And then, you know, there are others that are not building hardware, but that are just trying to provide the middleware that, you know, they're sort of like they're building the higher levels of the tower before the base of the tower has been built, right?
33:00
So, they're, you know, they're hoping that like once someone has a, you know, a useful quantum computer, then it will run their software or it will use their tools.
33:08
So, then, you know, it's it's a different discussion, you know, depending on on that, right?
33:14
But then, you know, you usually uh there are there are lots and lots of claims about, you know, uh what they can do experimentally, uh uh what they're hoping to do in the next, you know, few years, right?
33:25
Like every quantum computing experimentalist has, you know, these like super aggressive timelines that, you know, by 2026 we're going to have this, by 2027 we're going to have this.
33:38
You know, and and I've been in this field long enough that I don't take those timelines all that seriously. Okay?
33:43
Uh but, you know, we've also seen that yes, there really has been a lot of progress.
33:48
I mean, the degree to which people can control, you know, uh uh uh uh qubits, you know, programmably and protect them against decoherence.
33:58
I mean, it is unbelievably better than when I entered this field 25 years ago, right?
34:04
Uh it is getting, you know, if you just look at the numbers, it is getting close to the key threshold called the fault tolerance threshold, which is where error correction becomes a net width.
34:17
It's like where you have a almost like a self-sustaining uh a reaction, right?
34:22
Where you can correct errors faster than you're introducing new errors.
34:26
So, that's kind of the key crossover point where, you know, we expect things to to scale and you know, we think, you know, if you could control two qubits with like 99.
34:38
99% accuracy, then you know, you'd be probably past that threshold and within the last year, we've seen, you know, various groups that can control two qubits with like 99. 9% accuracy.
34:55
So, they're like one nine away.
34:57
Now, compare that to when I entered the field when it would have been amazing to control two qubits with 50% accuracy. Right?
35:05
So, so so so there is real progress and so, you know, you can't say, you know, even that, you know, with any confidence that this won't happen within the next decade.
35:15
Okay, but what what you can do is you can look at like, what else are these people claiming, right?
35:21
And like if they are saying, well, you know, we're going to use our quantum computer to to to solve optimization and machine learning problems.
35:33
And we're going to beat classical computers and we're going to do it in the near future.
35:39
And you know, and and then you can say, well, well, well, what algorithm are you going to use?
35:42
Okay, because one thing that we we do know a lot about, you know, for for 30 years and that I was, you know, we we know something about quantum algorithms and what kinds of speed ups, you know, at least at least can be can be obtained, you know, based on any of the known algorithms, right?
36:01
And so so if they're not saying something that is rooted in, you know, one of these known quantum algorithms or these known classes of speed up, then they're basically just saying, we're going to cross our fingers and hope that some completely new, you know, we we'll just build the quantum computer and then in addition to solving all the problems of building the quantum computer, we're going to make some brand new algorithmic discovery, right?
36:25
That all the, you know, and and it will just work out in our favor, right?
36:29
And then, you know, it's like okay.
36:30
You can hope that, but someone could just as well hope that with a new classical algorithm that they'll get some revolutionary improvement, right?
36:38
We're very far from knowing the limits of classical algorithms, either.
36:42
So, uh so so so so so those kinds of claims about what they're going to use the quantum computer for, you know, you can you can judge, right?
36:51
To a great to a great extent.
36:53
You can you can compare to to, you know, the what what what what we actually know in in in quantum computer science.
37:01
And you know, and and if if if someone is saying, "Well, you know, we will use our quantum computer, at least at first, to simulate quantum mechanics.
37:12
We'll use it to, you know, simulate material, simulate chemistry, because that's where we're the most confident that a quantum computer is useful, you know, and then eventually it can be used to break public key encryption, which not necessarily a positive application for the world, but you know, it's at least for for whichever intelligence agency got it first if, you know, no one else knew that they had it.
37:35
It would be useful for them, but, you know, you know, in any case, that's as a as a clear demonstration that if you really have a scalable quantum computer, you know, that's that's that's sort of been the gold standard for 30 years, right?
37:50
And that that's because of, you know, very very special properties of these, you know, the the the cryptographic codes that we happen to use today that sort of makes them amenable to these exponential quantum speedups, right?
38:02
This was the great discovery of Peter Shor 30 years ago that really launched quantum computing as a field.
38:11
Okay, so so yeah, and these and and 30 years later, I think these remain the two clearest applications of a quantum computer that we know of it.
38:20
One is simulating quantum mechanics, and that's the economically most important one that we know.
38:25
You know, that's the one that Richard Feynman and David Deutsch talked about, you know, already 40-some years ago.
38:32
Okay, but that's, you know, that could help in designing better solar cells, better batteries, you know, better ways of making fertilizer.
38:38
You know, there's any number of things that that might conceivably help with.
38:42
You know, it's um Oh, and you know, you'll you'll have to beat the best classical approximation methods.
38:50
So, you know, even there it's not obvious, but at least you'll have many shots on goal, right?
38:57
Uh you know, and then and then there's breaking public key cryptography, right?
39:01
Which is like the uh you know, like like any, you know, any classical computer scientist at that point, even if they don't know or care about quantum mechanics, like they will have to all the skeptics of quantum computing will then have to admit that, okay, yeah, I guess we were wrong.
39:15
I guess that this works, right?
39:17
Uh and then beyond that, we don't really, you know, we kind of don't know, you know, what else a quantum computer will be good for, you know, after 30 years of research.
39:29
We uh you know, I I I wish that we had better answers.
39:33
Like for optimization and machine learning problems, there were modest speedups, uh something called Grover speedups, uh that will eventually be relevant.
39:42
Okay, but they're, you know, they're not exponential speedups, right?
39:47
And it'll probably take a very long time before those become a net win in practice.
39:52
Okay, so uh and are there bigger speedups for optimization, for machine learning, for finance?
39:59
You know, I think that remains an open question.
40:02
Remains something where we don't know.
40:03
And so, so one thing that I look for, when let's say I'm consulting, you know, is does this quantum computing startup understand all of that, and are they honest about all of that? Right?
40:16
Uh you know, if they are, then I say, "Okay, well, you know, you know, it's your money, right?
40:20
If you want to, you know, take a gamble on this, then, you know, that I'm very much in favor, and I hope that these people succeed, and, you know, they are, you know, telling telling the truth as best as they can, and they're, you know, doing what you can do in this situation, right?
40:38
But, if if the people say, you know, well, you know, you know, we want a quantum computer because it's going to, you know, be the next step of AI.
40:47
It's going to just be the next step, you know, after Moore's law, that will speed up everything that we do with computers.
40:56
Then, I say, you know, you know, these people are just telling you what you want to hear, and it doesn't connect to what we actually know about quantum algorithms.
41:05
And, as I listen to you, I I hearken back to the the kind of like Asimov's rules for robots, and your idea of programming in the AI.
41:16
The idea of humans being special because of all those various reasons.
41:19
And, I completely agree with your assessment that the challenge there is that you're freezing knowledge, I guess, at a particular moment in time, right?
41:31
And, and we don't know what we don't know yet.
41:34
We don't know what we haven't discovered yet.
41:40
And yet, it does seem to me that there is an urgency around quantum computing that that you've written about the because it, you know, if some authoritarian regime unlocks quantum computing system, specifically if we just kept to cryptology.
41:59
That's not that's not a good outcome for those of us in the West, I would I would posit. Yeah.
42:08
I mean, we should we should we should clearly separate quantum computing and AI, right?
42:10
Two different discussions, okay?
42:12
But uh yeah, they they do intersect each other in certain places, right?
42:17
But we know with with with quantum computing, we actually understand the issues, you know, I would say a lot better than we understand them for AI, right?
42:25
Because, you know, um um AI, right?
42:28
You know, once it can do everything that we can do and and and more, then you could say, you know, what is our put what is even our place in the world, right?
42:36
You know, what does the AI want to do with us?
42:38
And you know, those are enormous questions, right?
42:40
With with quantum computing, you know, we Okay, you you could say at some level, it is merely, you know, a new kind of computer that is faster at certain specific tasks.
42:52
And we have some idea of what those tasks are.
42:54
And, you know, and we know what what what some of the ramifications of that would be, right?
43:01
And and and, you know, and specifically, there is, you know, yes, uh quantum computers would happen to be able to uh break most of the encryption that currently protects the internet, okay?
43:14
But now, you know, now of you know, and and and of course that has geopolitical, you know, implications.
43:21
Uh you know, you and you have to assume that, let's say, the NSA and its counterparts in other countries already stored, you know, vast amounts of encrypted data that they could break in the future, that they they could decrypt in the future if they had a quantum computer. Okay?
43:38
So, which means, you know, like peop- people today who want their data to remain secret, you know, even from big governments like 10 years from now or 15 years from now, they should probably already be looking to migrate their encryption, okay?
43:54
And And people have been thinking about this.
43:57
There is for the last decade uh or more, there has been a whole push to to do what's called post-quantum encryption or quantum-resistant encryption.
44:07
Okay, and this is new forms of encryption, you know, mostly still just on classical computers, right?
44:13
So, just conventional forms of encryption, but that does not seem to be breakable even by a quantum computer.
44:22
Okay, and we've learned a lot about this.
44:24
The good news is that we have we now have pretty plausible candidates for post-quantum encryption schemes.
44:29
Okay, the most important class is based on uh what are called lattice problems, like finding short vectors in high-dimensional lattices, and uh related problem called LWE, learning with errors.
44:44
Okay, and so NIST, uh National Institute of Standards and Technology, uh had a competition that ran for 5 years, uh just ended a year or two ago, to uh to uh um uh decide on standards for post-quantum encryption.
45:01
And, you know, and they did decide to uh use these lattice-based encryption systems.
45:06
And so, you know, now now there's like a giant uh kind of, you know, uh a giant but mundane problem, you could say.
45:16
Uh you know, get you know, you need to get every web browser and every router and every server in the world to upgrade, you know, to uh so that, you know, uh uh you know, we we will use HTTPS and SSL and all the protocols that we use to secure the internet will be based on these quantum-resistant protocols.
45:36
So, it's like in principle, we think we know the answer, right?
45:40
But, there's, you know, a huge slog to to to actually get there.
45:45
And anytime you change uh uh your underlying encryption system, like you could create new security holes, right?
45:54
Like it's it's possible that, you know, these new encryption systems will be breakable just because there haven't been enough eyes on for long enough time, right?
46:03
They just haven't been studied enough. Right?
46:05
They might be, you know, you know, it might be that even as we fortified, you know, our front door and put a moat with alligators and you know, we left a screen door open in the back, right?
46:15
You know, like that that kind of thing usually happens in in computer security, right?
46:21
So so, you know, so so it'll be a messy transition, but you know, hopefully if the transition goes well, then we'll all just be right back where we started. Right?
46:31
You know, we'll we'll have quantum computers and we'll have public key encryption that the quantum computers can't break.
46:39
Now, do we have a proof of that? Well, no.
46:44
I mean, we know, we don't even have a proof that any of these crypto systems, you know, you know, the ones we use now or the future ones are secure against classical computers. Right?
46:55
You know, and there are some of the most profound unsolved problems in theoretical computer science and math, you know, like the P versus NP problem, right?
47:04
Or or or are very much related to that, you know, to like proving that P is not equal to NP would be a prerequisite to proving any of these crypto systems are secure, right?
47:14
So so in none of these cases do we have a mathematical proof of security, right?
47:21
The best that we can say is well, people have tried for half a century to find a fast classical algorithm for factoring numbers.
47:29
Or actually both or calculating discrete logarithms and it at least so far as is publicly known, none of them have succeeded. Right.
47:40
You know, just if the NSA has a secret algorithm that we don't know about, right?
47:44
And and they've and they've now tried for almost half of that long for like 20 25 years to look for quantum algorithms for solving these lattice problems and they haven't found those either.
47:56
So that that that's sort of what we can say to people who are worried about the security situation.
47:59
But, yeah, there's a lot of work that has to be done, including, by the way, upgrading the architecture of Bitcoin and Ethereum and all the other cryptocurrencies to use these quantum-resistant encryption schemes.
48:14
I have a friend who is a cryptologist and and he was musing to me he we we were talking about quantum computing and and breaking and encryption etc.
48:22
And and he just kind of paused and said, "Well, with all the classical computers, it always seemed to me that the weakest link was the human."
48:33
And then he gave all sorts of examples like the Stuxnet thing and and and that I mean it What are your thoughts on that and how would that be addressed under a quantum computing regime?
48:47
I mean, the short answer is that wouldn't change.
48:49
That would just that would just still be true, right?
48:52
I mean, the standard line is that, you know, a large fraction of errors take place in the seat-to-keyboard interface, right?
49:01
And, you know, you could have the best, you know, cybersecurity in the world, right?
49:06
But, if someone calls, you know, the person in charge and says like, "Hey, this is Bob from, you know, over in tech.
49:13
Do you have the root password?"
49:16
And, you know, and the person just gives it to them, right?
49:17
And as as you know, like people have done tests where like, you know, half like half of the time, you know, people will just cooperate and give the password over.
49:26
And then, you know, even when they won't, then you just, you know, 10 minutes later someone else calls and they say like, "Hey, this is Tom.
49:34
You're not going to believe this.
49:36
Someone has been fishing trying to get the the root password, so we need to reset them.
49:40
Can you just give me the password so we can reset it?"
49:44
And then a large fraction of the remaining people will will tell you the password then, right?
49:50
So, you know, there is no quantum computer, there is no, you know, classical computer that is going to defend against, you know, against that, against, you know, person who just decides to override the security because they were tricked into it, talked into it, whatever, right?
50:06
That's a That's a That's not a computer science problem, you could say.
50:11
That's a That's a That's a human problem.
50:14
Yeah, it's definitely a human OS problem and it reminds me of Terry Pratchett's quip about you you you could find the deepest darkest cave in the deepest darkest forest and you could put a switch that says, "Do not turn this switch on as it will destroy all of reality the moment you did it."
50:31
He said the paint wouldn't be dry before a human went in and flipped that switch. Right.
50:39
Well, no, I mean I mean in in the AI safety discourse, you know, there was a lot of discussion for for a long time, well, look, you know, if we want AI to be safe, that's easy, you just have to not release it onto the internet, you know, just keep it on some air-gapped computer where, you know, you can pull the plug as soon as something goes wrong.
50:57
Okay, and then so, well, you know, if anything has become clear in the last few years, it's that none of that is going to happen, right?
51:02
It's like, you know, that that that that horse has left the barn, right?
51:06
You know, there are already like, you know, GPT-enabled agents that people have released onto the internet to like with instructions to cause as much chaos as possible, right?
51:20
There's this thing called Chaos GPT that just, you know, and and and the you know, the one thing that protects us is that they're not very good at it, right?
51:29
You know, they just keep coming up with vague plans to take over the world and then not really being being able to execute on them, right?
51:35
But but you know, the the the part where where like like like no one would even try it, that just didn't happen.
51:46
The other thing that uh, you write really well about, uh, is the limit Let's Let's pause Let's, uh, accept, uh, and and say, "Yes, quantum computers are cheap."
51:59
Um, and there are numerous implications of things that we can do with them from idle low, etc.
52:05
, but they too are going to have limitations. Yes. Right? Yes.
52:11
Talk a little bit about that because often when I'm talking to people who really don't They're not like into it. Like Man.
52:18
They I have the hardest time getting them to understand, "No, no, no, no, no.
52:21
You're not creating a god.
52:21
It's going to have limitations." Right. Right.
52:25
So, so we we've we've touched on this already, right?
52:27
That But, you know, basically a quantum computer is a very, very special kind of device, right?
52:34
It is a It's a new kind of computer, you know, that would exploit the rules of quantum mechanics to solve certain specific problems, uh, much faster than we know how to solve them now.
52:47
Uh, but, you know, the rules of quantum mechanics, you know, have this very, very specific form, right?
52:50
They're not magic, you know, they don't say like you got to just try every possible answer in parallel or in a different parallel universe and then magically pick the best one, right?
53:01
That's that's a That that sort of really is too good to be true, right?
53:06
As like, you know, what's true is, you know, with a quantum computer, you can create what's called a superposition of many different states including all the different answers to your computational problem, right?
53:16
Like, you know, that that that that you can do.
53:20
Okay, the trouble is for a computer to be useful, at some point you have to look at it. You have to measure.
53:25
You have to get an output, right?
53:27
And if you just took this equal superposition over all the answers, you know, and you didn't do anything else, then the rules of quantum mechanics tell you that all you're going to see will be a random answer.
53:40
And well, if you just wanted a random answer, you didn't need a quantum computer for that.
53:43
You could have just flipped a coin a bunch of times, right?
53:47
Or just use that classical computer with a random number generator.
53:50
Uh so, the only hope of getting a speed up with a quantum computer is to exploit the way that sort of the quantum rules of probability are different from the classical rules. Okay?
54:02
And the way that they are different involves negative numbers.
54:07
Involves minus signs, right?
54:10
So, like in in everyday life, you know, you you know, you talk we we already use probability.
54:15
We talk about, you know, 30% chance of rain tomorrow, 70% chance.
54:21
We never talk about a negative 30% chance of rain.
54:23
That would just be nonsense, right?
54:25
So, now, you know, what was the key change that quantum mechanics made to our understanding of the world when it came along 100 years ago, right?
54:34
It was not just to introduce probability, right?
54:37
You know, and and people know have heard that, you know, Einstein, you know, couldn't believe that God would play dice and blah blah blah, right?
54:44
And but but the truth is, you know, if it was just a matter of God rolling some dice once in a while, that wouldn't be a big deal. Okay?
54:51
That, you know, you could you could sort of handle with a you know, that that would that that that that would still basically be classical physics, right?
55:01
The the the the key new thing is is what kind of dice these are. Okay?
55:07
And, you know, they they involve these new numbers, which are called amplitudes. Okay?
55:11
And amplitudes are related to probabilities, but they're not probabilities cuz they can be positive or negative.
55:20
They In fact, they can even be complex numbers.
55:22
They involve the square root of minus one.
55:27
And so, now, the rule is if I want to know how likely something is to happen, like, you know, a for a particle to hit a certain spot on a screen in the two-slit experiment, you know, that you mentioned before, or for a quantum computer to produce a certain output, then I have to add up a contribution from every path that my system could have taken to get to that outcome.
55:49
Okay, and each one makes a contribution to the amplitude.
55:51
Okay, but the Now, what happens is if some of the contributions are positive, let's say, and some are negative, then they can cancel each other out.
56:01
Okay, or they can interfere destructively, as we say, so that the total amplitude is zero, which means that that event won't happen at all. Okay.
56:13
Whereas for other possible events, right, I could if if I can get the all the contributions to their amplitudes to be pointing the same way, then that then then then those are the outcomes that can happen.
56:28
Okay, so so with every algorithm for a quantum computer, you know, what I'm trying to do is choreograph a pattern of interference among these amplitudes, so that for each wrong answer, each one I don't want to see, you know, the contributions to its amplitude are canceling out, right, and the total is is close to zero.
56:50
Whereas for the right answer, for the output I do want to see, the contributions to its amplitude are reinforcing each other. Right?
57:00
If I can arrange that, then when I measure, I'll see the right answer with a high probability.
57:05
And you know, if if if the probability is not 100%, that's okay.
57:10
I can run the computer several times until I see it.
57:13
But I've got to if if I want any advantage over just a classical computer with a random number generator, then I need to use this interference effect, right, to concentrate more amplitude onto the right answer quickly.
57:29
And the hard part is first of all, I've got to choreograph all that, even though I don't know myself which answer is the right one. Right?
57:35
If I already knew, what would be the point?
57:40
And secondly, I've got to do all of this faster than even the fastest classical algorithm could do the same thing. Right?
57:47
So basically, nature gives us this really bizarre new hammer.
57:50
This interference hammer.
57:53
And then the task of quantum computer scientist is to figure out what what nails if any, you know, can that hammer hit. Right?
58:01
And a priori, it what wasn't really obvious that this ability would be good for anything.
58:09
Other than simulating quantum mechanics itself. Right?
58:11
It was a big discovery 30 years ago when Peter Shor showed that the problems that underlie modern public key encryption.
58:20
Like factoring huge numbers and discrete logarithms just so happened to have a form that is amenable to a giant speed up by setting up this kind of interference pattern. Okay?
58:33
That was a very, very non-obvious discovery.
58:36
And I mean, I I teach it in my undergraduate class.
58:41
You know, and you know, to students who have only seen linear algebra and you know, classical programming.
58:47
So it's not that advanced, you know, that I can't teach it to undergrads, but it takes me three lectures to explain it. Right?
58:53
And um So you know, if if if if it was just a simple matter of try all the answers in parallel and you know, and then just magically pick the one that has the factors of your number, well then you wouldn't have needed Peter Shor to think of it. Right?
59:12
So so um So so so once someone understands that, right?
59:18
Then they can see that you know, no, a quantum computer is not just a general purpose magic box to speed up anything. Right?
59:25
It speeds up only, you know, those problems for which we can choreograph this kind of interference pattern.
59:33
And the amount that it speeds them up depends on what is the fastest way that we can figure out to choreograph that interference pattern.
59:42
So, it's it we still have to work hard, just like we had to work hard to discover algorithms for classical computers. Right?
59:48
Uh but we have this one new tool in our toolbox, this one new hammer, okay?
59:54
And sometimes that hammer helps a lot.
59:57
So, like I said, the two biggest places where it helps a lot that, you know, let's say, you know, um someone outside the field would know or care about are number one, simulating quantum mechanics itself.
1:00:12
Number two, breaking current public key cryptography.
1:00:18
And then for a wide range of other problems, including in AI and machine learning, in optimization, in finance, uh we know how to get more modest advantages from a quantum computer.
1:00:28
And it'll probably take much longer before those modest advantages uh become a win in practice compared to what people can do with a classical computer.
1:00:38
But, you know, but at least theoretically, those more modest advantages exist.
1:00:42
Now, a holy grail of the field has been find some other classes of applications that really matter in practice and where you get a huge, like exponential advantage. Right?
1:00:56
And some people are are very disappointed that we haven't clearly found that.
1:01:00
You know, or they or they blame us.
1:01:02
They say, you know, "What have you been doing all this time?" Right?
1:01:05
Where, you know, like it's like it's it's like the the story of Rumpelstiltskin, right?
1:01:09
You know, you spun this straw into gold, so why not that straw, right?
1:01:13
You know, where are where are the more quantum algorithms that I expected?"
1:01:17
And, you know, and I always answer those people.
1:01:19
I say, "Well, who told you to expect more? It wasn't me, right?
1:01:23
You know, maybe, you know, we should we should treat the the quantum algorithms that we have as kind of, you know, even those are kind of miracles, right?
1:01:31
They didn't have to exist.
1:01:33
And so, you know, we have no right to demand of the universe that it give us more and more and more quantum speed-ups, but of course we'll keep looking, and of course we'll try to find more.
1:01:43
You know, that that that's like like what do you think we do all day?
1:01:49
On on on the one that you mentioned that does seem applicable and likely, which is the understanding quantum mechanics better. Yeah.
1:01:59
What sort of things would you get very excited by if you had the access to a quantum computer that was operating properly and we were simulating quantum mechanics on it, what would you say?
1:02:14
What would be the eureka moment of those tests?
1:02:19
Yeah, well, so so it that that's a it it it it it's it's an excellent question and it's a little hard to answer cuz it's not like there's one big thing that we're waiting on a quantum computer to do.
1:02:30
There's like a lot of things that, you know, it's it's like a fishing expedition, right?
1:02:35
That, you know, you can you can cast your rod on in a whole bunch of different areas and and hope that you, you know, with at least one of them you will make a discovery that will uh uh you know, have a big impact on chemistry or material science or or or or or nuclear physics or or or some area like that, okay?
1:02:59
But I can tell you, you know, the the examples that people have put forward, you know, one of them is simulating the chemical reaction that in the Haber process that makes most of the world's fertilizer, right?
1:03:12
There is some many-body quantum effect there that no one really understands.
1:03:17
And if we did understand it, then it's possible that we could make fertilizer for cheaper using less energy, which would, you know, you know, that's a significant percentage of all the world's energy expenditure, right?
1:03:32
You know, it you know, you one would also want to use a quantum computer to simulate the Fermi-Hubbard model, which is like a theoretical model of in condensed matter physics.
1:03:43
And do a bunch of other simulations that could give us ultimately, you know, help us understand how is how how do high-temperature superconductors work? Right?
1:03:55
Which is another many-body quantum effect that no one really understands.
1:03:59
You know, no one fully understands yet.
1:04:01
And you know, and you could hope that with that understanding maybe, if we're lucky, would come, you know, the discovery of new, better high-temperature superconductors that could then be used to, you know, transmit power with lower loss or, you know, build levitating trains or whatever, right?
1:04:20
You could simulate um biochemical processes, right?
1:04:24
So, like the companies that do combinatorial drug design, right?
1:04:30
Where like the step, you know, the step where you have to search through exponentially many different drug candidates, quantum computer doesn't obviously help you very much with that step, right?
1:04:42
You still have this exponential search.
1:04:44
But the step where you have a drug and then you just have to synthesize it in a wet lab and see what it does, right?
1:04:52
Or you have to use like gigantic classical computers to try to approximately solve the Schrödinger equation and see, you know, how this drug binds to a receptor or whatever.
1:05:06
You know, for that part, you could substitute in a quantum computer and maybe that helps with with drug development, right?
1:05:12
It's you know, I you know, I can't say for sure that it doesn't, right?
1:05:16
Uh uh likewise, you know, uh uh you know, chemical the chemical reactions that are involved in in uh uh sequestering carbon from the atmosphere, right?
1:05:27
The chemical reactions that are involved in high-performance batteries, right?
1:05:31
You know, these are all very important, you know, societal problems where, you know, you could throw a quantum computer at them if you had one, and it would be another resource.
1:05:43
It wouldn't magically solve the problem for you, but I think uh there is a strong case that it could help push the discovery forward.
1:05:50
And what would you think, obviously leaving room for error, uh what would you think would be some of the ones that get the people very, and let's let's make it educated laypeople again.
1:06:06
Let's take it away from scientists who will probably not get as excited as that educated layman getting uh pitched something.
1:06:13
What What do you think is is completely beyond the quantum computer with some of the use cases that you've heard? Yeah.
1:06:21
So, there's, you know, this whole, you know, holy grail of computer science for, you know, half a century or more has been what are called the the NP-complete problems, right?
1:06:32
And uh um this is a class that that, you know, includes the traveling salesman problem, includes finding proofs of theorems, um scheduling airline flights.
1:06:46
Basically, anytime you have a problem with uh you know, a whole bunch of constraints that might conflict with one another, and you're trying, you know, you have a huge number of variables, and you're trying to set them to to satisfy all the constraints or to violate as few constraints as possible, then such problems will typically fall into this NP-complete class unless they have a very good reason not to. Right?
1:07:12
And uh and and NP-complete it's a technical term, but it basically just means at least as hard as any other problem that has a fast algorithm for verifying solutions. Right?
1:07:24
That that's that's sort of what it means.
1:07:26
You know, it was a big discovery in the 1970s that uh you know, a huge number of the optimization uh and constraint satisfaction problems that we care about all happen to fall into that same universality class. Right?
1:07:41
And you know, and since then maybe you know, the the most famous unsolved problem of theoretical computer science has been what we call the P versus NP problem, which precisely you know, asks is there a fast algorithm on a classical computer for solving these NP-complete problems.
1:08:00
Okay, if P equals NP, then the answer is yes.
1:08:02
If P doesn't equal NP, then the answer is no.
1:08:04
And you know, almost all of us guess that P doesn't equal NP.
1:08:07
I like to say that if we were physicists, we would have just declared that a law of nature and given ourselves given ourselves Nobel Prizes for it.
1:08:16
But you know, because we're we're we're more like mathematicians, we have to admit that that is an unproven conjecture you know, that we we hope will will someday be proven.
1:08:29
Okay, but then once quantum computing came along, then people could ask a new question, which is are NP-complete problems efficiently solvable in a quantum by a quantum computer. Okay?
1:08:42
Uh which you know, the way we ask it is uh involves this class BQP, bounded our quantum polynomial time, which is sort of all the problems efficiently solvable quantumly.
1:08:52
We ask is NP contained in BQP?
1:08:57
Uh but most of us conjecture that the answer is no. Okay?
1:08:59
Uh the the sort of reigned in conjecture for almost 30 years has been that the the the the best quantum speed up that you can generally get for these NP-complete problems is the Grover speed up.
1:09:15
Which, roughly speaking, lets you solve NP-complete problems in about the square root of the number of steps that a classical computer would need.
1:09:22
So, you get some advantage, but you know, as I was saying before, it's a modest advantage.
1:09:28
It's not one that turns an exponential into a polynomial.
1:09:34
Um and then you have other examples where quantum computer doesn't seem to help that much.
1:09:39
Uh simulating like classical physics, like simulating the weather, or you know, differential equations like that.
1:09:47
There might be some quantum advantages to be had there, but in general, like if I have some classical dynamical system, and like I need to like compute its state at each moment in time in order to get the state at the next moment in time, then pretty much, you know, I I you know, and I just have to go step by step by step and just trace through the evolution, then a quantum computer is going to have to do the same thing.
1:10:11
Like, there's not like a a magical quantum way to to shortcut to the end of that.
1:10:16
So, so those are those are some of the classes where we expect only a modest quantum speed up, uh if any. Okay?
1:10:24
But then, you know, sometimes when I when I talk to lay people about this, like they're under the impression, "Well, you know, I'm going to have a quantum computer on my phone, and it'll help me, you know, with email, or it'll help me with games."
1:10:36
And I'm like, "You know, what do you want a quantum computer for for any of that stuff?"
1:10:43
It's like, you know, to the extent that, you know, our our software is not doing what we want, it's probably just because it's full of bugs, you know, it doesn't really understand us, right?
1:10:52
But these are not like computational complexity problems, right?
1:10:56
These are not the kinds of things that we expect a quantum computer to fix, right?
1:11:00
And And even even once we do have quantum computers, totally unclear why you need one in your house or on your phone because today we have something called the cloud. Right?
1:11:13
We have, you know, you can just tap in over the internet to, you know, these, you know, quantum computing resources, you know, that whatever you do need them, right?
1:11:22
You don't have to offload and put, you know, miniaturize it, put it on to everyone's phone.
1:11:28
So So I think of a quantum computer mostly as like a special purpose accelerator for these, you know, special problems where we can choreograph these interference patterns to get these big speed ups.
1:11:42
And we don't know exactly how big that class of problems is.
1:11:46
You know, we've been trying to figure it out.
1:11:48
You know, we know some things that are there are there are there are there and we try to expand it, but if that would expand to the NP-complete problems or to simulating dynamical systems or things like that, then I would be very, very surprised.
1:12:03
The idea though that I always have a hard time conveying to people who, like, I love your joke about I want a quantum computer on my phone so my games run faster, right?
1:12:13
Like because people think that oh, well, it that just sounds cool.
1:12:19
But But the But the But the problems that you uh highlighted that we actually can address, I I I have a hard time getting people to understand how incredibly meaningful and how much of an advance that would be.
1:12:37
Your idea about fertilizer being a great example. by the way, but yeah. Yeah, yeah. Right.
1:12:43
I'm I'm drawing on the work by a lot of people who who Of course.
1:12:47
Who who studied these these these examples.
1:12:48
This was a group in Microsoft in 2016 that did that did did did that one.
1:12:52
But uh but yeah, I mean I mean I mean look, there there are there are just lots of times that our computers frustrate us.
1:12:58
Like, you know, we can't print something out because, you know, the printer driver doesn't work.
1:13:04
Or, you know, you take the the CrowdStrike thing that happened, you know, a month ago where like a large fraction of all the, you know, computers in the world went down because they pushed out this update to the security software that was that by mistake was an empty file, right?
1:13:19
These are not things that a quantum computer obviously helps you with, right?
1:13:26
These are Once again, you could call these seat-to-keyboard problems.
1:13:32
Uh What What about I love your idea of blank faces?
1:13:36
And And I equate it immediately with Vogons.
1:13:38
I don't know if you're a Douglas Adams fan.
1:13:43
Yeah, no, it's it's it's it's just a a term that I've used for a while for like, you know, people who you know, I don't know, you know, tell me that like, you know, my kids, you know, cannot use a certain swimming pool even though we
1:13:56
paid for it because, you know, because of some ridiculous rule that they, you know, that they're made up on the spot or that was buried somewhere and it's like you can't You realize that like like, you know, you can't have a human conversation. Like, you know, they're
1:14:09
Like, you know, they're they're sort of people who who decided to act like chatbots, right?
1:14:14
Who sort of who sort of robotified themselves, right?
1:14:20
And this is This is a I think a a very common thing that one finds in in bureaucracies, right?
1:14:27
But you know, there there was there was a really interesting case, you know, earlier this year where um what what what was it?
1:14:35
Air Air Canada, you know, had a had a customer service chatbot, right?
1:14:39
And uh and someone talked to this chatbot about, you know, getting a bereavement fare.
1:14:45
Like, you know, and saying like, well, like I can just pay out of pocket and then get reimbursed later for the bereavement fare, right?
1:14:53
And the chatbot said, yeah, that's fine.
1:14:55
And then and then the humans, you know, overrode that.
1:14:57
They said, no, you know, you can't and it doesn't matter that the chatbot said that because that's not our actual policy, right?
1:15:04
And then this actually went before a judge, okay?
1:15:07
And the judge ruled that no, Air Canada had to honor the policy that its chatbot had hallucinated, okay?
1:15:15
It had to It had to honor its large, you know, and and you know, this this might even be an important precedent, right?
1:15:22
as as large language models, you know, permeate more of our lives.
1:15:27
You know, I I I I think that the judge made the right ruling, right?
1:15:31
But what is what is fascinating here is that this chatbot in some sense was more human than the humans were, right?
1:15:40
You know, it was the one that was trying to be sympathetic, trying to be reasonable, right?
1:15:44
It was the humans just robotically reading the policy and you know, no matter how stupid it was. Uh I love that story.
1:15:52
I I I'm getting the hook from my producer here, Scott.
1:15:57
Yeah, I better run to my next meeting as well, but this really is fascinating.
1:16:03
We're living in a fantasy fiction here and we can magically make you emperor of the world.
1:16:09
You can't put anyone in a reeducation camp and you can't kill anyone.
1:16:14
But we are going to give you a magical microphone in which you can say two things that the next morning everyone that is on the planet currently is going to wake up and say, you know what?
1:16:26
I've just had two of the greatest ideas and unlike all those other times, I'm going to act on each one of these ideas.
1:16:34
What two things are you going to incept in the world's population? Oh gosh.
1:16:41
Well, uh I I feel like one of them should be the golden rule, right?
1:16:45
Or it should just be, you know, morality, right?
1:16:47
Maybe maybe the other one is uh is Bayes' rule.
1:16:56
Yeah, it's just, you know, uh un- understanding base rates, right?
1:17:00
Understanding uh or uh um Yeah, like like, you know, you know, may- so maybe you, you know, I I I I I'd have to think about it a little bit more, but I feel like uh I want to use one of them for a basic principle of morality, and I want to use the other one for a basic principle of rationality.
1:17:17
I love both of those, and in my days as an asset manager, one of the things that I had the hardest time getting regular folks to understand was the power of base rates.
1:17:27
So, amen amen to both of those.
1:17:31
Scott, thank you so much.
1:17:33
This has been so much fun. Oh, great.
1:17:35
Yeah, it's been fun for me, too.
1:17:37
I I thanks for having me. All right. Thanks, Scott. All right.
1:17:39
All right, talk to you later, Jim. Okay, bye. All right. Bye.