Dec 21, 2024 ยท 2024 #46. Read the transcript grouped by speaker, inspect word-level timecodes, and optionally turn subtitles on for direct video playback
Edit labels for this show, save them in this browser, or download a JSON override for the production folder.
Transcript Playback
That Was The Year
Human Transcript
Timed transcript
Blocks are grouped by speaker for readability. Expand a block to inspect word-level timing.
Speaker 2
Hello, everybody. 2024 is winding down. What kind of year has it been? I do a weekly show, as regular viewers and listeners know, with my old friend Keith Teer. That was the week where we cover all the news in technology. And Keith has a wrap up. for 2024. He's made Claude, of course, artificial intelligence, person of the year. I'm not sure whether that's a fair description of Claude, maybe non-person of the year. Keith, we're not going to get to 2025 yet. We're going to do another show looking forward to 2025. But you make...
clawed your person of the year, and you make AI dominant. In your editorial this week, you say, if I start by telling you that I use AI every day, often for hours, AI is changing your life, isn't it?
Well, look, I wake up every morning with a smile on my face and a project in mind. And I usually get it done the same day. And that would have been unheard of. I would have needed months and engineers before. And I chose Clawed because despite my regular Clawedits for OpenAI, the fact is when I use AI, I am almost always choosing to use Clawed, which is the anthropic chatbot. rather than OpenAI because it's better at coding. And a lot of what I do is coding. That grid of pictures you just saw, that's 42 pictures, one from each front page from this year's, that was the week. Literally took me five minutes to ask Claude to write some Python that would go through my directory structure where I store this stuff and pull it all in and make a grid out of it. And if I did that manually, one can only imagine how much time that would have taken.
I wonder whether we might think, Keith, of 2024 as a very transitional year in terms of AI, because my experience is entirely the reverse. I'm not a Luddite, but I'm certainly not as technologically sophisticated as you. I don't have any need. Maybe I should have need for programming. I'm open to using AI, but I have to admit that whilst I've flirted with Claude and OpenAI and all the other platforms, it hasn't convinced me. So might 2024, when we look back, be remembered as the year where, and you've always been an early adopter, where early adopters embraced AI, but it hasn't quite caught on with everybody else?
Yeah, well, I think most people don't know how to use it. Let me rephrase that. It's not their fault. They don't know what to use it for. And I think to know what to use it for, you have to come to the table with a preconceived idea of an end goal that you can't achieve easily. And the AI could help you with. Just to give you another example, yesterday I pulled the top 150 investors in private equity, just names of companies. And I said to Claude, I want the email address, the name and the title for everyone in these companies, for one person who is responsible for decision making about investing in private equity. And I was focusing on SignalRank, obviously. It was about a day's work. I only finished a day because finding email addresses and titles is non-trivial. It used Perplexity's search engine, as well as Google's, as well as something called Apollo, which is a LinkedIn app. But it eventually found all 150 names, titles, and email addresses, which now gives me the platform on which I can do my next work. Now that isn't that technical. I just had to be able to know what to ask for. It was technical. It wrote code to do it. Crawlers and spreadsheet builders and confidence level on whether an email address is likely to be true or not. It wrote a bunch of stuff. I didn't have to know any of that. All I needed to know was, here's the companies and I want an email address.
So basically what you're doing is using AI to do sales. Well, to prepare for sales. To get the emails of people that you think might be interested in buying your product.
Are you going to boast to them when you're pitching them? Are you going to boast, well, my person of the year, Claude, he was the one or she was the one who it seems to be a he. He found your email for me.
Of course I'm not. It's interesting also for people just watching, well, everyone's listening, I hope, but watching as well, that the image of Claude is a grey-haired man with a beard. Was that Claude's idea of himself? Is that how Claude thinks of himself?
No, I... I did something blasphemous. I used Grok to imagine what Claude would look like. And Grok is Elon Musk's AI built into Twitter or X. So that is a Grok rendition of Claude.
And some people might be listening and watching, especially women might think to themselves, this is just another version of male domination, bearded men who seem to know everything. How would you respond?
Well, men do dominate. It's not a lie. It's true. It's not acceptable, but it is true. So you would imagine AIs would reflect reality to some extent. In fact, they got criticized earlier for not doing, like making the founding fathers various shades of brown skin to black. So reality is women are definitely challenged to be equal to men. And that's been a lifelong struggle for most women and for most men who consider themselves feminists.
So what kind of year has 2024 been, Keith, for AI? I mean, 2023 seems to be the year that open AI revolutionized the artificial intelligence world, convinced most of us that LLMs were for real. Is that a fair analysis? Is 2024 the year that 2023 became a little bit more substantial? It wasn't a revolutionary year, was it?
No, I think it was incremental. You went from, at the end of last year, OpenAI was already known about and existed. There were few competitors then. There's a lot more now.
It was the case that, you know, the debate with Gary Marcus and the rest of us was finally balanced. I think Gary now sounds a little bit out of touch because the success...
Well, AI has vastly exceeded his predictions of how bad it would be. The hallucinations, which was the dominant orc last year, really no one ever mentions it anymore. and it's because it's good. This morning, by the way, Andrew, we've both been busy, so there's no reason you would know this, but Sam Altman announced the O3 version of OpenAI, and they've only just released O1. They're skipping O2.
I think it's going to be a huge deal because it's reduced in accuracy down to close to zero, whilst making it possible for it to carry out a much wider range of tasks. The latest is the ability to act as if it's you on your computer, which Google also does now.
Well, you can give it all of the resources of your computer, Excel, Word, the cloud, web search, you name it, programming tools, and tell it you want to achieve something, and it will use the tools it has access to to achieve whatever that end result is. That's another way of describing what an agent is. An agent is it replaces you in terms of having agency.
So am I just dumb, Keith? Am I missing something? I'm trying to write a book, as you know, about America. I'm traveling around. I'm putting together information. I'm taking photos. I'm doing many interviews. I still have no idea of how to use AI, for example, to write a better book. How would I even begin to do that?
I don't think you would. I mean, it isn't really designed to be good at that. It's not going to replace... your ability to pass knowledge through an intellectual framing and a point of view. So it could never write a book, and that indeed is not what it's designed for.
Well, it will. It probably won't be a very good book.
Words and timings
Well,itwill.Itprobablywon'tbeaverygoodbook.
Speaker 3
It won't be a good book. It won't be an original book. Yeah. The closest I could probably get to it, and you've got to be a bit of a science fiction fan to understand this, but The computers on Star Wars or Star Trek basically serves human need. And human need is very specific. It's not like there's a general need. There's millions of specific needs. And any given need, it's probably a very good tool to help you address. So let's say in writing your book, you wanted to contrast the the views of person A with person B on topic C, it would be a pretty good research tool. More and more good. Less hallucinations, if any, anymore. If you wanted to find specific quotes, it might have a harder job because it's not an index and Google might be better for that. So it's really what do you want to use it for? And that is so strongly aligned to what your goals are. I do think we're getting to the point where you have to add the word robotics to AI, whether it's physical robots or software-based robots that carry out things that you don't need to pay attention to until they're finished. I think there's task-orientated stuff like that.
We're talking about 2024, not 2034. No, 24, even now.
Words and timings
We'retalkingabout2024,not2034.No,24,evennow.
Speaker 3
No, we are. That email thing that I mentioned, I didn't have to sit at my computer and wait. It just kept cycling through code, and eventually it finished.
I'm still not convinced. I mean, for most people who don't even understand why they would do programming, you're a programmer, you have your own startup. I can see people wanting a personal assistant. I would like a personal assistant. I have no idea how to use it. There don't seem to be any front-end products to do that yet. What is the best product in 2024 to show how AI can actually change ordinary people's lives?
I think AI isn't a product. That's the misconception. AI is an intelligence capability that you have to engage with. It's not an interface. I mean, you've got to interface with it somehow.
Well, it's simple. It doesn't just appear. It has a very simple interface. You choose to use, as you say, Claude. You go on Claude and you use its technology.
Yeah, Claude has a canvas, that is to say a place it saves things for projects, and it has iterative work through a project, and you can go back and look at it. So it's a chat-based interface with a canvas. Cursor, the one I use, is different. It has access to my computer. It can write code and run it.
Right, but you're not answering my question. I take your point. I'm not questioning your use of cursor or the way in which AI has changed your life, but how could it begin to change most of our lives?
You have to not be lazy, you have to use it. Let's say you want to cook something tonight and you've got various things in the fridge. It'll probably come up with a fantastic recipe way better than you would yourself. It's just, it's task-based. So you've got to have a task, any task. It'll be better than you at almost any task.
And it says, for a pleasant two-mile walk, it's giving me the neighborhood loop or the Pearson Astrodero Preserve or the Balins Nature Preserve. And it's giving me directions, how long it is, how hard it is. It's actually come up with five different ones that I could. It's even told me what the weather might be like and what the trail conditions are. And this is on Claude. Actually, I just use ChatGPT, but if I do the same on Claude, it'll be just as good. Claude's better at programming.
One of the things I'm struck, Keith, with you is that you have complete trust and faith in this technology. Earlier today, you and I, we strongly disagree on Elon Musk. And I mentioned to you that I thought that his backing of Germany's far-right AFD was... Just one more reason why he's a loathsome human being. You disagreed and you responded by running through, I don't know if it was CLAW or OpenAI, something on the German AFD in comparison with the German Social Democratic Party. I don't want to get into a debate about Musk on this. But are you more and more, you seem more and more reliant on... AI to settle disagreements, to make sense of the world. It is your ontology, isn't it? It's becoming your ontology.
That's a big word, Keith. How many letters is ontology? I'm going to guess seven or eight. No, but in all seriousness, ontology is a big word. It's how we perceive the world.
Well, ontology is really organizing principles for knowledge, labeling in a way. and the relationship between labels. What this is is the Library of Alexandria having been read and digested and you being able to ask the orb anything about anything in it and have a reasonable expectation of good outputs. I mean, it's so much beyond ontology.
Yeah, and is it... I mean, how different is that from Google? Is that a profound shift? I mean, we used to rely on Google as a search engine, maybe not so much for plotting out walks, for asking questions on books or... facts. So AI doesn't seem in the way you're suggesting to be that different, or is it?
No, it's very different. Google is just a very large sheet of paper with everything on it. There's no knowledge or intelligence. It has keyword matching and link counting to try to get the best sources for anything from that big piece of paper. But, um, AI, it's read everything in the paper and can not retrieve it because that's a search concept, but use it to answer questions or complete tasks that you ask it to. So it really is a doer. AI is a doer and Google is a store.
How has it changed? I take your point on the programming and your life as a startup guy. But how has it changed your day-to-day life, your relations with your wife, with your kids, with your friends? Or has it not affected those?
Well, my wife used to think I was an idiot. Now she thinks I'm a genius because I produce all this output. And she's kind of astounded by what you can do with it. My kids engage with it quite a lot, actually. I mean, this week, I'll just give you a couple of anecdotes from this week. My youngest son, Luke, he has a computer science class using a programming language called Scheme that I'd never heard of. And he had his final yesterday, and he was trying to understand some stuff about Scheme that the book wasn't really helping him with. So we started asking, in this case, Claude, which immediately said, yeah, I know Scheme. The best book is this book. What do you want to know? And we gave it some of his questions and asked it not to answer them, but to explain how it would think about answering them. Within about two minutes, he was leaning into my screen. And he, by the way, never pays any attention to either me or my screen. He was leaning in and saying, wow, this is actually better than my teacher. And it is. I mean, why wouldn't it be? Because it's got way more resources to draw on than the teacher has, and more time has been spent reading and putting everything into its ontology, as you say.
So it's just very, very good. Is that the darker side of things? I mean, so many books have been and will be written on the impact of AI on jobs. If, as your son Luke says, it's better than my teacher, the teacher who gets paid presumably to dispense wisdom will become redundant. Is there going to be... Does any of this in your complete embrace in terms of your lifestyle and how you... You interact with the world, your ontology. Does it make you in any way worried or are you gung-ho on the potential?
It doesn't worry me at all. I think it's fantastic. I mean, just what an amazing thing. What are the teachers, Keith, going to do? Children still need teachers. They just need them for different things. The most mundane things a teacher does is mark work. and score it. The second most mundane thing is to repeat the content in a classroom so that the children are aware of it. Those two things can be automated. The thing that you can't automate is the teacher's caring and compassion and mentoring of a kid through his journey or her journey. And so a teacher can become way more human by having to do many, many less automatable tasks.
You know, there is a body of work that came out this week. Well, it's been around a long time, but it got publicity this week, which is work that is about trying to understand how the brain works. And this work posits that the brain is a quantum computer, not a binary computer. And by being a quantum computer, it can have consciousness. And this particular professor, his work is all about understanding consciousness. And consciousness is very different to what AI does.
And so the thinking there is that humans are really not reproducible in AI at all because you can't reproduce consciousness or awareness or even context mostly. And that's where human empathy comes from. And so now that said, an AI can become a pretty good fake therapist because the content of therapy is something it has and can use. But it's understanding of when to use what part of therapy and who with is almost non-existent.
Your post of the year touches on this. It's from Alex Kantrowitz, who I think is an excellent writer. And it's a post that he put up on his, I think it's his Substack page, on a Neuralink patient called Noland Arbor. Tell us about this and why you chose it as your post of the year.
Well, I think because I'm so astounded by it. When Elon Musk started talking about Neuralink and putting chips into the brains, I think it was pigs initially, I just thought it was not real. I couldn't see any way that you could take anything that happens in the brain, plug it into electronics, and have control, fine-grained control coming out of it or understanding. And this guy is the first human patient, and he's quadriplegic, if I'm not mistaken. And this article is saying... This is Noland or RBAC. Yeah. And Noland is... Alex Kanjovitz, by the way, published this story, but he also published a long interview with him, which is itself a challenging thing to do. He says, I often forget that I am disabled. And for a single brain transplant to be able to produce that in somebody that is highly physically challenged, it just blows me away that that's even possible. So I think it has to be the invention of the year. It goes way beyond SpaceX or Tesla.
You chose Claude as your Person of the Year, which is a bit of a cheat because Claude isn't a person. He's, I guess, an aggregation of human intelligence. I cheated a little bit too. My Person of the Year is the US. There's an interesting piece earlier this month from Martin Wolf of the FT about the exceptionalism of the US economy and its success. is uh is am i onto something keith was 2024 the year where the american model for better or worse became unarguably the leader in the west we've done a number of shows on the decline of europe the failure of europe to come up with models to compete with the us to keep up with the us what kind of year has 2024 been for the u.s economy particularly in terms of the tech sector
It's been a good year, but I think my framing of it probably, I think you've got to separate a few things out. I mean, the US is far ahead of the rest of the world when it comes to innovation leading to business outcomes. So all of the top companies are tech companies, and they're all US companies bar maybe one or two. So the U.S. certainly is good at financing and allowing creativity and innovation to result in business success. But I think there's another U.S. which is summed up a little bit in Trump's victory, which is the anti-state capitalist economy movement And if you look at it in that way and look at what's happening in Argentina, for example, in contrast to Brazil, there's a US kind of way of thinking that goes beyond the US. I think this German ADF party is a bit like that as well, which it kind of blends libertarianism with fierce independence and and somewhat antagonistic to the rest of the world.
Yeah, I'm intrigued that you're not deeply hostile. Given your internationalism, your commitment to globalism, why would you even flirt with Musk or the AFD or Millet's economic nationalism in Argentina? I'm surprised that you're not deeply offended by these developments.
Well, I think it's all to do with the relationship between economic success and human good. And they're often talked about as if they're opposites. And I actually think that it's very hard to imagine human good that doesn't live alongside economic success. Because economic success creates the means of life and hopefully makes them cheaper, theoretically tending towards free through automation. And so I really think there's a new framing, a new ideological framing to do with progress and innovation and civilization as well that doesn't really fit into the old paradigms of socialism and capitalism.
But if you're right, and a lot of people would debate whether or not you're right... Isn't it troubling that these ideas go with hostility to immigrants, what some people at least see as racism, certainly hostility to people of different color skins? Isn't that troubling?
Well, it's troubling, but I think you've got to understand the cause. I think a lot of modern... sentiment against immigrants comes from mistakes made by the left in benefiting immigrants above the local population and the resentment that arose from that. So I think it's the abandonment of poor people by the left and the left's embracing through identity politics of the underdog that meant there was a blind spot on the left about the impact of that on their normal base, ordinary people. And I think most of this nationalism is the price the left is paying for having had that blind spot. I don't think the human beings who have those thoughts are bad people. I think they're self-interested people who feel let down. And in order to get them back, you have to embrace their interests. which I personally don't correlate to nation state interest at all. You're right, I'm a globalist, not a nationalist. But I do understand why people have arrived, especially poor people, at those thoughts. And I think populism is the consequence. Populism is the price we're paying collectively for having had that blind spot.
I wonder whether 2024, Keith, will be remembered as the year where populism became, if not mainstream, more and more acceptable. One of the interesting pieces in the Martin Wolf FT piece was the way in which the US is increasingly becoming a plutocracy and, of course, Musk is exhibit A in that. But are we seeing in 2024 a new kind of system emerging? Maybe a plutocratic one run by what some people might view as enlightened libertarians like Musk, others see as anything but libertarian. It's something structurally changed in the Silicon Valley. And we talked about this in a number of shows this year. The Silicon Valley stormed the gates of Washington, D.C. in the form of men like Musk and Sachs.
Yes and no. I think mostly no, because I think a year from now, we will find that Musk and Sachs met opponents that were far more strongly in place than they realize. I think they're going into DC thinking they can win, and they may well... have their ass handed to them by the bureaucracy and the state, because it's very powerful, and they aren't. So I wouldn't predict their success. If anything, I think there is an undemocratic reality, which is that despite the fact that Trump was elected with a majority, he almost certainly is going to fail to be able to carry out his policies due to what I just described. And so capitalism, looked at from the point of view of its governance,
Yeah, and that, I think, contrasts with your arguments about AI. I mean, you may be right, AI might be changing your life, your world, but it doesn't seem to be changing most of our lives and certainly not political life. It seems to me as if... maybe I'm being a little abstract here, in that 2024 is most interesting because it's setting us up for 2025. In other words, Trump seems to operate in reality television terms, where it's always the next show that's the big deal. He always leaves you with a cliffhanger. And the most interesting thing it seems to me about 2024 is it set us up for 2025.
I'm not, as you know, a great fan of Trump or I certainly didn't celebrate his election. But I do think that 2025 is going to be really interesting in terms of we're going to do another show on 2025. But I wonder whether 2024 historically will be remembered as the year before 2025.
You know, I think it's hard to predict. The most striking event to me was what happened in Romania in the last couple of weeks, where a populist came from nowhere to be the majority winner of the presidential election. And the courts just annulled the election, arguing that Russian interference created this person's success, which of course would be super hard to prove, as they admit. But they've annulled an election because they didn't like the result. And that was, in a sense, what would normally be called the left, taking democracy away from what would normally be called the right. I don't think those labels work anymore. I think it was a populism against the incumbents more than anything else. But I do think we're in an era where the general population is so reasonably
Yeah, and I wonder whether ultimately that's the big deal about AI. You talked about how it can make you into a better program or do a lot of the programming that used to take you days can be done in a few minutes or a few seconds. But whether ultimately... the revolution that we're living through, this AI revolution that began in about 2023 and will certainly run through the 20s, maybe even to the 30s, will ultimately create a new operating system where we'll somehow trust better.
Well, you trust it enough that you've made Claude look very wise with a gray hair and a beard. The old Christian version of God. In other words, is AI just another version of Christian theology?
Well, if you believe that God was created by man, yes. Yes. because men create things bigger than themselves as a goal. And by men, I mean humans, not the gender men. So I do think that AI is a human product that shows human capability and there's nothing about it that isn't human. And so in that regard, As an atheist, I think God is the creation of humans as well. I'm not one of those atheists who scorn people who created God or believe in God. I think it's a fantastic thing that they believe in something bigger than themselves and want to because that is the aspiration of humanity generally. So I'm good with it.
Well, finally, I know you made Cursor. You've talked about the AI code editor Cursor, your startup of the year. But what kind of year has it been for you that was the week? You've done tremendous work, Keith. I ride off your coattails. You aggregate all the news of the week every week, and we talk about it every week. Do you feel that you're wiser at the end of 2024 than you were at the beginning? Leaving aside AI, I know you feel AI makes you wiser, but do you feel you know more at the end of 2024 than you did at the beginning?
I think my assumption is always that I don't know anything, really. and that everything is a temporary point of view until proven wrong. That's my basic belief. So that was the week forces me to read and think. I don't think it makes me write.
Well, that was the year for 2024. Keith, you and I will continue into 2025. Our next show, we're going to Look at our crystal balls and imagine what might happen in 25. But have a wonderful year, Keith. It's been a good year. You and I have covered a lot of ground. We seem to be a little bit ahead of the curve. And I hope we'll continue to do that next year. Thank you so much.