Business, Innovation, and Managing Life (October 4, 2023)

Sam. Hello, everyone. Welcome to another episode of Q and A about business innovation and managing life. I must apologize that I'm at the tail end, I hope, of a cold. So I'm a little bit subdued and probably sounding a little bit more congested than usual. All right, let's take a look at some of these questions here. Let's see. Yoshimando asks, what do you think is the most important aspect to focus on or dedicate the most effort to when running a business? You know, businesses are different. All kinds of different businesses have different objectives, different personalities. You know, some businesses are very driven towards kind of sales, you know, do everything the customer wants. Some businesses are more driven towards make great products, and then, you know, we build it. They'll come. Type, type approaches. It really depends a lot on what the nature of the business is, what one should focus most on. I think the thing that some. Depending on type of business, if you have a service business, for example, it's about kind of dealing with customers and their expectations, selling, making sales to the customers, then having the right people in place to deliver the services. All those kinds of things, really very different. And I think different people have. Different. People who run businesses have different things that they're good at and that they concentrate on. And I think the real trick is to figure out what things you can successfully delegate to other people, what things are worth you paying attention to yourself and what things. When to notice that things have gone sufficiently off track, that even though you thought you delegated it, you have to kind of parachute in and do something yourself. I mean, I think it can be the case people will do things like they'll delegate strategy. To me, that's a completely bizarre idea. I mean, kind of the people in charge of the business are the ones who should be figuring out the strategy, but there are people who run businesses and even successful businesses where most of what's involved is just keep the machine running. And the strategy, the doing of new things is something where you hire a consultant, you do this, you do that, to kind of wheel in strategy. I mean, for somebody like me who's very much involved with sort of product and strategy and so on, it's. There are other aspects where I'd like to just kind of wheel people in to do those aspects, because those are things that I, I would like to, would. I don't. I think. I think I can delegate, so to speak. So it's. It's. It's complicated. I mean, I think that the. The thing one has to understand is what, what is the nature of one's company and then what is one's way to contribute to that? And I think people sometimes, when they're running businesses, take the point of view, oh, I've hired an expert to do this or that. I'll just let them do this or that. And, you know, sometimes that will work great, but sometimes that expert will kind of walk right off the cliff. And they're just, they don't really have the same commitment to the company. They don't really have the same overall vision. They're just doing the, you know, turning the crank in the particular way. They, they, they know how to turn the crank. Now, I have to say, in my own experience running businesses, there will be occasionally people where you think that's what's going to happen. They're just going to turn a crank the way they've done it before. But they'll surprise you and they'll do fantastic things that will really fit into the business that you have. And other times you'll discover that people are just like, well, there's a crank that they used to turn some other way somewhere else, and you thought they were going to do something different. And it turns out they just turn that old crank and it doesn't really work in the business that once actually built. So a few comments at least. Let's see. Marco says you're a speaker at the all in Summit 2023, which was a conference, they say, aimed mostly at venture capital folk. What were your impressions of the summit and its attendees? Did you attend the parties at the All In Summit? I really enjoyed the Orland Summit. It's a conference that's sort of born of a, of a, a podcast which is kind of a discussion between four people. I have to admit, I was, I hadn't seen this podcast, but I mentioned this to various people and they were like, oh yeah, that podcast is really good. And they, it was really a, it was a very unusual thing for me because usually I, I'll go to conferences and I'm really bad at sitting through talks and so on. But at this point, conference, the, There were, they weren't really talks, they were just interviews and they were much more engaging than sort of typical talks are. And the people were interesting people. And I sat through almost all of them and I, I found the, the. I, I went to the parties and I was sort of kind of. It was interesting because a lot of times when I go to conferences, I know a lot of people and so I'll know a good fraction of the people who Are there this case, I knew almost nobody, but it was very asymmetric situation because they almost all seem to know me. And that was something I was actually, frankly, a bit surprised by, because these are people who are from all kinds of different sort of industries, walks of life, levels and so on. And they're sort of pulled together, I think, by an interest in kind of thinking about things, actually thinking about things and kind of understanding things from an intellectual point of view. And I suppose that demographic is one that intersects quite a bit with folks who follow some of the things that I've been up to. But anyway, it was. It was interesting, and I think I had a very wide diversity of conversations with a very wide diversity of people at this conference. And my goal in all of these things is to learn lots of stuff and talking to people from all these different industries. And my usual question is, what do you do? So to speak? And then you will typically get into some really pretty interesting conversation. I found that, at least in this crowd, an interesting conversation about things where you could never predict what the thing would be. You know, would it be something to do with greenhouses or something to do with the legal system or something to do with venture capital for cryptocurrencies or something. Very hard to predict what it would be, but it tended to be something interesting where people were actually thinking about what they were doing. So it was. Was fun for me, unexpectedly fun, actually. Aqua asks, do you get demotivated to do things that AI might be able to do in a fraction of the time in the relatively near future? Well, I think that I've spent my life trying to automate things that I do. And my attitude has always been, if I can automate it, then I've kind of zeroed out that cost and I can go and sort of build further up in the tower, so to speak. And I think it's something that people. There's a certain kind of their activities where it's like, well, you can just do it. And you're kind of like, you spend the hours and you get something done, and it's kind of satisfying. But for me, if I can zero out those hours by having automated the thing that I'm doing, I just always consider that a win, so to speak. I don't think the kinds of things that. I think it's a general principle that if the things you're doing are things that involve kind of human choice about what to do, that is, you're going to invent something, you're going to invent a strategy for what you want to do, you're going to invent something that is kind of a new direction. Those are things that are very much for us humans to choose. There are an infinite number of new directions in the computational universe of possible directions that can be taken with computers and so on. But it's sort of up to us to choose which direction we decide to go in. And that's something that's always going to be. It's like, okay, we're going to make a company. Okay, AI, go and make a company. What should the company do? The AI has no idea. It's like you've got to pick what your company is supposed to do. That's the place where kind of humans are necessarily valuable because there are these infinite set of possibilities. And it is the sort of, the human aspect is to pick which one you want to pursue. And so I think an awful lot of what I do is kind of sort of figuring out strategically what to do. And the thing that I've tried to build for myself is the best kind of superpower tools to take the things that I think of doing and actually implement them. And mostly what I built is Wolfram language for the last 35 years or so. 40 years altogether. 40, no, 44 years altogether. And the goal there is set up a framework of formalism where if I can cast my thoughts into that kind of framework, if I can think about things computationally in computational language, then I've automated the process of getting those things done. People say, well, gosh, isn't it amazing that the AI can write this low level programming language code? It kind of amuses me because for the last 30 something years, the goal that I've had is to just encapsulate that sort of boilerplate code in the functions and the kind of human level functions that we provide in Wolfram language and so on. And so it's kind of like, yes, there's just a function that does that. It's already automated. The fact that there's a slab of low level code that you can sort of maybe get written by an LLM or something is kind of like, that's, well, yes, you could do that. I don't think it's really the right way to do it. But the general idea that you can take something and automate all of that, kind of all of that manual labor, in a sense, well, that's the thing I've been pursuing for the last 40 something years. And so to me, as I say, whatever I can automate is something that I can Then use as a platform and build higher and higher. And it's worked really well for me at least, and I think it can work really well for other people. If you have decided that you will not take any initiative and you're just going to do the programming task that you were assigned to do, then, yes, what you're doing will probably be automated. People like me have been trying to automate that and succeeding in automating that for decades. And AI will do more of that. I think, again, if your task is one that is just sort of turning a crank, then yes, you can expect that it will get automated. And I view that as a good thing. I view that as a thing where we humans get to do the things that kind of are much more unique to us in figuring out what to do and building on something where we've automated away kind of the more mechanical parts of what's going on. I will say that there are cases where there's a question of, should I do this project or should I just hang out for another decade and have the ambient technology of the world reach the point where that project is easy to do. I'll give an example back in 2009, 2010, something like that. Well, we had built over the previous 15 years or so, we. We built a bunch of good image processing capabilities in Wolfram language. But those image processing capabilities were very much things at the level of detect edges in an image, change the colors in an image, these kinds of things that were really kind of talking about things at the level of the structure of images. But there's also the question of, okay, we've got these images, find all the zebras in this image, find all the cars in this aerial photograph, these kinds of things. And I had thought, well, gosh, you know, using the kind of curation technology that we built for WolframAlpha, where we're kind of curating all these different things about the world, we should be able to curate kind of the set of sort of, at that time, kind of fairly hacky image processing operations that would allow you to, for example, pick out cars, an image or something like this. And so I was all ready to do this. Fortunately, it didn't get done because a couple of years later, kind of deep learning rolled in, successfully doing a bunch of image identification. And that project would have been a waste of time. That project would have been kind of overrun by what then became possible with that technology. It was a fairly rare case for me where I kind of thought, look, by the way, I was not without having Done my homework. I mean, I had. I knew quite a bit about the whole neural net sort of development over the course of a couple of decades, three decades by then I had talked to some of the kind of leading people in the area and they were like, well, we think we're going to be able to do it with neural nets, but it might be decades from now type thing. And so it was not a no homework situation, but it was one where, as it turned out, the sort of ambient technology, kind of the tide of ambient technology came in and made that particular initiative not worth doing. Now, usually when I think about doing projects and so on, I'm pretty aware of whether there are sort of things coming that will make this project not worth having done. I mean, for me personally, maybe it's a piece of egotism. I like to do projects which I think nobody else will do in the foreseeable future. And so if there's something where there's kind of a tide that's coming in that has nothing to do with me, that's a place where I'm not going to really put a lot of effort. I mean, of course there are plenty of things we've done. Like amusingly, before the Web existed, we were interested in distributing what were then Mathematica packages. We had this thing called Math Source, which was partly a thing that would. We'd send out CD ROMs. But another part was a kind of a telephone direct response thing where you did various things where you would, how did it all work? It was a crazy system that was built for a time before the Web existed and we put some effort into it and it was somewhat useful, but then again, all sort of washed away by the arrival of the Web. As I say, in my life, I've been perhaps fortunate, perhaps, perhaps put some judgment into not doing things where I think there's about to be kind of some tidal wave of other development that overruns it. And I think in terms of sort of the current round of AI and LLMs and so on, my feeling is very much like what happened in 2012 and so on with image recognition kind of, we see an amazing and very scientifically interesting kind of jump in capabilities that happened, well, about nine months ago now, a little bit more than that with the arrival of ChatGPT and so on. And this is a big jump. It's very interesting. I don't think that there'll be, you know, the fact that there was this big jump doesn't mean that things are going to go on jumping up that quickly. You Know, going forward, I think the big jump already happened and I think now the challenge is to understand how to take what's there with the reliability and unreliability that it has and fit it into use cases where one can make use of the things that it does well and not get skewered by the things that it doesn't do well. And part of that story, which is one we've been working a lot on, is integrating computational language with LLMs. Having the LLM be able to use computational language as a tool, being able to let humans read the computational language that the LLMs produce so that they can say, yup, you got it right. This is terrific. Just like when you get search results, for example, you'll see a bunch of results and some of them might be right and they're exactly what you want, and some of them might be way off and you just ignore those. And so similarly here, sort of surfacing from the LLM, here are some pieces of computational language code that might be what you want. Okay, great, that's what I want. On I go. And that's then a success. Even though some of the other things that might have been surfaced by the LLM are kind of off track. That's okay in that kind of use case. Let's see. Gosh. Satoshi is asking, what's your take on privacy, especially for digital services and devices, regarding companies using data to manipulate people and similar things? Well, you know, there are all these complicated trade offs. I mean, the fact is I'm sort of amazed that the whole kind of ad based way of monetizing things has worked out. I didn't think it was going to work when the folks starting Google went in the direction of we'll spend a lot of money, we'll give this thing away, we'll support it with ads. I didn't think it was going to work. I mean, of course it had worked for things like television before, but it was very counterintuitive to me that that would actually work. And of course it is very difficult in a situation like that when the customers, the people who are the users, aren't the people who are paying the money. You have this kind of complicated sort of skew set of incentives and that skewing has caused all kinds of issues. I mean, in my own world, I've tried pretty hard to have it be the case that the things from which we make money, the organizations, the people from which we make money, are the people who are actually using the products that we make, not some kind of skew situation because I think it provides one a much simpler, much sort of more direct set of motivations about make the products better, provide better value to the people who are paying one for the value that one is providing. And I think as we maybe see some paid search engines come in, it'll be interesting to see the extent to which they're able to provide different services because they are, their, their motivations are very much keep those subscriptions coming, keep the people who are using the search engine happy, not keep the advertisers happy by showing their ads to the right people or as many people as possible and so on. So it's, I mean, to me this kind of complicated kind of skew incentives is, is, you know, leads to all kinds of issues, I think. And by the way, there are many, you know, in, in, in many industries there are skew incentives not just because of the business models, but also because of the role of government and a lot of these things. There's a lot of complicated situations where people, where it's not clear that we're looked at sort of from a big picture point of view, it's complicated to see what all the levers that are in operation and what results that'll produce. Now in terms of, of privacy and so on, I mean, societally it's kind of complicated. Give you an example. Let's take medical records. Let's say you take the point of view all medical records are just completely private. Nobody's ever going to get access to anybody's medical records. Well, that means that if there are sort of health results, health discoveries that could be made by looking at a million medical records, you'll never make those discoveries because you never get to see, nobody ever gets to see those records. On the other hand, you can take the point of view that. And so then what one ends up with is sort of these technical ideas and issues of, you know, can you look at the million in aggregate and not invade the privacy of the one, so to speak. And there's a whole bunch of sort of mathematics and algorithms around trying to do that. In the end, it's tough to do because in the end there are all kinds of things where you fuzz out the data and you look only at averages and things. But then you worry about, well, what if you have five different groups and they all look at different slices and they all collude with each other and when you take the intersection of those slices, they've honed in on one person, all those kinds of things. I think it. So that's a case where there is clearly societal good in exposing some of that data, but there is potentially individual harm in exposing the data of the individuals, so to speak. And so, you know, one hopes that there are technical solutions to that and there are partial ones. They're not really very complete, I think. And sort of there's a question of where should the needle go in terms of sort of the societal good and harm and individual good and harm and so on. I think it's difficult to say. But, you know, there are other places where, you know, there are cases where sort of the breakdown of privacy depends sort of what's the downside problem? And there are some potentially horrifying downside problems. And then what's the sort of. And how does that, you know, people trade off, oh, I'm going to get this free service. Oh, but my data is going to get used to and I'm going to get fed all these ads and so on. And if you look at what people actually do, an awful lot of people just say, hey, I want the service for free. I really want it free. Well, somebody's got to pay for it. So how is it going to get paid for? Ultimately it's going to get paid for by spending your time seeing the ads or potentially, you know, your money buying the products that are being advertised in the ads. So I, you know, and people are, I would say the, you know, this pendulum is swung in different ways of, you know, do you just tell people, look, you know, just pay for the services you're getting. And that way we don't have to worry about all this kind of, you know, grabbing data and all this kind of thing. I mean, in our own business and company, we have been very privacy oriented because it just doesn't for our business. It's like we're about selling things to people who are going to use them. And it doesn't have anything to do with our business to kind of invade their privacy, so to speak. It's just irrelevant to our business. We're not going to do it. But if we had a different kind of business which involved, you know, providing ads and all this kind of thing, then it would be critical to our business to kind of try and find out as much as possible about the people who are sort of in our orbit. So, you know, I think it's a, it's ultimately it comes down to sort of business models and also to what people are prepared to do. I mean, if you say, well, gosh, you know, we're going to make this business where people pay for the stuff that they get value from. And everybody says, no, no, no, we don't want to do that, we just want everything for free. Well, if you want everything for free, as I say, at some point somebody has to pay for it. So you're going to wind up with something where you're essentially paying with your time or with your data for what's happening. Let's see, Let's see. Pam is asking, could you imagine the web being washed away as it was by other technologies? You know, I think there have been various sort of threats to the web, so to speak, which haven't panned out. I mean, for example, it, the, you know, the mobile apps for instance, were in some sense a threat to the web. If you could get everything you want through a mobile app, why would you go on the open web where you kind of don't know what's happening and go there? Why would you not just sort of get everything through your more curated sort of walled garden of mobile apps? Well, that didn't completely work out. And you know, the web has become this kind of just sort of place where the information of the world gets put out there and you know, what will happen to it in the long term. I mean, I suppose if we look at history, things like books, they've been around for a solid, well, in printed form for a solid, what, 500 years or more? Six, 500 years. A little bit more than 500 years. And they are, and the mechanism of you publish a book, you print a book, you make lots of copies of it and you send it around the place that's been a surviving technology for a long time now. Has it been washed away by the web to some extent? Not completely. Some form factors of sort of. I'm going to make something which is a, a book size lump of information that survived even with the web. So I think, you know, what I would expect is that the idea of, you know, you go to the domain, you go to the website that'll probably survive. Just like you pick up a book that, just as you might have done 500 years ago, I pick up a book about this thing, you know, I go to the website where I can find information about this thing. Now what is that website? Is it some kind of thing in some augmented reality sort of presentation? Is it something where lots of the kind of navigation and so on is very different? It might, you know, it may evolve. I think books are a good example because books honestly haven't evolved that much in hundreds of years in terms of Their, you know, some technologies got easier to do, but you know, the Woodcut illustrations of 400 years ago are not, you know, it's easier now to sort of intersperse those with the text and so on. But that's not a huge change. And the, you know, I would expect that the sort of the, the user interface of the web is something that's going to be around for, for a long time. And although it's, it's particular presentations will probably change. I mean, you know, and, and there are all these things that are so kind of hard to predict. Like for example, you know, you always think, oh, people are going to want the highest resolution screens, the most kind of the biggest way of presenting data. But actually more and more people are like reading everything on their phones. It's like, how can, you know, the convenience of having it in your pocket exceeds the kind of the richness of having that giant, you know, four foot wide screen or whatever. And it's sometimes hard to predict how people's preferences will fall out in terms of those kinds of things. But I think the concept of a website like way of presenting information with a domain you go to, to sort of get there, that'll probably survive. I mean, I think it was a question of whether everything is sort of based on search and you just sort of type in keywords and that's how you get to places or whether there's a definite notion of kind of a place on the web. I think the place on the web thing has survived people. Yeah, I mean lots of details, but somehow these things, if you look at the lessons of history, there are things that survive. Once they have a certain level of absorption in the world, they survive in one form or another for a really, really long time. And I would expect that for the web. Let's see, Another question from Prab. How has the concept of intellectual property evolved? Is land a good analogy for intellectual property? You know, the concept that intellectual property is kind of a thing that you can own has, it's a, you know, it's a comparatively recent phenomenon. I mean what happened back in the day was sort of individual people, families, guilds, whatever had certain kinds of secret knowledge. You know, only the Freemasons or something knew how to make, you know, how to set up, how to do masonry construction for a, I don't know, a cathedral or whatever it was. You know, there was some sort of group that had what amounts, what today we would call sort of trade secrets that were about how to do some particular kind of thing. Only this particular Group knew how to make swords that were strong in this particular way, knew how to make this particular kind of food, whatever else, you know, the idea of kind of a trade secret held within kind of a guild of people, maybe a family, maybe some more organized kind of externally organized kind of association of people was. Was. Was a thing for, you know, that's been a thing for a very, very long time now. The idea that you could expose how things work and have people not kind of rip off what you were doing, that's a much more recent thing. That's a thing that kind of relied on people believing in kind of the rule of law. And not just. It's by the time when you're protecting yourself by having a castle with tall walls and ramparts and all that kind of thing, and drawbridges and moats, then that's not a time when, if you've exposed your core piece of intellectual property, when you can expect that nobody will steal it, so to speak. I think it was really until probably the 1700s or something when people started sort of believing enough in the rule of law that it was possible to expose your intellectual property and still have the government pretty much guarantee that intellectual property will be maintained for your benefit and not just. Not sort of taken by the people. Now, I think, if I remember correctly, the history of the patent world, I think it starts in the reign of Elizabeth I. And I think it has to do with particular industries where there were kind of where there was a goal, where in a sense it was the crown, the Queen, in that case was, or effectively the government was giving a monopoly to somebody and saying, you're the only people who can do this or that. I mean, there were many such monopolies. I mean, I know that the. I don't know, like in England, the printing of the Bible, the most popular book, you know, only the Oxford and Cambridge University presses were sort of given the license to print the Bible. And, you know, that was a kind of a common thing that it was in sort of the gift of the monarch to say, okay, you have the ability to do this thing, which in principle, other people could do, but if other people tried to do it, the government will come down on them and consider them criminals and things like that. And I suppose another place where that happened was with money. The whole question of whether you have to, in order to have money as kind of a way to run an economy, you have to be able to defend the validity of the money. And so even though everybody can see what coins look like, it had to be illegal. And the government had to police the idea that, you know, you couldn't make your own coins, you couldn't be a forger, so to speak. Even though that was a, you know, it was a, was a common issue back in, in past centuries and so on. So I think, but I think, you know, the, the idea, I think things started with kind of the government giving monopolies to people, which is of course something that still happens in, in almost every country. And you know, the, the ways in which it happens in a country like the US are very different from the ways in which it happens in countries that have a more kind of, well, basically more corrupt kind of system, but a system where there's more kind of, where there's. But, but in the US it happens too. And plenty of places where, you know, the government will say, okay, only companies that have this regulatory compliance can do this particular thing. And in order to have that regulatory compliance, you have to be of a, of a, of a great, of a very large company or something. Or sometimes they're just individ companies which are the only ones that sort of satisfy the things which the government has decided are going to be kind of the regulatory compliance thing. So you know, it happens even in a country like the U.S. but I think this idea that on the part of sort of smaller inventors that there should be a way to protect intellectual property, that's a thing that I think by the 1700s that was a thing in, in England, I mean in the U.S. in the Constitution there's a line I think was put in by Jefferson who, I think intellectual property lawyers have told me that he, Jefferson himself was somebody who had gotten some, I think wine making methods or something from some folks in Italy and had kind of just taken those and was going to use them. But he was the one who sort of added to the, to the Constitution the, the statements about, let's see, to provide inventors, people who work with science and the useful art, the useful arts, the kind of, I can't quote it, I'm afraid in verbatim but for a limited time, the exclusive right to the, to benefit from their inventions. So that was kind of the, the line that originated patent law in, in the U.S. interestingly, that line mentions science. But science has never been protected by any form of intellectual property protection other than kind of the folk protection of people saying, oh, you plagiarized that or whatever else. It's not a, it's not something unlike patents where it's kind of like I made this Widget, and I'm describing a widget, and this widget is useful for at least something. And you can't make a widget that's just like my widget for 20 years or whatever it is, 17 years. That mechanism is. When it comes to scientific ideas, there is no analogous mechanism. And I think the patent law is. I mean, like so many of these things originally it had sort of good consequences, it had bad consequences. People ended up figuring out how to game it and have done all kinds of crazy gaming things with it that, you know, sometimes there have been moments when there's sort of huge, sort of huge kind of torrents of, you know, patent trolling and things like this, which then eventually the pendulum swings back a bit and it becomes more difficult to do that and so on. And you know, I think it's sort of an inevitable thing that it's sort of. It's gamed in various ways and if it's game too much, the system shifts a bit. But I think in, I mean, my own feeling about, you know, patents are complicated to deal with, particularly if, you know that they are something in sort of the arsenal of value. And sometimes people will say, oh, there's this company, it has lots of patents, so that means the company is more valuable in some definition. And, or, oh, you've got all these patents so you have a very defensible position. Nobody can kind of, you know, kind of stomp on your turf. Well, if the, if the stomper is a very big company, they're going to end up being able to sort of squiggle around what you have. And even if you get into a pitched battle, the chances are the big company will defeat the little company just by the mechanisms of the legal system having nothing to do with whatever one might consider the sort of true merits of the situation to be. And that as a realistic matter, the question of what you should do with patents and when it's worth getting them and when it's not worth the cost and what do you do with them once you have them? And do you go out and tell people who you think might be infringing them? Do you make a big fuss about it or not? And all those kinds of things. I have to say my own experience with this has been that at least with the kind of life I tend to lead, patents are not a big factor in it, that it's really something where even though I have lots of ideas and I could turn them into patents all the time, I don't know what I would do with Those patents, it would be kind of more like patents for wallpaper, so to speak, than patents that I know what to do with. Because most of my sort of mode of operation at least is to actually take ideas and try and make them into real things myself. And I tend to feel that a large part of the value is in making the ideas into real things, not the raw ideas. But you know, there are other forms of intellectual property like copyrights and there are all kinds of, all kinds of issues in today's world about, you know, can AIs read the copyrighted material and can they, you know, are their brains, are their minds effectively like human minds where it's kind of okay for them to read it or not and, and all those kinds of things. But I don't know what, let's see. I mean, I think that the concept, I think this original question was, is land a good analogy for intellectual property? I don't think so. I mean, land is very definite. It's kind of like, you know, where it is. It's got these geocoordinates intellectual property. It's like you've got one patent, you've got another patent, you've got this invent this widget, you know, how does the widget relate to these patents? Well, you could kind of ask an LLM that, you can ask lawyers that. It's a big messy process. I mean, if you put, if you have a, a, you know, a marker on the edge of your property and somebody sort of starts building something inside your property or whatever, you, you know, that, that they're, you know, it's very clear cut thing. It's not something like it doesn't have the same kind of muddiness of, of typical, well, at least patent type intellectual property. I would say that the copyrights, well, they're complicated as well because it's kind of a question of, okay, if you've got the literal slab of words, you know, big slab of words. Yeah, you know, it's pretty obvious that that's covered by some kind of copyright. But when you've got the slab of words processed through an AI, you've got the, the picture that's been modified and this and that, you know, it's kind of complicated. And I think the, you know, there's a question of is intellectual property protection a good thing? I would say that the answer is, you know, the answer is, so far as I can tell, basically yes, it is necessary. Certain forms of intellectual property protection are necessary as a motivation for people who are going to create things and that was what the original line in the Constitution effectively said. And it's true that if you think that anything you do is going to give no value to you, it's hard to be motivated to do it. Now, sometimes that value might be, I'm going to give it to the world, but somehow I'm going to gain reputation or some such other thing through doing this, and at least then. Or I'm going to feel good about it in some way. But if, you know, I. And for some people, it's like when you make an anonymous donation to something, it's like sometimes it's purely the internal feeling that matters, not how the world reacts. But in many cases, if the world kind of like just sort of steals what you give it and stabs you in the back, so to speak, then it's not a good inducement to give the world something else in the future. And so I think the. Insofar as intellectual property protection provides a way of having that not happen, it's an inducement to people to actually bother to create intellectual property, which is a large part of what represents advances in the world. I think this problem of intellectual property around science is an interesting one. I mean, I've kind of had the thought experiment. If you're 1687, you're inventing calculus, you're figuring out mechanics, you're pretty sure that they're going to be trillions of dollars of value in the coming centuries that are generated from what you invent. Is there any way that you can get any of that value into the present day? And for things that we invent today, is there a way? I mean, this is a very practical thing for somebody like me. I'm trying to invent things in science and so on, where I know that the more that's done on the science now, the faster the results and the benefits from that science will arrive. But to get more things done costs money. And even though sometime in the future a trillion dollars of value might come from the science that might be 50 years from now, and today it might be. Well, we don't really know what's going to happen. Well, you know, it's all in the future. It's kind of science fiction for now. And the question is, is there a way to kind of take that distant future sort of expectation of large, even commercial payback and somehow deliver it to what one can do today to let more of it be done today so that what might otherwise be 100 years gets turned into 25 years because more work was done? And if work wasn't done. Well, it's going to be 100 years and that's going to be. And there may be plenty of things that sort of society will lose on because, well, there was an extra 50 years when these pieces of progress were not made. Let's see. Oh boy. I'm being asked here from Biggie about antitrust cases against Google and Amazon. Yes, of course I know about them. I think, you know, if you look at the past, for example AT&T, for example Microsoft, big tech related antitrust cases, things have played out differently in different cases. And I think one of the things that often happens with companies is they get to a point of sort of in some sense great arrogance and sometimes kind of very well bullying tactics with respect to other folks in the industry. And you know, what seems to happen in these antitrust cases often is that sort of, the bullying character seems to be kind of seems to be diminished. And actually I think it's a great benefit to the companies themselves in many cases. I think it's probably a, a feature of, you know, you're the dominant player in some industry. You get to be sort of big and arrogant and a bully. And that's something where, you know, well, in the end nobody will work with you. And you know, when there is some new innovation comes from somewhere else where you can try and buy it, you can try and do this, you can try and do that. But fundamentally people don't want to work with you because they're afraid of, of, of your company. And that's, that's a bad situation. And I think my feeling is that that's one of the things that happens sometimes. I mean, it depends on how any of these cases actually develop and play out. And I think it's always a complicated question, what's the point of antitrust and what precisely are the rules for what counts as, you know, an antitrust violation? And how much is that determined by what the law said before and how much is that determined by sort of political forces and how people feel about things right now. And I think the tech industry, you know, from having been, I mean the tech industry has certainly had some, you know, has, has not always been perfect, let's say in, in the way that it's treated the public, so to speak. And I think that, you know, has a habit of, of coming back to bite it to some extent. I mean I, I could, could talk about, you know, every, any industry has a lot of things that really shouldn't be the way they are in the sense that they are the way they are because somebody had the opportunity to do something that was sort of outrageous but. And they got away with it because of the way the industry was set up and the positioning and so on. But being sort of outrageous doesn't necessarily mean it's against the law. And you know, if one's operating in kind of a, under the rule of law, well, whatever is the law is what you get to do and not do and so on. And so I think it's a little bit, you know, if we wanted the, you know, how could we make the tech industry as productive and as responsive and as responsible as possible? Well, I'm not sure, but kind of some capitalist forces go in the direction of that being the case. Some against it, in my observation. Mostly government action. Well, government regulation has a nasty habit of just making sort of dumbing down everything and making and making things sort of, you know, making innovation very difficult, etc. Etc. Etc. But, but I think in. So, yeah, so I consider it complicated, but I think that the main, the main thing that I would say is that with some, you know, just an observation about kind of the dynamics of the industry, in the time I've been involved, what is it, 45 years now, there have been companies where it's just like they're just too friggin arrogant and evil to want to deal with them. And it's whether that's a. And when that kind of breaks, then you end up getting companies that are, yeah, okay, it's a big company and it's, it kind of, it's like much easier to deal with. And everybody is sort of, well, some people at least are happier. And you know, that's a cycle that seems to exist in the tech industry. Let's see. Brady asks, have I ever gone through patent process personally? Well, yeah, if you look up on patent databases, you'll find all kinds of patents of which I'm an inventor. I have to say that. Have I been involved in drafting them? No, I've not. Do I understand them once they're written only with effort? It would help me to just have. I would expect an LLM to help me actually understand them. It's a weird business because you're kind of turning what was sort of a set of ideas into something formalized. But there isn't a way to formalize the ideas of patents yet. The only way to formalize it is with formalized English or whatever language in formalized legalese. It's my longtime expectation that in the future, with computational language and so on There will be ways to describe what one would be describing in something like a patent using computational language rather than using legalese natural language. And, for example, there is at least some chance, although there are many sort of computational hurdles to this, even theoretical ones, that some of kind of the definition of does this patent relate to this other patent? And so on, will be much more definitively answerable in the case where the thing is written in sort of computational language than when it's written in legalese and where you have to interpret that word and so on. But, you know, it's always a strange process because you're going from a bunch of ideas to kind of a description in this very formalized way. Memes asked, maybe ChatGPT can make patenting things easier. Oh, boy. Maybe so. I'm not sure that's a good thing. I think that it's one of these cases where sort of there's. There's often, you know, when. When there's bureaucracy in the world and you think, oh my gosh, this bureaucracy is such a pain. But then you realize the bureaucracy is sort of slowing things down. It's making things. It's making it be the case that when somebody just says, oh, I want to do this, it's like, well, you have to file this thing. And it's, you know, you really have to want to do it. It's not like just on a whim. One day you just say, I want to do, and then you can do it. It's actually often helpful to have this kind of slowing down of. And sort of effort that has to be put in. Are you really sure you want to do this? Put effort in to prove that you want to do it? And I think if it comes to be the case that an LLM can write a patent, I can just tell it my idea and it'll write the patent for me, and the patent office will deal with that patent. I think that will be a difficult time. And I'm not quite sure what. Because, you know, in. In. In a world, let's say there was a world where. There. Where the patent Office is using LLMs as well. And let's say there started to be a billion patents in the world. What a weird situation. I mean, and. And then sort of every. Yeah, I. I can only imagine what, what kinds of weird things would happen in a world like that. Let's see. Yeah. Ki is commenting that maybe ambiguity in natural language is actually a good thing. And certainly the way that a lot of legal work is done in today's world and laws are written and so on. The ability to have some degree of human judgment wiggle room because it was just words and they have to get interpreted by human is thought to be a good thing. Now you could have said, but we adapt to different kinds of things. When we have programs that run and do definite things, there are plenty of tasks where in the past it will be, a person is doing this, a person is doing it with that extra little piece of human judgment, should I do this, should I not? And then we replace it with a program which just deterministically does whatever it does. And sometimes that's not a good idea, but it becomes there's sort of a way of adapting to the fact that that is now a deterministic thing that just works according to the program, so to speak. Even though it might be hard to predict what the program will actually do in any specific case, it's still sort of a slightly different way of dealing with things. Well, just like, you know, it's always amazing to me that, you know, Anglo American law, which is very much based on precedents and, you know, what was what happened in previous cases versus continental law, that's much more based on kind of a code that's defined once and for all. The fact that, you know, lawyers and legal systems can work with both of those foundational, with those very, what seem like foundationally, very different approaches and work in a somewhat similar way is kind of a reflection of a certain kind of universality in the operation of those kinds of systems. And I think the same is true for sort of the computational law without ambiguity and the written legalese law with ambiguities. My guess is that in the end they sort of can achieve the same purposes, but they may have different economic characteristics. Like, it may be much easier to deal with much larger volumes. At this moment, when you're pressing this button, it's like 10 contracts could sort of fire that are computational contracts. But if each one involved sort of the operation of humans doing things, nothing like that will be possible. Whale comments With LLM lawyers, the patent disputes will end up being a bunch of robots arguing all the time. You know, what's really ironic about this whole thing is back in the 1660s, I guess Leibniz was writing his PhD thesis, and his PhD thesis was about resolving legal disputes through logic. And he spent in his life he had this kind of idea of a characteristic of universalis, a kind of universal language that that would represent kind of what meaning and so on. And his big use case was going to be Resolving things like legal disputes. So you know that that characteristic of Universalis idea you can view as being having a pretty direct line to my kind of computational language directions and so on. Although the technology has changed a lot and our ways of thinking about things have changed a lot. But it is kind of ironic that Leibniz already had this idea of, you know, forget the lawyers, just have logic, resolve these legal cases, and now we know more and we can say sort of let's just have LLMs arguing these cases. I think that that's going to be. Well, it's going to be a. There'll be, you know, it's one of these things where there'll be interesting adaptations that happen as a result of that. That coming in. I think that the, you know, as of right now, well, certain kinds of things, like given a set of points to make that have to be written up as kind of an essay, as a report, as a filing or something. Yes, an LLM is a good way to go from those points you want to make to the fluffing out of the thing with actual language. It's probably a bad way right now to be kind of arguing your legal case. I know there was some disaster with that of an LLM that just made up a bunch of citations, which is what LLMs do. They make up things that are statistically plausible. And if you're trying to say this is the specific legal precedent. Well, one that sort of has the right structure of pager numbers doesn't really cut it. Let's see. Baraki asks, is diversifying my professional ventures a worse outcome than focusing on one or two occupations, which I'm really good at? You know, again, it's, it's really dependent on people and the extent to which they can multitask. For example, I mean, there's some people where it's like they can only do one thing at a time. And you know, I must say that I quite often run into people who say I really want to do, let's say, science as a hobby. But, oh, I have to have this job. I'm doing software engineering or something like this. And that means I can't do my science as a hobby. Well, why not? Well, well, you know, my software engineering job is really only a 9 to 5 job, so there's a lot more hours in the day. And for somebody like me, I, you know, I tend to work a lot, like all time. I can and I do a variety of different things. And I think that it's really just a question of kind of One's force of will, so to speak, to decide one is going to do those other things that were sort of the hobby that wasn't what was part of the main flow of what one was doing. One's just going to decide to do those things and then one can. And I think in terms of this question of whether I think this whole issue about do you concentrate on one thing you're good at? Do you spread. I tend to think that doing things one isn't good at isn't a very good idea. There might be things where, gosh, you know, it's a, It's a more prestigious, it's a more, a more lucrative, it's a more whatever kind of thing, but you're not very good at it and you don't really like it. And I think that the, the, you know, it sometimes takes some sort of figuring out how to manage the world, so to speak. But the things you're good at and really like you're going to do better. And you know, maybe they're not the most prestigious or most lucrative, but maybe you can turn them into something more prestigious and lucrative because you're really good at doing them. And it's a better deal than going into something which is already prestigious, but you're really pretty bad at it. And it doesn't really. And you're not, you know, it, it doesn't. It's. It takes more personal effort to do the thing that isn't already prestigious or whatever and, and you know, make something good out of it than it does to take something where everybody knows. Oh, if you work at Place X or something, you must be cool type thing or you do particular occupation, you must be cool. It's a different personal dynamic in terms of what you put effort into. I think in, let's see. I think, Yeah, I mean, without further information, my $0.02 worth would be if they're things you're good at that you like doing, if you're good at them but you hate them, that's a different story. Those are the things to kind of double down on and see how you can kind of expand from those rather than just sort of casting out in some other completely different direction. I mean, it's very dependent on specific circumstances. I mean, some things it's like, yes, it's worth trying something because, gosh, you may turn out to be good at it and you didn't know you were good at it. Sometimes when you first try it, you don't know enough and you're not very good at it. And then you gradually get better. It's a slow process to get really good at doing something. Sometimes people come in with a sort of natural ability to do something. Some fields it's easier to kind of have the natural ability be the thing that sort of carries the day at the beginning. Other fields, you really have to have this kind of slow ramp of getting better at doing it. And there isn't likely to be a, you know, I came in at the beginning and I was immediately good at doing it. All right, let's see maybe one or two more things. And then I should. Des Comments. They're somewhere between 500,000 2 million cuneiform tablets, apparently just sitting in warehouses, untranslated, unscanned, inaccessible. What can we do about that? Well, I didn't know that fact. Cuneiform was the little language with wedges and things used in Babylonian times. Actually, funnily enough, one of my kids is very interested in these kinds of things and is actually right now trying to see whether you can get an LLM and other machine learning kinds of tools to be able to do things like read cuneiform tablets. So there really are a couple million cuneiform tablets sitting around somewhere. There might even be a piece of technology that will help solve that problem. And maybe we'll discover that the Babylonians actually had figured out all kinds of things we had no idea had been figured out 4,000 years ago. And we'll be a bit embarrassed that we didn't know these things for the last 4,000 years. That will be very interesting. I mean, I think it's always interesting that it's a surprising thing that gradually more and more of history gets kind of peeled back as archaeology and sort of scientific archaeology gets better, and as people decode that secret code that somebody wrote 400 years ago or something, or somebody finds those documents that have been sort of lost somewhere or another. It's sort of interesting the way that more history actually gets learned. You might think that, you know, whatever we know about the past, we're never going to learn more about the past than we know today. Let's see, Let's see. A couple of questions here, which he says might be simple from fellow. When you first started making sales with Mathematica, was it mostly to academics or to companies? And how did you find these first customers? Well, times were different. The web didn't exist. It was 1988. And how did people find out about things? Well, they read newspaper articles and we had quite a few of those. And they went to bookstores and we had the Mathematica book which I had written, which was in sort of all the relevant bookstores. So the main way people found out about us was I guess, reading the newspapers, word of mouth and to know more details. There wasn't a website you could go to, but you could go to the local bookstore and pick up the Mathematica book and either flip through it or buy a copy and read it. And that was a very successful thing. And I would say that one of the things that happened, gosh, a lot of very well known scientists, technologists and so on were among our first customers. And what I sort of realized only long after the first fact is that what the tools we built, and that's true with computational language today as well. If you're a person who kind of has an idea and has conceptualized what you want to do and now you just want to do it, the things we've built, the kind of superpower automation that we've built is exactly what you want. That's kind of, you know, you've got a pent up sort of idea and it's like, let me actually implement this. Okay, we got a way to do it, Great, let's do it. And so, you know, I would say that the, and this has been true throughout the history of mathematical language and so on, that the, the, the, I would say the strongest usage is in kind of intellectual leaders and people who are doing things where they conceptualize what should be done and they want to just get it done. The kind of dynamics of software engineering and things like this where people are more kind of saying, I'm going to write a program, I'm going to work nine to five, I'm going to write this program, I'm going to grind out X number of lines of code each day and so on. Those are not so much the people who are the people who've absorbed our technology because for them it's kind of like, well then you have to have a sort of a big idea about what you're going to do next. That's not what they do. They grind out code. Maybe they're realizing now that that's not really a surviving profession, but that's a different issue. But I think the thing where people recognized and recognized today the value of the tools we're building is where you've got an idea, you conceptualize it computationally now, just make it happen. And it was really, really striking. I mean, if you look at the early registration cards, and they were physical cards from Mathematica, a lot of very well known people, both either people who are well known already at that time, not necessarily young by the way, and people who became very well known and very successful and very innovative and did a lot of invented, discovered lots of things. They were among our earliest users. And the same is true anytime new technology comes out. There are companies and so on where there are people sort of paying attention, whose job is basically go find the new stuff and see what's happening. And obviously there was some of that going on as well. But it was a, it was a mixture of, I would just say a very interesting crowd of people. And you know, it's often the case when sort of new things happen that there's a, that kind of. It selects for kind of a certain set of people who are interested in taking initiative, so to speak. The things where, oh, it's been grinding for a long time, that can be done and is done by people who don't exhibit lots of initiative. Whereas the things that are kind of on the leading edge, that sort of selects for initiative. And that's what we had. Art asks, let's say an amateur claims to find a big breakthrough. How do you judge if it's worth reading? Well, there's an awful lot of meta information in the way that things are presented. So for example, there's in science technology type area, there's just formal technical stuff. It's like, okay, you got a big breakthrough and you can, there's a piece of code, there's a formula, there's a picture that is non trivial and shows something interesting. These are kind of technical thresholds of you have to have a certain degree of togetherness to get those things, to make those things happen. If you're just saying I've got an idea I can describe in words, any old LLM can do that. And it's much less obvious whether there's something real there by the time there's sort of formalism. And it's like, here's this thing, here's this piece of orphan language code. It runs, it produces this interesting picture. We can describe what's going on. Here's this kind of formal statement of things. Here's maybe the positioning of this with respect to other things that already exist. That's not absolutely essential, but it's useful. And it's also useful as a way to kind of prove that you didn't just come up with this and you haven't done your homework, so to speak. Because like I've got an idea and it's like, well, yeah, that's Great. But people had that idea 250 years ago. If you haven't, you know, if you give evidence of having done your homework, that really helps. And I think the, it's, it's always, I think other matter information is also useful. I mean when people. I like to think that I'm very non judgmental when it comes to assessing what people have to say. But it is useful often to know the backgrounds of people. Like I was doing something just, just recently where for various reasons I was sort of exposed to a, a remarkable new claim about something where it's kind of like, well, you know, what kind of a claim is this? Is this a claim that's based on a big tower of mathematical knowledge? Is this a claim that's based on, you know, hacking around with some physical devices? What really is this? And you know, if it turns out the person making the claim is this person who's been doing math for the last 30 years, then it kind of gives you an orientation to what kind of a claim this is going to be. Even if so, you know, that kind of thing is helpful. I would say that the presentation of things, how much it is sort of formalized, how much it's positioned, contextualized and a bit the background of the person. Now the background of the person might be, well, they dropped out of high school and they've just been having ideas and they're really brilliant and these ideas are really good, that's just fine. But it gives you some contextualization because then, you know, if you say, well, you know, how does this relate to formal thing X which they've never learned, it's not going to be useful. They may have discovered something that goes around formal thing X and they never needed to learn that. I mean, it's often a multiple edged sword, this question about how much should you know. Sometimes if you know too much, it kind of drags you down and you can't ever figure out anything new because you're just convinced. Or sometimes it's really a question. If you know a certain amount and you're convinced that everything is known somewhere, then you're not going to go and try and discover things. But if you know too little, then you can end up just coming up with a bunch of stuff where people can immediately say look, that just can't be right. Because XYZ and I've certainly seen a vast number of kind of things where people, particularly in science, where people say I've got a theory of physics, whatever else, and the theory uses the language of and the methods of essentially high school level physics. And it's like, well, you know, there's the 20th century, which is pretty much physics that is not taught yet in high schools. And that brought us much of the formalism of physics as it has existed, I would say, until our recent efforts, where there's kind of a new sort of step up in ways to think about physics. But there's kind of all of 20th century physics. And if you don't know any of 20th century physics, it's a vast handicap in coming up with a theory of physics today. And so that tends to be something where again, it's partly meta information of. Is this stated in the terms of high school type physics? Is it stated in using the formalism that is known from 20th century physics, for example? How does it relate to that? Is it something where you can. One of the great things about computational language and computational things in general is that they're really. It's very easy to get them specified right down to the bedrock. There's this piece of code, you run it, it does stuff it's not, oh, we can wave our hands and this or that thing roughly seems like this or that. It's run the code. Here's what it does. Sometimes there can be issues of interpretation of what the code did, but that's, you know, that's a separate challenge. So there's a few thoughts. I think that the kind of, the sort of the. I have an idea and I'm going to write it in words in two pages. And you know, if it's a small idea about something, well, that's okay. But if it's like, it's a big idea, I've solved the fundamental theories of everything, so to speak, and it's two pages and it's just words. It's just from sort of the texture of the thing. It's like, look, this is just not convincing. And it's interesting to see of the different sort of surprising discoveries that have been made over the years, they almost always have a lot of technical detail in them. The big idea, as it's quoted in subsequent years, is often something that can be said without a lot of technical detail. But when you look at when that idea was first presented, it was buttressed by lots of technical detail. And that's a good kind of way to identify that. All right, well, I think we should probably wrap up for now here. And. Well, I was fearful that the end of my cold would cause me to be much, even, much less coherent than I've been so I'm thank you for asking questions that have livened me up a bit here. And thanks for. Thanks for joining me. I think I may not be able to do these live streams for a week or two because I'm traveling and doing all kinds of things. But we'll see. Maybe I'll do them from an unusual location. We'll see. But thanks for joining me and bye for now.