The Exchanges

Q Yeah. Do, do you, do you use like any of the systems, uh, for the book, for your articles, like as a, as a journalist and editor?

A Um, so, uh, yeah, I figured that would be a question people would ask me. Um, I did not, I really didn't use it for the book at all. Um, I didn't use it to do any writing. Um, and in part cause I just find like for writing a book length thing in the, in my own style that would sound like me and I felt like would be have the quality that I demanded. Um, I didn't find the systems that useful actually. I thought they were really good at like, they're really good at crafting a business letter. They're really good at Writing a quick email response. Um, I at least struggled to get them to write something that sounded like me. Um, so that was the problem. Maybe I, I don't know, maybe I, it's interesting cause I've, I've talked to Reid Hoffman a number of times and he seems to have had more success in getting, uh, fine tuning his LLM to, to actually imitate his style. I've, I found it actually kind of difficult to take, uh, a current system and get it to, uh, match my style properly. Um, so I didn't really use it for the book. I, I occasionally use it in my work for phrase finding. It's very good if you're, Sort of stuck for a phrase, and you know, if you're stuck for a word, or if you're overusing a word, you can always go to thesaurus, and that's very quick on, you know, on Google, you just go to thesaurus and look it up. But if it's not a single word, if it's actually like a kind of…

AI assessment note: “I did not, I really didn't use it for the book at all.”

Q it's worth as a techno optimist is that actually, uh, I'm in the camp of people that say, well, in a world where a lot of people are lonely, it's actually better than nothing, and actually maybe, uh, AI will be able to listen to To you, you know, whether nobody can, but like, I think you have a different view. You view it more as dehumanizing. Is that fair?

A Yeah, I'm worried about it. Um, I am concerned. I mean, I think you already see this a little bit with, uh, users of things like replica and, um, to some extent, character AI. Um, and there's an attractiveness to these, these companion chatbots. There seems to be a certain subpopulation that really enjoys interacting with them. Uh, but to the extent that I, I worry that they're sort of, they become very addictive. It becomes very easy to kind of have this replace actual human companionship. And I think for people who are already lonely, you, you could say, oh, well, this is better than nothing. These are people have nothing right now. Isn't this better that they have this outlet and, and something to talk to. On the other hand, I do think they, it tends to be a crutch and people will then not be, they'll say, I'll have this. And now I have this outlet, um, for my emotions and my expressions. I can unload my thoughts about the day. Um, and they are not going to seek out a real human relationship, and I think that's detrimental. I think, well, we should be, there are other people who say, well, these things are good practice for real social interaction. I think that's interesting, but then the model, I think, you know, and then this, the, uh, the companion Shabbat should, should prompt people to go out and actually, okay, now we've, we've tried a little practice, now go out and s…

AI assessment note: “Yeah, I'm worried about it. Um, I am concerned.”

Q And a little bit to the earlier point about, um, UI and other, um, you know, one of the many sections and tidbits, uh, that I found super, um, interesting that really caught my attention was this idea of, um, automation bias, uh, versus automation surprise. Can you talk to that?

A Yeah. So I think as we design these systems to work within enterprises as, as co-pilots, um, there are those, these two, uh, human cognitive biases around technology that we really need to be careful of. And one is automation bias. That's the tendency of people to become overly reliant on technology and to assume it is right, even in the face of data that would, you should prompt us to think, uh, it's not right, that it's making a mistake. Um, and the other is automation surprise, which is when an automated system, uh, goes haywire. It often takes humans a lot longer to figure out what's gone wrong and to recover back to some sort of manual process, um, than it would in, in the case of a mechanical failure, for instance. Um, we seem to have less, um, less intuitive understanding of how these systems work, and I think our tendency to assume that they're right, uh, and, and the, if you have an automated system that is mostly reliable, um, it is a problem when they fail because it is, it's an unusual event, and unless you've been specifically and, and very well trained to kind of deal with that failure, Um, there's a tendency to become very flustered by the, by the failure, and I, I use the example of, uh, aviation, where, which has had more automation for longer than, than most of them.

AI assessment note: “one is automation bias. That's the tendency... the other is automation surprise”

Q And so speaking of skills, uh, there is a section, uh, called, uh, Winner Takes Most, where you talk about AI's tendency to help the best performers and across professions. So, you know, you're going to end up with a, a star system for lawyers, for doctors, for, do you want to talk to that?

A Yeah, sure. I think that's, uh, gonna be one of the effects we see of this technology. Um, is that in sort of industry after industry, it will create a kind of winner take all economics, um, where enhanced by this technology, you have sort of the stars, um, of a field being able to charge a real premium. And in part that premium will be based on sort of how much better they are than, than the average person is assisted by, uh, this kind of AI co-pilot technology. I think one of the things about the AI co-pilot technology is it tends to have the biggest impact on kind of lifting less experienced people Uh, up to the average performance. So then, but it, so that means that I think there'll be many more people who can kind of perform at an average level and their ability to commit, and, and any sort of market economics, you, if you increase the supply, uh, the price tends to drop, right? Uh, assuming demand is constant. So I think what will happen is the, the price of the average value of legal services or of accounting, um, of marketing will go down. But if you can perform above average, I think people are going to put a real premium on that, because that's still going to be a very rare skill. And the fact that you can, you can value your human labor, um, that much more highly, I think, in, in a world where most, most people are, are able to achieve average, um, the ability to ac…

AI assessment note: “it will create a kind of winner take all economics”

Q Actually, to, to the extent that you can, uh, share, uh, in the process of writing the book and having all those conversations, um, Who were like some, the most interesting folks and the most interesting conversations, the more surprising ones, the, or the ones that you enjoyed the most?

A Um, yeah, sure. Um, well, I talked to, I don't know, I talked to like hundreds of people for the, for the book. Um, I thought I had a, had a really good conversation with Demas Asabas from, from DeepMind, Google DeepMind, uh, for the book. Um, he was very enthusiastic about agency and sort of, and I think in part because agency When you start thinking about training, uh, AI agents, it, it gets back into, um, potentially a zone for reinforcement learning, which was what DeepMind was always known for. Um, and so maybe, uh, he's excited about it because it, it will allow DeepMind to sort of draw more on its, its, uh, pedigree and its, uh, kind of background, uh, and expertise. But, um, yeah, I thought that was really interesting. I, I really enjoyed the conversation, uh, with some of the artists I talked to actually for that creative chapter. I mean, it was really interesting how people were using and incorporating AI into their creative process. Um, there was, like I said, there was this painter I talked to, sorry, he's a photographer actually, and he takes these huge kind of monumental digital images, but then he feeds them through an AI system that creates interesting effects. And then, but then he applies a lot of sort of Photoshop skills and editing on top of that. And then also the printing of these, uh, very large, um, photographic works and the way they're displayed is aga…

AI assessment note: “I thought I had a, had a really good conversation with Demas Asabas”

Q Great, great. Let, let, let's, um, Definitely going into this in a, in a minute. Um, maybe to help from the, the conversations of the book, um, is called Mastering AI. Uh, why that title? What is the, uh, underlying thesis behind the title?

A Yeah, so I, so the book's called Mastering AI, A Survival Guide to Our Superpowered Future, and, um, I wanted a title that would capture, I think, to some way to play on people's anxieties, that we're going to lose control of this technology in some way, that it, uh, would pose a risk to us, um, and, and to say that essentially, no, I mean, this is a technology we can control, we can master, um, and I'm going to hopefully illuminate some ways I think we can do that, um, And then the subtitle, I mean, a survival guide to our superpowered future, I really wanted to capture both elements of the, the kind of concern that this is a very, very powerful technology that, um, has inherent risks, but I think, you know, there are ways to mitigate those risks, and I think if we can mitigate those risks, then we really can give us all superpowers. I think it can do, you know, incredible, um, incredible things to, uh, to make us more productive and to be really transformative.

AI assessment note: “I wanted a title that would capture, I think, to some way to play on people's anxieties”

Q Talking about co-pilots, which you have a chapter about, you say, um, we're all middle managers now. What do you mean?

A Well, I think that's one mode of looking at these things, and, and, um, And there's been several other people have written that this, this phrase that we're, it's turned us all into middle managers. I guess it's the idea that, uh, the, uh, co-pilot technology can be used as a kind of junior colleague, um, uh, junior assistant that it will do the sort of first draft of things and, um, we will, we will supervise its output and, um, that puts us all in a little bit of a management role where we're not doing the, the work ourselves, but sort of overseeing the work and supervising the work and correcting, uh, course correcting other, you know, The, uh, the entity's, uh, efforts. Um, so I do think there's a way in which you can think of co-pilots in that way. But as I point out in the book, there's also this other way of thinking about co-pilots, which is, ah, you know, they can both be a kind of junior colleague, but also you can prompt them to act as a kind of senior mentor. And instead of having them do the first draft, which you then oversee, you could do the first draft yourself and then ask, uh, the co-pilot to sort of critique what I've done, or, you know, here's the sales pitch that I'm thinking of using and You know, what would you, what would you improve on, uh, from that pitch? I think, um, they have tremendous possibility in both roles. Um, either way, I guess it puts the…

AI assessment note: “puts us all in a little bit of a management role”

Q Yeah. I mean, you, you, you go into, um, You know, a lot of this, and, um, I mean, pretty much the worry is, like, AI is gonna make us dumb, right? Is that, is that fair?

A Yeah, yeah, exactly. I, I do worry that AI is gonna, gonna hurt our human intelligence. Um, that, again, that over-reliance on these things can, can make us a little bit, uh, less sharp than we used to be, and certainly lose some cognitive skills. And I think there are examples of that from previous technologies, as I, as I try to point out. I think it really is true that Google has hurt our memory. Um, I think people memorize far less than they used to. And I think generally that's fine, but I think you have to realize that that's a trade-off. Um, because I think Google and the ability to just have any, you know, the world's information at your fingertips like that is a, is a huge advantage, but you have to realize that people memorize less. And then there's some interesting studies on learning and memory that, you know, the, the, the more you actually know of a subject, uh, and have memorized about it, that tends to help your, you to achieve kind of breakthroughs in that field.

AI assessment note: “Yeah, yeah, exactly. I, I do worry that AI is gonna, gonna hurt our human intelligence.”

Q All right. So maybe taking some of the sort of Big topics and just like doing a quick, um, you know, review for some of them. Uh, you know, we can start with anyone you'd like, but like you go into AI and workforce, AI and art, AI and science. Maybe let's start with AI and science. What's the, what's the high level?

A Oh, so in general, it's the area I'm probably most, uh, enthusiastic about AI is potential, uh, effects. I think, uh, AI is a sort of super tool for science. It's going to, uh, I sort of We've compared a little bit in the chapters, the microscope or something for biology or the telescope for astronomy. Um, it's a thing that is going to become an essential tool for science. Um, the idea that you could be a biologist without a microscope is sort of anathema. And I think in within 10 years, the idea that you can be a sort of any science in any science, uh, without using AI models to help assist what you're doing will also be anathema. Um, I look a little bit as particularly the drug discovery case. Um, I think what's happening with AI models for Genomics, uh, for chemistry, for enzyme and protein design is incredible. And I think, uh, we're going to very quickly start seeing the results of that in, you know, better drugs designed faster at a lower cost. And I think that's, you know, tremendous impact, but I, I think it's sort of across the sciences. I think it's in, in chemistry and in material science, uh, where we're, where I think AI is going to help us find, uh, new materials that will have transformative effects on our lives, help us be more sustainable. Um, I think we're seeing it, uh, but even in, Even in the social sciences, there's very interesting things about using sort…

AI assessment note: “AI is a sort of super tool for science.”

Q Yeah. So we've alluded a little bit in the conversation about the need to do something, uh, from a policy standpoint. What is it? What, what can we do? Is that more regulation? Is that self-regulation?

A Well, I think we need actual regulation, like government regulation, and I think it has to be Both industry specific regulations. So looking at the use of AI within particular industry verticals, and there are probably very specific things we need to do that are very different in banking than what you might need to do in medicine that are different again than what you would do in e-commerce. And I think, um, so I think we need some industry specific regulation around some of these things, but then on some of the higher level risks, I do think it would be useful to have a kind of AI regulator at the federal level that has enough expertise to kind of look over the shoulder of what these, uh, the tech companies are building and Make sure they aren't taking undue risks. Um, and that we aren't accidentally sort of tripping ourself into a situation where, um, the models do pose more of a kind of catastrophic risk. Uh, I, I think that's sensible. Again, I don't, I wouldn't want that agency to be funded, uh, more than, uh, you know, the DOD or something, but it's like, but a little bit of money towards that makes sense to me. And, and having some sort of outside independent entity that can look at what's happening in the, in the, uh, industry And draw some clear boundaries. I think makes sense. I do not think that companies can be trusted really to do this completely on their own. I th…

AI assessment note: “Well, I think we need actual regulation, like government regulation”

Q You and I, um, and presumably there's an intermediary risk, um, as we linked AIs together, uh, do you have any thoughts on sort of agentic AI or systematic AI, uh, from that perspective?

A Yeah, well, I think the, the more power we give these systems to actually, uh, take action in the world, obviously there's more risk. So I think as we move to agentic systems, um, the reliability becomes a bigger factor. Um, I would not want, it's interesting because I think we're very close to some, uh, some system with some form of agency and yet, you know, our hallucination rates on current LLM based systems are still relatively high. Um, you know, even with some of the best techniques, you know, if we get, if we get hallucination rates down to like three percent, which is considered very good. Um, I don't know if you're, if, if three percent of every time you send it out to buy something for you on your behalf, it gets it wrong. I don't know. It's like a fairly high.

AI assessment note: “as we move to agentic systems, um, the reliability becomes a bigger factor.”