'Having Their Say' in AI. Altman, AI Researchers & Big Tech. ARD post 178

0:00 Welcome back AI curious folk to AI Ramblings Daily Podcast episode 178. I'm Michael Parakh, your human AI guide. It's Tuesday. Today's theme is having their say in things that matter in AI. On to today's show, getting your turn to speak on something that matters, putting it on the record, saying it the way you want it once and for all. We all want to do that. AI is no different. In this AI tech wave, three groups had their say this this week and the past few days, each in their own way. OpenAI's Sam Altman with a lot of words to a lot of different outlets in several days.

0:41 And then a group of AI researchers, about over half a dozen of them, they've claimed all kinds of safety issues around the AI. They've had their say, and we're going to examine that. And then Meta and Microsoft, they're speaking with their budgets and having their say in the context of where they want to go to be the leading the frontier AI models. My takes on each and an overall take. Let's go and get started. First of all, OpenAI's Sam Altman. He had a lot of interviews with a lot of publications.

1:12 They'll all be in the show show notes, and so you'll be able to read them in full detail. Indeed, there's even a big two-part piece, a set of pieces in Vanity Fair, which I'll again have in the show notes. He went on a media blitz, and this is in addition to his postings on X. Remember in the last few days, he's been in DC with the president along with the other tech leaders. And basically what he's saying is several things. He's saying he's very uncomfortable with people describing religious force to AI models or surrendering human judgment to them.

1:51 And he's talking about safety issues, which is obviously the core thing that everyone is talking about. The religious element comes in because the New York Times had a piece that Anthropic co-founder Chrisola was back again meeting with religious leaders. You may recall I had some write-ups on his meetings with the with with the Vatican a few months ago. Apparently, that is continuing. The issues are ranging from AI models could be cautious and potentially how to give them morals.

2:22 And then Sam Altman wanted to have his say on some of this as well. And he had a particularly interesting interview with the Politico where he said there's a lot of daylight between him and Dario Amodei, the co-founder of Anthropic. His words, Sam's words, "The world should accept some bad things happening for the benefits of of the technology." Obviously, by itself, it's very sensational, but he has a broader point that, you know, most technologies have dual uses and every technology from the fire, the wheel has had downside, some bad things happening to society. So, in that context, if you look at the overall context, it's not a very dramatic statement.

3:02 But he said also that he doesn't think one lab in San Francisco should hold the most powerful AI. And no one's going to disagree with that. And then, you know, his point is, you know, there should be a lot of them, which again is not a super controversial point overall. And then he's go he will goes on to do this two-part interview with Vanity Fair where he talks about a lot of details on both how OpenAI is thinking about their opportunity set, their timing for IPO, which has been talked about. He talked about the capabilities of the most recent frontier models from their own shop, Astra 6.1. How it it was important that OpenAI delayed it a little bit because of security and accident risk.

3:47 And he goes into slightly detail that maybe worth a little bit more time to kind of understand where he's coming from. My take on all of this is he's taking this time to position himself and his company, obviously, versus Anthropic that is going public imminently between now and Thanksgiving. I've already talked and written about that. And that's why he's on a barrage on the media front to make sure he's positioning himself relative to the conversations he knows he's going to have with the same investors who are meeting with Sam and his team over the next few weeks. And so this is his way of paving the road on some of those issues and some of those differences.

4:27 But the core issue here is that both of them are both Anthropic and OpenAI are threading the needle. They're running a pragmatic course between making sure they run their company for revenue growth and eventual profit and building compute and getting their share of the of the capital dollars out there. And at the same time, make sure that they are careful as much as possible in the laissez-faire scenario scenario that we have in terms of the US market, the much more controlling market in Europe, and obviously the competition from open-source models in China. So the daylight between the two companies, Anthropic and OpenAI, is is real, but it's far narrower than the rhetoric might suggest. Both companies very much want the upside of AI while mitigating the downside. So important to keep that in mind always.

5:19 Next item I want to talk about is the AI research folks that have been negative on AI, many leaving OpenAI, Anthropic, Google, DeepMind, etc. There's a big New York Magazine article where they sat down with about eight of them, including Jacob Coxson, who's the 27-year-old who got a lot of attention cuz he quit Anthropic last month and said he I could there's a possibility that AI I kill a 10% chance that that PR could kill a lot of people.

5:49 And there were other similar researchers that have had similar moves and they all got to have their say in in these interviews with New Yorker magazine. So, you may want to take a look at that. And and we're going to see a lot of this. This is again in in AI, this is not unusual. This is just the latest beta folks who are saying what they want to say. But remember that the the doom versus boom arguments in AI have been from there from the very beginning from the earliest meeting with between Larry Page all night with Elon Musk in the mid 2015s where they basically lost a friendship on on their difference between the risks of AI versus the benefits. So, this is a very long decade plus kind of a scenario.

6:35 We've not seen this in prior AI tech waves and this will continue to be debated. I think nothing to be resolved over the next few quarters. It'll take years because we'll have to see what AI actually ends up doing. I've been in the camp that people like Jensen Huang have been at and China, which is let's just build the technologies carefully, methodically, year by year. And like any tech wave, it will take a long time to figure out the true benefits of this stuff and we'll have ample time, I still believe, to control the downsides and they don't need a whole mega set of policies because you generally you cannot anticipate what technologies are going to be used for. They always surprise you on the upside in terms of how they have an impact. That's been true from the PC to the internet to mobile phones, social media, and cloud, and so on. AI is going to be no different and it will take years to figure this out.

7:32 The third item I want to talk about are Meta and Microsoft. Both there's a piece on this in the information that's worth taking a look at, are working actively to kick their Claude habits, meaning they're both trying to their These are the two of the biggest corporate customers for Anthropic, and both are cutting how much their staff use Claude. So, they're token managing how much Claude is being used from Anthropic in their shops, not only for cost reasons or competitive reasons, but also there are a lot of options, including their own models, obviously. Meta has its own Musepark.

8:07 Microsoft has its semi-AI models and a bunch of others that they're partnering with, and they're trying to mitigate their expenses, both the internal spending at Microsoft and the article highlights for Claude alone was on track at least a billion dollars this year, and now that's by a down by a third. And Microsoft had made OpenAI the the default for a bunch of years because of their close partnership, but that's also also shifting as well. And Microsoft is now using a lot more of their own models and open-source models. Meta, the number of employees using Claude code fell by half from about 60,000 to about 30,000.

8:45 Part of that, obviously, because of layoffs, but mostly it's Meta's own coding tools running on their own Musepark 1.3 models. So, remember both of these are still big Claude users. They'll be represented in the S1 numbers for Anthropic's growth in terms of revenues. Microsoft's payment to Anthropic for its customers and Co-pilot features keep rising. So, the total spending with Anthropic is is roughly flat. So, take a look at that if you were interested. My take on all of this is, again, this is, as you'd expect, the logical moves by these companies to manage their costs and their dependencies and make sure that they're thinking through the tactical and strategic implications of how they're investing and using AI models frontier and otherwise, and that is going to show up in the budget. There's going to be a lot of ebbs and flow and it's important to track especially for those looking at Anthropic's IPO imminently between now and Thanksgiving. My overall take on all of this is we're seeing again clarity the the specific individuals whether it's founder CEOs like Sam Altman, the people critical of AI like the people who are fearful of of AI that we've talked about and the big companies who are trying to figure out with their budgets how what they're what they're saying on the big models themselves. Everyone's having their say and that's important to kind of track and understand.

10:13 Let's move to gadget AI. A little fun here developers it seems are leaning into Apple's upcoming duo fold phone. This is the one I've talked about a lot. It's going to be released on October 16th with wide availability on the 23rd. Remember this is Apple's first fold phone about eight years after Samsung has had annually eight versions of their fold phones and other companies Pixel on from Google has had maybe five or six or so.

10:44 But this is the first time developers seem to be getting excited by what they could do with a fold phone from Apple and they're showing a lot of creativity this article that I've highlighted that in in Axios give you some example one independent developer is going to basically recreate a Walkman in very clever ways with the sounds and the mechanisms and the tactical feels that he'll build in into his app. You're seeing a range of creative applications hackathons are being held etc. of apps being developed ahead of this this thing which is something fun to see it's something we've not seen that as much as I would have anticipated with other folds and there've been so many different versions of folds from lots of other companies on the Android side but it's Apple and they sometimes do provide the software tools and inspirations on the operating systems and so on. And developers are leaning into it. Remember Steve Ballmer famously chanted developers, developers, developers and in this instance we have, you know, over 60, 70 developers at the early blush who are saying, "Hey, we've got something cool to do with the Duo when you get one." So, the early adopters like yours truly and whoever gets to try out these things in the Apple Store or Best Buy will have fun to try and check it out.

12:03 So, it's fun to see this, fun apps being developed for the Duo. I'm all for it. Let's go to the questions for today. What do I relish most about a folding phone? Well, it's interesting. I would say I'm my behavior is consistent with how most people use them. About 70, 80% of the time people don't open their phone. The the the thing that I relish is that at any moment I could, that I could at any point open this thing and have a screen that's 8 inches diagonal, so it's a mini iPad, but for the most time it's just a small phone that stays in my pocket. It's just the luxury of knowing that you could use it that way and most people don't. That, I think, is one of the core aspects of the experience that is under appreciated.

12:50 So, I would highlight that and I think that's going to be a lot of the appeal for people who do get the Duo. It's like my love for the Apple Air iPhone, which is a lot lighter, a lot thinner than the 18 super 18 Pro Max over here. I I prefer this one because it's it's because I when I pick it up, it's what's not there that's fun to relish. It's just like with the with the folding phones, it's the it's the luxury of not need not opening the phone, but knowing that you could.

13:23 So, it's a subtle thing, kind of a fuzzy qualitative one, but I wanted to highlight that. Question two, what's been the most disappointing element about foldables so far? And again, the the excitement that developers are showing to build cool creative apps for the Duo is something that's been a disappointment not to have seen over the last half decade or so, where developers haven't done much creative stuff with with the foldable phones. They've made sure that, you know, their apps open in the two screens and so on, and you can share two apps on a screen, that sort of thing. But I've not seen creative things done, and that's been a little bit of a disappointment in the sense that in a market of 4-5 billion people, you know, millions of these things have been sold, but it's it's it just highlights how early these things are, despite having over 5-6-7 years of models annually from these companies. So, that it's just it shows you how slow some of the stuff is when the technology is way ahead of users and and developers. So, that is one of the things that is is a notable item here.

14:31 Well, today's AI Reset to Zero, which we went out this morning, it's 12:31, it's part two of our three-part series about the AI agent web arriving and how it interacts with the app environment and e-commerce and so on. A lot of detailed narrative and discussion on on this these changes that are underway that are going to be relevant to regular folk. So, do take a look at that if you have an interest. Tomorrow, of course, we'll be back with Ramblings 179, and of course, Reset to Zero 12:32, which is part three of the series that I just talked about. Thank you again for joining us today, AI curious folk. Stay tuned.