ChatGPT Gets Lazy, Elon Musk Blasts Advertisers and Releases The Cybertruck, Jack Ma Returns

0:00Is ChatGPT getting lazier? We've got lingering concerns about the world's leading AI companies. Elon Musk tells advertisers to go F themselves and sets the Cybertruck free. Jack Ma is back and plenty more coming up right after this. Welcome to Big Technology Podcast Friday edition where we break down the news in our traditional cool-headed and nuanced format. Wow, what a week of news. It just continues here and we're going to break it all down. Joining us today is a great reporter, Luis Matsakis of Semaphore is here with us to talk, of course, about the week's news and some of her great reporting on the latest in the AI dramas and plenty more.

0:37Louise, welcome to the show. Hey, Alex. It's great to be here. Great to have you. You're about to come out with a story talking about how ChatGPT might be getting lazier. Now, this is something that we all maybe feel spiritually, but what did your reporting show you on that front? Yeah, so over the last few weeks, a number of mostly professionals, I would say. So these are people who already pay for chat GPT plus, which is the$20 a month subscription, which gets you access to GPT four, which is, you know, their most advanced model right now. And they're basically like, look, when I asked the chat bot to like produce 50 lines of code or, you know, just to do tedious work to maybe like, you know, put a list together of upcoming dates for me.

1:22it won't do it. Or in, I think the cases that are the funniest, it will instead give people instructions on how to do it themselves. So it'll be like, hey, like here's a template, you know, you can fill out the other 50 lines or whatever, which is, I think, super interesting. And it's not clear to me exactly why this is happening. So a few OpenAI employees have said that this is part of, you know, the result of a known bug. So there is something going on here that, you know, OpenAI admits is a problem. But I think the reason that this is interesting is that it speaks to this wider, really unique issue with large language models, which is that they're a black box.

2:02You put something in and something comes out on the other end. And even the people who make these models don't really know what goes on in the middle. And that can be really frustrating for users. And I think what you're seeing now is as people try and bring ChatGPT into their workflow, this is really frustrating. You know, this happened a few months ago where people said, oh, you know, ChatGPT is getting dumber. It can't do the things that I used to ask it to do. But at that time, I felt like a lot of people who were saying that were like kids trying to get their homework done or, you know, maybe people who are using this in an experimental way.

2:39But I think this is a bigger problem now for the company when you have these, you know, startup founders, tech executives who said, you know, I'm doing this workflow and I'm using this chatbot and it's not working anymore. And that's really frustrating. Can I just say how amazing it is that a known bug within ChatGPT is making it tell people, nah, I'm good when they ask it to do stuff. It even makes me more in awe of this technology that it has sort of incorporated such a decidedly human trait in the way that it works. Totally. I heard people joking like, oh, you know, it's like hiring a bunch of really capable interns, but they're lazy and avoid work in the exact same way as human interns would.

3:25So I totally agree. It's remarkable. And I think it also speaks to how, you know, the reality is that OpenAI is making tons of tweaks all the time. And they can't always predict what those tweaks are going to do to the models. And that's, you know, I think a problem for users. And we just also don't have a transparency system or transparency norms. You know, there's no release notes. There's no like, you know, hey, we made a big tweak today. Like you might see this, right? Like we have no established system for communicating with people about how these models are going to change over time or, you know, what the impact might be.

4:07And so I really want to see how that is going to develop over the next few years. Right. And it sort of brings us to an area which has become more discussed since the open AI situation, which is that we seem to be heading towards this commoditization of all these bots. It was discussed at DealBook with Andrew Ross-Ork in the conference this week about how they all seem to be coming together and starting to mirror each other. And you're not going to have one chatbot that's extremely capable and one that's not. It's just the functionality seems to be colliding. What do you think about that? Totally.

4:42I think that right now, especially enterprise customers, have a lot of different options for the types of models that they want to use. And a lot of the companies that I've spoken to say, we're not going to commit to a single model. It's too early. We want to see what works best for us, what's the most affordable for us, where we get the best customer experience. So I think that when there's, you know, rumors flying about the chatbot getting lazier, in that kind of environment, that's a bigger risk for an open AI, right? Because, you know, it used to be, oh my God, chatGPT is down. Like, what am I going to do?

5:16Am I going to have to write my own essay that's due tomorrow? Whereas now, I think it's a lot easier for people to say, okay, chatGPT is not working for me. I'm going to try, you know, Meta's Llama 2. I'm going to try something from Amazon. I'm going to try, you know, a different open source model that's free. and I'm going to run it myself. So I think that's really exciting, but it also means that differentiating and ensuring that you kind of just have the basics of like good customer service, a good interface, you know, reliability, those boring things are going to become more important than just the capabilities.

5:50Right. And maybe customizability as well, which is where these open source companies can get a heads up. But you really just kind of lead in perfectly to this story that I've written this week about Anthropic and their governance, because it does really seem that it is Anthropic's moment. They've raised$7 billion, including multiple billion plus dollar rounds since late September from Google and Amazon. They have a bot that works extremely well called Claude. They sell the underlying technology as a model. They've released an update in the middle of this open AI catastrophe. And they're so well positioned to be that number two or that alternative that people sub in if they don't like what's happening with open AI.

6:34And then you start to look at the governance. And that's where things get interesting. Because they're not a non-profit, they're actually a public benefit corporation. But the people that choose the majority of its board will eventually, and this means within a few years, be people that are made up, people that make up this long-term benefit trust that Anthropic has set up. And the trust has five people on it, initially picked by Anthropic, but eventually they'll select replacements. And that trust within four years is going to be able to pick three of five Anthropic board members. Now the board members are going to have this fiduciary duty, so it's not a non-profit like OpenAI, which you would imagine insulates them a little bit from the type of thing that happened at OpenAI.

7:23But it's just so fascinating that you have this company, which was set up by OpenAI expats, people who had left OpenAI and they wanted to build a company more focused on safety. So of course, they're not going to go with a traditional for-profit board structure, but they build this long-term benefit trust. Their mission is to make sure that the AI is developed safely. and the way that they live out their mission is by selecting people who would effectively channel it as members of a for-profit board and it just puts Anthropic in this very fascinating position where it can be this number two it can be this opening this this second or yeah this opening or this hedge for companies that are trying to build away from open AI but that being said the governance is fascinating and it's not as clear cut as a standard for-profit.

8:15Hearing that, I'm curious what you think about Anthropics position and when people, when they're thinking about this company, should Google and Amazon now kind of have their eyebrows raised given that the structure is not the traditional for-profit structure. It does have some notes of that open AI structure, although perhaps a bit more stable. I think it's a really good point to make And it's so interesting to me because when Sam Altman got fired, I looked back at how some of OpenAI's investors were talking about the structure and they loved it. You know, I think it was beneficial to them. You know, a few days before Altman's ouster, he, Brad Smith, the president of Microsoft, actually said that this nonprofit structure was what made OpenAI more trustworthy than a competitor like Meta, you know, that is profit driven.

9:05So I think there's been sort of this rude awakening to what this sort of funky structure will mean. But I think unlike maybe Microsoft, which is OpenAI's biggest partner, you've seen Google and Amazon, which are, you know, the big investors in Anthropic, hedge a little bit. You know, Amazon is allowing other models onto AWS. They're developing their own, you know, their own infrastructure, their own AI technology. you know and Google is also very much doing the same so I think that by having Anthropic you know on their team so to speak they're you know able to sort of see what's going on and to have some you know interest in one of you know what is being referred to as the foundational models you know and Claude is considered one of them but I think at the same time you're not seeing that sort of over reliance you know to the same extent at least like you're not seeing you know, the CEO of Anthropic on stage next to the CEO of Google or the CEO of Amazon.

10:09Like, I think that that branding is not as clear, I guess. And I also do wonder, I think that this firing of Sam Altman did give an opportunity for Claude. You know, there was some reporting that more customers were going over to Anthropic, even if, you know, maybe the same governance vulnerabilities are there. But I think the real issue is that OpenAI is a household name and that's going to be really hard for Anthropic to compete with. Like, you know, even, you know, companies that are in the tech space, sometimes I ask them like, hey, like, are you using Claude? Like, what other models are you using?

10:44And they look at me and they're like, what's that? You know, and I think that that speaks also to how Altman has just been more of a corporate, you know, actor, right? Like, He's extremely ambitious. He's sort of positioned the company in a very public way. And I think that a lot of the people who are at Anthropic are academics, sort of have this high-minded idea about what artificial intelligence should be and what it should do for the world. A lot of them are former academics or have PhDs in things like philosophy. And while I think Anthropic has been really shrewd about getting sort of a front row seat at the regulatory table.

11:24They were much smaller than Google or OpenAI, these other companies that were invited to these closed-door meetings at the White House or invited to testify before Congress. But I don't know if that's going to translate to profits. Yeah, and they're not just academics. So there are also plenty of effective altruists in that company. And we spoke about effective altruism at length on Wednesday, so we're not going to just continue to do it so deep here. But I would be remiss if I didn't point out the fact that these trustees on this long-term benefit trust, two of them that I found, at least two of them have direct ties to effective altruism.

12:04There's Zach Robinson, who's the interim CEO of Effective Ventures US, which is inherently tied with the effective altruism movement. And then Paul Cristiano, who's the founder of the Alignment Research Center, who's also a prolific writer on the EA forums. You know, you've done some reporting on this. You spoke with the CEO of Skype who said that they're backing away a little bit from these EA ties. Let's just briefly touch on it. Is this something that folks should be concerned about? And how do you view what's happening to the EA movement right now? So I think the number one thing that is important for, you know, investors, consumers, everyone to realize about the impact that effective altruism and sort of like, you know, related movements like rationality are having on the development of artificial intelligence is just that it's a very specific view.

12:59Right. Like when you talk to these people, you know, there's totally variation, you know, in this community, which is, you know, thousands of people on multiple continents. But what I find is that they read the same texts, they read the same blogs, they all know the same people, they go to the same effective altruism conferences. There's nothing wrong with that, you know, and I think there are plenty of good ideas that are circulating within these communities. but it's really important to remember that it's one perspective. It's one community. And I think that this technology is too important.

13:30It's too powerful for only one viewpoint to be considered and how it's developed. And I think that eventually something has got to give. And I think that the good news is that you're seeing sort of at the government level, I think that the White House, other lawmakers in other countries are seeing like, okay, we can't just listen to these people to decide how we're going to steer this, which is great. But I think that the risk for a company like Anthropic is that if they're not letting in these other perspectives, what are they missing, right? What are they not seeing? What are the problems that they're not accounting for?

14:05And that's what I worry about when you have a board that is, you know, primarily consisted of people who read the same things, talk to the same people, have the same perspectives. Yeah. And the philosophy. So by the way, for the, we did get a few people writing in after Wednesday's podcast and I appreciate it. I've read it already. I'm going to respond. Um, and we're definitely going to try to get somebody who's more aligned with effective altruist thinking on the show just to talk, talk it through and nuance. But, um, but yeah, it's, uh, it's, uh, clearly like baked deeply within, within to anthropic.

14:41And the thing is, and we talked about this on Wednesday, that there is a tendency to sort of act rashly if you're ascribing to this long-term thinking where, you know, the lives of people in an imagined future universe or future Earth are just as valuable as the lives of people today, even though you have no idea what the lives and the society will look like in the future. And so, yes, it's one strain of thought, but it also, we've seen it in places like FTX and potentially even with this OpenAI board, there's a tendency to, well, not a tendency, but there's definitely moments where the thinking can lead to what seems like rash action from the members.

15:21Does that sound right to you? Yeah, I think that that's definitely true. And what I think is positive that's coming from this moment is that people are realizing that this is a community that shares a lot of the same views and has, you know, its own perspective. Because what I think was happening before is that you saw people say, oh, well, like this company is saying this and like this company is saying it. And, you know, that think tank, that research organization, like, oh, wow, like they all agree. So like, therefore, there's clearly consensus about like what matters the most with AI or like what people should be worried about.

15:57Whereas I think now people are saying, oh, wait, like all of those institutions are like people who all know each other and they all are being funded by the same handful of billionaires. Right. Like that's really important to know. It's really important for lawmakers to know. It's important for consumers to know. It doesn't mean that you can't agree with them or that they don't have good ideas. But, you know, knowing that the three policy papers you read are actually all funded by the same billionaire is important context for evaluating whether you want to believe those claims. Right. Or like whether those claims are like, you know, the totality of the thinking on this issue.

16:31Yeah, I was going into some of these organizations on the long-term benefit trust of Anthropic and digging into their financials. And they have murky statements like 70 % of our funding is from one source. And it's like, come on, just name the source. Who's it really helping to keep that opaque? Right, right. And we should just say that in a lot of cases, it's Dustin Moskovitz, who is, you know, an early Facebook employee who is worth billions of dollars. And I think that he is sincerely funding these initiatives because, you know, he genuinely is worried about how AI is going to be developed.

17:07But I just think in other contexts, when we see a web of organizations that are being funded by the same person, we don't ignore that fact in evaluating what they're saying. Great. And yeah, I have asked Dustin to come on the show multiple times and no avail. But let's also turn to OpenAI because I'm curious to hear your thoughts about what's happening in the aftermath of the Altman coup and counter coup. You have a story this week saying that a number of open AI employees are looking for job opportunities elsewhere after this cataclysm. So, you know, 700 open AI employees signed this paper demanding Sam back.

17:52Sam is back, but some are now starting to explore working at other companies. What can you tell us about that? Yeah, I think first, one caveat worth noting is just that like these people make oodles of money and they are so in demand. And so, you know, it's not hard to find opportunities elsewhere if you want them. Right. But I think that what it demonstrates is just that this was really crazy. Right. Like this was, I think, an unprecedented incident in many ways in Silicon Valley. and I think the number one issue is that we still just do not really know why Altman was fired right the only thing we know is that he was you know not consistently candid in his communications I still have trouble saying that phrase which is all that the board really has revealed so far and I think that the board would say that they're still worried about that and that you know Altman coming back was not, you know, indication that their concerns were not valid.

18:52They believe that, you know, there's going to be an independent investigation. He lost his board seat, you know, his close ally, Greg Brockman, who was also back at the company, also lost his board seat. So I think that the fact that these people stayed on, you know, during this like incredibly terrible, you know, intense five-day saga, you know, in which headlines were swirling constantly in order to get these concessions, I think shows that these concerns were serious. And I think until we know more, it's difficult for really talented, really in-demand employees to necessarily bet, you know, the future of their career, particularly like we talked about earlier, when perhaps AI models are becoming more of a commodity and there's all these other great companies that you know that they're going to exist, right?

19:42And that the CEO is going to be there at the end of the day. And maybe that's a better peace of mind for you and your family and your future. Right. Did you see Amazon, speaking of everybody building a chatbot, they have their own chatbot for business now called Q. They announced it this week at their big AWS re-invent conference in Vegas. It comes a year after Microsoft backed open AI launched its ChatGPT chatbot. This is according to CNBC. And it's an interesting name. They can't quite determine where Amazon selected Q. It may be from the James Bond movies, maybe from Star Trek. But it's really an enterprise chatbot.

20:26It plugs into applications like Jira and Salesforce and Slack. And it seems very useful for business intelligence. And, you know, my instinct on this one was that this is a boring chat bot. And, you know, of course, Amazon comes out with like the least exciting bot. And it's a very enterprise focused bot. But then I'm reminded of this belief that I have that we're going to end up moving from these broad, large language models to more narrow, specific use case type bots. And this actually might be something that can be very successful and useful to people who are using it in a business context.

21:01Maybe not as big as ChatGPT, but definitely something with staying power. If it could, for instance, help you speak to all your organization's data and figure out where you need to get selling better or what's broken and create a ticket right there or potentially aggregate some of the conversations within your organization happening on Slack and get a pulse right away. What's your perspective on Q? Q? Well, first of all, maybe I just have like internet brain worms, but I couldn't believe that Amazon named this Q after we just endured, you know, years of like the QAnon conspiracy theory. But that aside, I think that you are spot on that.

21:43Look, the reality is that Amazon is not an organization of a bunch of, you know, high minded researchers who, you know, think that they're building, you know, super intelligence, right? Like this is a company that the majority of its profit is coming from its cloud computing service. It's talking to the biggest companies in the world every day about how they're storing their data, how they're accessing it, you know, what tools can make their business better. And I think that Q is, you know, responding to those concerns, right? Like they are trying to build something that will slot neatly into your existing AWS experience.

22:20And I totally agree that what people are going to need is something that's customized, something that's fast, something that integrates with the tools they already use. You know, it's kind of interesting that one of the first things that chat GPT did was, or sorry, that OpenAI did with chat GPT was those integrations, right? It was like, you're going to use chat GPT to like order from Instacart, or you're going to like use chat GBT to like buy things online. Right. And it's kind of interesting that for now that didn't really work. Right. Like no one, I don't know anybody that's using those integrations where I think you're definitely going to see enterprise customers using something like Q to integrate with the services that they're already using.

23:01Right. I think we're a way off from something sensitive, like using Q to like do financial transactions or to access, you know, maybe potentially sensitive customer data or, you know, things like health records. But there's no reason that you couldn't use something like this to say like, hey, like, what was our profit last quarter? You know, like, how much should we make from this? Or like, you know, how many customers signed up today? Or, you know, these sort of like, run of the mill business questions. And I think that, yeah, it's boring, but it's going to make more money, perhaps, at least in the short term than, trying to get super intelligence or make something that everybody uses both in their personal and professional lives.

23:45Q is here to save us, Louise. You know the truth. Not again. Listen to Q. I can't do this again, Alex. Q is finding these patterns that no one else is seeing. All you have to do is trust Q. Like, were these marketing people living under a rock? I just don't understand how that happened. Maybe they knew exactly what they were doing. True. Because it kind of fits. Louise Metzakis is here with us. She's a reporter at Semaphore. We're going to talk in the second half about Elon Musk telling advertisers to go F themselves. So stay tuned. We'll be back right after this. Did you know your credit card points and miles can lose value to inflation?

24:26Credit card companies often reduce the redemption value of your points and miles. Now, imagine a credit card with rewards that can grow in value. With the Gemini credit card, you can earn Bitcoin or one of over 50 other cryptos instantly with no annual fee. Every swipe at the store or gas pump earns you instant rewards deposited straight to your account. Plus, sign up now for a$200 Bitcoin bonus to kickstart your rewards. Visit Gemini.com slash card today. Check out the link in the description for more information on rates. Again, if you're looking to invest in Bitcoin but don't know where to start, the Gemini credit card makes it easy.

25:05The Gemini credit card is issued by WebBank. In order to qualify for the$200 crypto intro bonus, you must spend$3 ,000 in your first 90 days. Some exclusions apply to instant rewards in which rewards are deposited when the transaction posts. This content is not investment advice and trading crypto involves risk. The Gemini credit card cannot be used to make gambling-related purchases. What the hell is going on right now? And why is it happening like this? At Wired, we're obsessed with getting to the bottom of those questions on a daily basis. And maybe you are too. I'm Katie Drummond, the Global Editorial Director of Wired.

25:41And I'm hosting our new podcast series, The Big Interview. Each week, I'll sit down with some of the most interesting, provocative, and influential people who are shaping our right now. Big interview conversations are fun. I want a shark that... That eats the internet. That turns it all off. Unfiltered and unafraid. So in a lot of ways, I try to be an antidote to the unimaginable faucet of reactionary content that you see online, to the best of my ability. Every week, we're going to offer you the ultimate luxury of our times. Meaning and context. True or false, you, Brian Johnson, the man sitting across from me, one day, at some point, as of yet undefined in the future, you will die.

26:25False. Tell me more. Listen to The Big Interview right now in the same place you find Wired's Uncanny Valley podcast. Subscribe or follow wherever you get your podcasts. And we're back here on Big Technology Podcast with Luis Matsakis, a reporter at Semaphore. Let's talk about what might have been one of the more remarkable tech interviews, I don't know, of a decade. I haven't seen many like this. So Elon Musk is sitting up there at DealBook with Andrew Ross Torkin. You knew it was going to be good. And Sorkin immediately starts talking to Musk about this advertiser. Drawback, boycott, exodus, after they didn't like some of the tweets that he was writing about Jewish people.

27:10And Andrew Sorkin was very, I would say, tender in his line of questioning, talking not about how Elon Musk needs to be liked, but he did need to be trusted. and Musk replied in, I guess, typical Musk fashion. He goes, if someone's going to try to blackmail me with advertising, blackmail me with monies, go F yourself. Go F yourself. Is that clear? Basically telling these advertisers that he doesn't want to play games with them. Either they wanted to advertise on the platform or they didn't want to advertise there. It will probably sink Twitter's advertising business. Louise, what did you make of this moment?