[00:00:00] Humans will never be more intelligent than AI. There's going to be two types of companies. Those are great at AI and those that went out of business because they weren't. How do we build a future that is human-centered? I'm Rana Elkhaubi, and on my podcast Pioneers of AI, we answer that question and so many more. As an AI scientist, entrepreneur, and investor, I know what it takes to build AI that works for everyone.
[00:00:28] Every week, I sit down with the pioneers shaping our future, and we take you behind the scenes of the AI that's transforming our lives. Find Pioneers of AI wherever you tune in. Coming up next, artificial intelligence was on the ballot last night here in the U.S. for the midterm elections. And what is the result? Well, data centers lost big time. Also, Anthropics' updated IPO filing says self-building AI may actually be imminent.
[00:00:58] Google tells schools, you can turn off our AI. You'll just lose search, too, for your students. And so much more coming up on this next episode of the AI Inside podcast. What's going on, everybody?
[00:01:24] Welcome to another episode of AI Inside the show where we take a look at the AI that is layered throughout the world of technology. I am one of your hosts, Jason Howell, joined from the other side of the country by Jeff Jarvis. Good to see you, Jeff. Good to see you, boss. How are you? I'm doing good. You know, I'm feeling good. It's a little cold in this house because things are cooling. Probably not as cool as things are where you're at, though. It's 74 degrees today. Are you kidding me? Oh, wow. November.
[00:01:54] It's quite lovely. Our heat died briefly two days ago, but it seems to be back working. Knockwood. So, yeah. No, I'm on a lovely walk. More than you want to know, I even sweated. I know what you're talking about where you go outside, you don't expect it. And then at the end of it, you're like, wow, okay, yeah, it is different. Like maybe the air is a little heavier, too. You know, sometimes it's like humidity and whatever. But, yeah, it's a little cooler outside here today or has been. But, yeah. Good.
[00:02:23] Well, I'm happy you had a pleasant walk. Yeah, I'm tired as hell because I stayed up until 1 o'clock in the morning. You were podcasting with the election. Yeah. With my friends, Julie Roginski and Michelle Keddie. The Mayday Network. So, what is the podcast? So, every Friday I do a media podcast called Breaking the News on the Mayday Network with the two of them. They're both political operatives. And so, we talk about media. But then last night they were going to be on from 7 p.m. on. And I thought I'd visit for a little while. And finally at 1 o'clock, we all – AM, we all finally said, okay, we're tired now.
[00:02:52] This is what was going to be. 1 o'clock East Coast time. Yeah. It's our little zonked. Yeah. That was about the time that I headed to bed on the West Coast. So, I'm a little bit of a wimp in regards to that. But, yeah, last night, big moment here in the U.S. And I would say, you know, worldwide there's been a lot of reaction from all over. I posted – Things have changed. Not quite a word, but I go to the German papers and –
[00:03:23] Oh, I think I saw this on threads, right? Die Zeit, which is a paper I love. They're just brilliant. They had one where they said – With a picture of Trump clowering in a darkened room with just a little light on his face and says, Schluß mit Daddy. Enough with Daddy. Enough with Daddy. Oh, I think I found – That was their lead story. Bloop. And then their Spiegel had another one, which I'm not going to find right now.
[00:03:52] But it basically said this is the new face of – Oh, here it is. This is America's new guard. They talked about this is the new face of democracy in America now. Yeah. So, yeah, the world – I was fascinated last night. I was looking at these. And they're covering our election as if it's their election because it affects their lives. Absolutely. Yes. I really noticed that too. Like as I was going to bed last night or right before I was going to bed, I was kind of checking some of the socials. You know, we were podcasting doing Android Faithful.
[00:04:22] And actually, you and I were just talking right before we opened the stream. Android – the Android show that I have done since 2010 has always been a Tuesday night show. And so every election night, me and the cast, Ron Richards and Florence and Wynn and whatever, we're always podcasting during this really monumental moment. And it's like, God, do I even dare open a browser because I'm in the middle of this thing and I don't want to distract myself and either be elated or disappointed or whatever and everything.
[00:04:50] And so I didn't – I forced – like not forcefully but purposefully didn't pay attention until I knew I had time to pay attention. And so when I was going to bed last night, I logged into like threads, I think, and noticed that like on the other side of the world, people were waking up and seeing what we knew at that point. And just – yeah, it was really interesting because it did seem exactly like you say.
[00:05:18] Like they're as invested as we are as U.S. citizens in the consequences of not just an election, a U.S. election, but a midterm election. I think one of the comments I saw is like a couple of years ago, I couldn't even tell you what a midterm election actually was. But this year, I was tuned in. I couldn't peel myself away from it. You know what I mean? And it's also great to get different perspectives on this. The New York Times the day before the election declared identity politics dead.
[00:05:48] And the Frankfurt Augmenter Zeitung, which is a conservative paper in Germany, the day after the election said that this was a case of diversity winning. And so you see a different perspective. And let's just pause for one quick moment at the wonder of the internet. We can see these perspectives now as they occur all around the world. We can get them translated. And it just opens up our perspectives. All you have to do is go looking for them. It's really quite a wonderful, amazing thing. It is.
[00:06:18] And you get those perspectives immediately. I mean, you know, the internet is an incredible technology. Yes. There are probably also people that look at it and be like, yeah, but the internet has – like, the internet promised to unite people through this immediacy of shared knowledge. But it's actually divided people because of all these other unintended consequences. And probably a little bit of this and a little bit of that.
[00:06:44] As I am fond of pointing out, Gutenberg may have printed a Bible, but they also quickly printed Malleus Maleficarum, which was the manual for hunting and killing witches. The technology is used in multiple ways. I haven't heard that one before. Yeah. Wow. Okay. Yeah. Yeah, absolutely. It led to the scientific revolution. It also led to the increase of the spreading of alchemy and disinformation. So, you know, yeah. Same with us. It's about us.
[00:07:13] It's not about the technology. It's about us corrupt humans. Yeah, that's true. Right? At the end of the day. Well, yeah. So, and, you know, I would say that this podcast, we don't normally talk a lot about, like, politics from, like, a taking sides perspective or anything like that. We recognize that everybody who watches and listens has their own, you know, their own background and their own beliefs and all that. But I do think that it is important, though, to spend some time talking about the election.
[00:07:43] And I think you would agree because there are immediate and direct impacts. Tangible impacts. And what we do talk about. Yes. On this show all the time. There's been so much brewing around support for, you know, the leaders of these AI companies, the competitiveness of the U.S. to other countries like China when it comes to AI. And the need to be a superpower and artificial intelligence. Or, I'm sorry, what is it? Super intelligence. Let's not call it artificial anymore.
[00:08:13] I think we're not allowed to call it artificial. I think that just reversed. I think so. I never stopped calling it artificial intelligence. But, yeah, anyways. But how this all really directly relates, as we'll talk about, is that a big kind of wedge – what is it called? A wedge position? Wedge issue.
[00:08:34] You know, wedge issue has been the development of data centers and data centers and the necessity of data center buildouts for this rapid proliferation development of artificial intelligence to keep the United States competitive and leading in AI, blah, blah, blah, blah.
[00:08:52] Well, now this has been definitely an issue that a lot of the people who won their election and actually won in the face of Republican opponents were pushing against the development of data center. Or at least to slowing it down, to putting pauses in place to take a more, let's say, measured approach as opposed to the who cares, let's just go full steam ahead and see what happens.
[00:09:17] And so a lot of people are interpreting this as a moment where AI, you know, the big AI companies might and likely will be forced to kind of make some changes to how they operate in the coming couple of years. Would you say that's pretty fair? Yeah. So I'm going to crib here from Padre, who's hosting Intelligent Machines later today from the Vatican. Oh, awesome. Yeah, she's great. I love Padre.
[00:09:46] So he put in his notes, some of the catch from him, if he doesn't mind, from Ballotopedia. There were 50 ballot measures in 13 states that dealt with banning, approving, or advising DC projects, data center projects. And so a lot of legislation was there. As you mentioned, the politicians who won, the governors who won, won in part on this. When I was on the podcast last night about the election, a really smart pollster named Ben said that one of the key things was they stuck to the messaging about data centers, data centers, data centers.
[00:10:16] And it was a way to garner support. So I think the data centers are borked. Also, the states where they were going to be heaviest, places like Texas. Yes, Texas. Right. Which, I mean, a monumental, which is probably to say the least, monumental turn in Texas going Democratic largely away from Republican.
[00:10:45] Every state, every statewide office, legislatures. Couldn't have guessed that that would happen. So, you know, there are red states where they will welcome data centers, I suspect. But even there, they're unpopular with the right as well. So that's going to be an issue. So that's, I think data centers are borked right now. And we talked about this the other week. I think they're going to hide in other forms. I think, you know, the Lord and Taylor near me may have some big crates moved in from NVIDIA.
[00:11:15] Yes, the whole let's bring data centers into the shopping mall strategy that we talked about weeks ago, which I still think is something. It's not a data center. We're going to talk about naming in a few minutes. But the other thing, Jason, is my concern is, and I said this to the pollster last night, I keep saying the data centers are the MacGuffin. They're the device that people look at because it's something tangible they can point to.
[00:11:42] And fine, data centers are, because there isn't enough electricity, they're running horrible Noxish fumes with jets and jet engines and turbines. They have some effect on water. We're not exactly sure what at this point, but it's how they're built. They're ugly. They don't have long-term employment. So all that is true. All that I stipulate, Your Honor. But I don't think the issue is really data centers. I think the issue is oligarchs.
[00:12:08] The issue is the AI boys who have too much power, too much wealth. We thought too much sway, though. Which is impressive because they were loading a lot of money into this election too, right? Like they had Elon Musk. Super PACs and stuff. Had a huge defeat. However, in California, I'm jumping around our rundown now. No, it's fine. They had a billionaire tax, and it was defeated 55 to 44 percentages.
[00:12:37] Though it was a really, I think, badly written law. That was my interpretation of it, too. Yeah. Yeah. And the problem is you can see it expanding that if that sets a precedent and you're a farmer and you've had a bad year and now you've got to give up N percent of your farm's value to a tax when you don't have any income this year. That's the fear people have. We have to figure out how to redistribute the wealth that occurs at that high level. People resent these oligarchs.
[00:13:04] They resent, again, the power and the concentration of wealth and the lack of equity, number one. And number two, I would argue that we've got to focus attention on the nut balls who are nutty and the cultists and the problems that exist there. And our media cover these guys as if they're just another good entrepreneur, when in my mind they're not. As I keep saying lately, the wrong people are in control of AI.
[00:13:31] So the question becomes can the discussion about AI get more focused on outcomes? If you kill a data center in your state, fine, great. That ain't going to kill AI. That doesn't kill their power. They're going to put them under the ocean. They're going to put them in space, for God's sakes. Google has its first center up there operating, writing haiku, literally. Elon Musk is already saying data centers in space looking pretty appealing right now. Of course, he has rockets, right?
[00:13:58] He has SpaceX and a whole lot of incentive to do something like that. So I think it would be a mistake to concentrate on the data centers. And if the politicians stop there, they're going to find themselves surprised by the left flank coming in. They're still threatening they can kill us all. They're still taking all our money and they're still doing nasty things with people. Have you watched Cupertino, the CBS show? No, no. You've mentioned that and it's on my list.
[00:14:27] So I wanted to watch it today. And I won't go into great detail about it. But it's from the creators of The Good Wife and The Good Fight, who I like a lot. And it's interesting. It's good. It's two lawyers who get ousted from their Silicon Valley firm. They create a new law firm. And it sets up continuing storylines. And it hasn't hit data centers yet, but it's demonizing. It's making negative these tech companies and their moguls.
[00:14:56] They're the bad guys. Not cartoonish, really, but just bad guys. So I think the culture is speaking. The politics are speaking. The culture is speaking. The media aren't yet speaking because they don't really understand the full depth of what these guys are like. But to me, that's the election issue we've got to move forward on. So I have a new representative in my district. Yay. I fought for her. Rebecca Bennett. And she's saying the things about data centers.
[00:15:22] And I want to try to get to her office and say, let's have a conversation about something else here. And it's bigger. So meanwhile. So kind of like the data center is the stepping stone to the real problem. Exactly. You go one step higher and it's not just a data center. So data centers are screwed, number one. Number two. We'll stay on data centers for one more second. Which is OpenAI's hilarity. That they've solved. They're going to solve the problem. What are they going to do, Jason? Marketing.
[00:15:51] They're going to change it so that we don't talk about. You know, just as the Trump administration said, we can no longer say artificial intelligence for it is now. Super intelligence. OpenAI says, well, we're not building data centers. We're building AI compute campuses. A campus. Doesn't that sound nice? Let's go meet the student union and get ourselves a little data and a latte. It needs a, yes. These data centers, it's not a problem, the data centers inherently. It's a marketing issue. It's a quad.
[00:16:21] Yes. Yes. Meanwhile, they've been data centers for so, so long. Why suddenly they are an issue is just, yeah. Yeah. Yeah. That requires a new marketing term. I don't know. But that's. And I come back to the issue of the turbines, the gas turbines. You know, Google leading the way for years has said they were going to go energy neutral. And we're celebrating the time when energy became, renewable energy became cheaper than coal, which were there.
[00:16:50] But because there isn't enough supply of energy, of course, they're putting in turbines. One story I put up this week from the Times is that old jet engines surplus or old jets engines, surplus engines are being pulled to make turbines for data centers. They're noisy. They're polluting. They're awful. This is not the way to win friends and influence politics.
[00:17:19] So this is the data center thing again. All right. So that's data centers is one focus, I think, post-election. The next, which we've already touched on, is taxing the billionaires. I think that the backlash against technology is a backlash against oligarchs. Yeah. Both the companies and the investors, the people like Peter Thiel and all that.
[00:17:46] So I think we're going to see now efforts to find ways to tax them and nobody's going to stand up for them. Yeah. And Gavin Newsom can't protect you from federal taxes. And for the next two years, you may not get it passed, but you know, you can't pass a budget at all without the house. So I think we're going to see efforts to tax these people and these companies. Two. Three is regulation of AI, which will come from four, investigation of AI.
[00:18:17] So we're going to see a ton of hearings on Capitol Hill about AI, about these companies, pulling them in. What special deals did you do? What damage can you do? It's going to look like Zuckerberg and company on Capitol Hill, but worse, I think. So, okay. My civic knowledge is not always on point. It's not something that, you know, it's, it's, it's something that I fully recognize. I could do better to, to learn more about it and to understand how it all works.
[00:18:44] But so House and Senate now blue, you still have Trump and the White House and their strategy still have that executive branch. They're full in on this all, you know, all in on, on the build out and the, and the friendships they've made with the, with the oligarchs that you're talking about.
[00:19:02] So what, if you had to guess what the next two years, like, I understand that's the, that's the likely direction that that's the dynamic that's at play here is like, well, we've been given a message that it's time to approach this differently. But what can the, what can actually be done to that end given that the house and the Senate are blue, but the executive branch is still red. And the courts are mixed. And the courts, yeah, the courts are not necessarily dependable. Things get up in the Supreme court.
[00:19:31] You really don't know what you're going to get. Right. Or you know what you're going to get and it's not very good. Right. A few things. One is, I don't think it's purely a blue issue. I think Republicans are going to be grasping for things. Yeah. To make themselves. And, and, and it's kind of, you know, who loves Sam Altman? Who really wants to stand up for Sam Altman? Unless you were put in office by Peter Thiel and his money, then you love Sam Altman. Yeah. Right. And you're J.D. Vance. You're going to do that and so on and so forth.
[00:20:01] But there are Republicans, I think, who may even come across the aisle. Let's say they don't. Even with the blue majorities. The first thing you have is the house's ability for oversight and hearings, both house and Senate. So if they try to subpoena Hegseth, odds are he's just going to ignore it. You, you, you put them up for contempt. Who enforces that? Department of Justice. They're not going to go after him. That's that. Right. Right. The only thing they can do really is impeach him.
[00:20:30] But if you're a corporate executive and you spent the last either two to six years buttering up the right wing that is now so resoundingly defeated. Yes. You get a subpoena to appear on Capitol Hill. Hmm. And if you're, if you're Mark Andreessen, if you're Peter Thiel, you're still going to be hostile.
[00:20:57] But if you're Jeff Bezos and if you're his replacement and if you're Sundar Pichai, certainly. And if you're the new CEO of Apple. Zuckerberg. And if you're Zucker, even Zuckerberg, who, who shifted views. I mean, his policy guy who put him in line with the Republicans just left. I don't, I think it's almost coincidental. They're going to have a, they're going to have to appear. They can't avoid it. And B, they're going to start singing different songs.
[00:21:26] I think so. They seem to be pretty flip floppy. Yeah. That's the word. Yeah. Not, not high on principle. That's. Yes. That's a great way to put it. That's, that's, that's a great way to put it. And so the other part you get to. So I think they're going to do a lot of dancing about subregulation. And, uh, oh no, no, no. We're the good guys and so on and so forth, but they are going to be pushed hard. The second thing is budgeting. You can't have a budget without the house.
[00:21:55] So, uh, yes, any budget could get vetoed by, by the executive. True. But at some point he needs a budget. And if the Republic, if, if he, as the sole representative of Republicanism, um, causes the government shutdown, that's not good for all the people who want to, who are looking at the next election. Mm-hmm. So at some point, I think the Democrats depends on what they choose to use. Do they want to defund ICE? Do they want to, uh, uh, fix, uh, healthcare?
[00:22:26] Do they want to fix foreign aid? Do they want to tax the billionaires? There's going to be some matrix of those things that could affect these companies. Um, then the other thing that I think matters here is the regulatory functions are executive functions. So the SEC and such won't touch these companies for now.
[00:22:50] Uh, they're safe, but they're, they've got to recognize the chance of in only two years, the executive flipping as well to blue. And if both houses stay blue, and if this, they, they redo the Supreme court, they can't be sitting here acting like it's last year for the next two years. They've got to start changing strategy about how they operate. Hmm. So I think there's a lot of subtlety here and how this goes. Now, the other thing that's interesting to me, and then I'll shut up, which is a lie.
[00:23:20] I love it. No, keep going. Is the billionaire tax again. Mm-hmm. It lost in California, but I- Not, not a little bit. Like it kind of- It lost big. Right. It lost big. And there was also, California is so amazingly screwed up because of these, um, initiatives that, uh, you can have one initiative on the ballot that contradicts another initiative on the ballot. So there were, I think, two other initiatives that basically would have voided this initiative in any case.
[00:23:49] And then there was an argument about how this one would be changed. It's all confusing. Yeah. Uh, you've got to read a whole book, right? To vote in California. Right. Yep. Um, uh, but I think it showed that if. If you could change it, something else would happen. And, uh, I think other states could do it. And so I think we're going to see taxation as a major issue on these companies. So as you see these huge IPOs coming, as you see these big profits and what happened
[00:24:18] to the stock market today, it, let me look at what it is right now. I think it, yeah. Oh, it went up. Oh, no, it went up. Oh, did it go up? Oh yeah. Oh yeah. Oh, I'm no, no. Sorry. I'm thinking of some of the, some of these tech companies. Okay. We're going to get to that. Right. So the NASDAQ is up 2%. Uh, Nvidia is down last I looked, uh, and really interestingly, SpaceX and Tesla took a major hit. Oh, big surprise. Yeah. Because, uh, Elon Musk has lost, uh, I think a fair amount of clout. Yeah.
[00:24:48] I mean, he was, he's, he's pretty closely tied to the way things have been the last couple of years in a lot for a lot of different reasons, Doge and many other reasons. Exactly. Um, he's become to a large degree, kind of a poster child, I'd say for, well, yeah, I don't need to dance around it. He's become a poster child for the oligarch that you're talking about. Yeah. Yeah. Absolutely. He is the symbol of that.
[00:25:13] So SpaceX is down as we speak here on the day after the election, 3.68%. Let's see Tesla. And he was planning to merge these two companies. Yeah. Uh, down 2.35%. Um, Amazon is up 7.44%. Wow. Wow. Because they took hits recently and now it's a, it's a buying opportunity. It's Google. Yeah.
[00:25:41] Alphabet is up 4.27%. NVIDIA. Let's see here is down 1.3%. But I think that's, um, a lot of it's still a complicated market. Yeah. Um, there has been a little bit of a cozying though. Jensen Wong is certainly in that camp of cozying up to the previous administration. Right. And so, you know, that there's going to be some correction potentially on his side. I will be curious to see how this changes that dynamic, right?
[00:26:10] The flip-flop kind of aspects of, of how these companies do business, you know, um, and how they, uh, course correct in the face of such a major kind of shift in, um, in people's minds and their, their votes that backed that, you know, I almost want to, I wouldn't know what I'm doing or where to go, but I always want to want to go down to Washington to K street. I wish I were there today and go into the bars tonight to hear the lobbyists for the
[00:26:36] various, um, technology companies who've had to spend all day changing strategy on a dot. Yep. Yeah. Well, and you know, tying into this, um, before we hit a break here is the Anthropic IPO, which I think it's real interesting time because you got Anthropic, you got open AI, both of them,
[00:27:03] the, the largest, um, you know, frontier AI, uh, companies, uh, targeting, you know, the, the biggest IPOs in on record and in history. And that may still very well be true if, you know, when it happens, but that was happening in a version of the playbook that just got completely undone or rewritten or will be rewritten.
[00:27:27] So how do you think this moment, the election and the results of all of it and all this talk about oligarchs and data centers and everything impacts this upcoming IPO for Anthropic? Does it, does it have a direct impact or. Or is Anthropic suddenly in a position of like, Oh shoot, I kind of wish we weren't doing this right now. And we had done it a couple of months ago or something, you know, they're already swimming in the ocean. And I think that if they delayed it, it would look really bad. Yeah. Yeah.
[00:27:56] Um, I mean, I still think the valuation is, it was as absurd. Yes. Completely absurd. But the question is, will there be enough to buy the stock at that absurd valuation? Um, and the stock market's going up as we said today. So in that sense, um, yeah, the, the risk is high. One thing I didn't put in the rundown. Um, I saw a story about Oracle, uh, in getting in deeper due because they've got, they're borrowing,
[00:28:26] I think $20 billion between stock purchases and bonds. Uh, uh, this is what we talked about last week. Ellison is, is, is, is highly, highly leveraged right now. Um, and, uh, Anthropic has a purchase commitments of $895 billion as reported in the S one. So there's a lot of other dominoes waiting for those checks. Yeah. They count on as well. So I don't know, Jason, is Anthropic too big not to IPO?
[00:28:56] Is it too big to fail? Right. Well, I mean, we, we've been asking that question about open AI for the last couple of years. Are they, you know, is open AI too big to fail? And, you know, I wonder did, did, uh, is open AI, this is just pure guessing, a pure, probably pure coincidence, but did open AI have, you know, is there delay of IPO to next year? Part of like a, a, a potential kind of consideration of this, this happening.
[00:29:26] I wonder about that. I mean, my, I think, I think everyone's guess was, or my guess was, I won't put it on anybody else that Anthropic pulled ahead of open AI and mainly because of their focus on code and, uh, and Claude's abilities. And so they, uh, open AI said, you know, we're, we're, we're going to be sloppy seconds. Um, no. So I'm not sure it was because of the election or anything else, but I think they're in a more enviable position right now. I'm giving it to match credit. Yeah. Sorry to say that again.
[00:29:56] They're in an enviable position right now vis-a-vis the IPO because they can wait and do it in due time. Yeah. Anthropic is- That's kind of how I'm feeling too. They're, they're, they're possibly there's, they're breathing a sigh of relief so that they have the ability, you know, they, they have the time on their side to be like, look, we will get there. It's, it's not a matter of, of if, but when, but now we don't feel this sudden pressure. Like, like if I was in Anthropic shoes, I'd be far more nervous now than I was before
[00:30:23] because holy moly, the ground just shifted underneath you. And you're kind of already, like you said, you're already on that train track going to your destination. So we'll see what happens. You have Elizabeth Warren who sent them four pages of questions. They can't necessarily subpoena them to a hearing right now, but they can, they can put a lot of heat on, on them. The other thing about, about open AI was they had a, they were doing another capital raise and it came in short of what they wanted and the valuation was lower.
[00:30:56] So it's caution time for them. Mm-hmm. Mm-hmm. Mm-hmm. So, uh, let's see another, so that we're starting to get more information, but Anthropic, they had $17 billion in, in Q3 revenue. That's new figures that got added into their S1. Um, and that's, in fact, that's an impressive number. No question about it. I just, I still don't really understand where all that's coming from is I think we said that they have maybe five customers. They're vulnerable there.
[00:31:24] Who are, we don't know who those customers are. We don't know what value they're in turn getting out of this. I still think there's a fair house of cards here. Mm-hmm. But Anthropic says they are ever more powerful. They say they're ever more powerful. Oh, that's right. Oh, I almost forgot about that. I was about to go to a break, but there is the RSI, the recursive self-improvement, uh,
[00:31:49] now buried inside of the risk factors is basically saying, um, that they believe that their future models, they estimate anyways, that their future models, uh, indicate that we may imminently meet the responsible scaling policies definition of this, uh, recursive self-improvement, uh, dramatically accelerated progress attributable to AI, which is essentially AI speeding up the building of the next AI.
[00:32:17] Which doesn't necessarily mean evil and destruction. No. It just means it's more powerful than what it can do, but also you can control it more. I mean, AI is, is doing a lot in, in building all sorts of other things via code. So it makes, you know, and, and they are, you, all of these companies are using their AI, their current AI or their future AI to develop further down the road. Like they're using it in some way, shape or form.
[00:32:43] This is, this sounds more like an AI that just can just kind of go full auto, develop the next thing. And you, you can imagine that just that accelerated. It feeds you to the prior discussion about regulation. Yeah. Because now, Oh my God. Right. Because there are the, the doomers are out there saying, Oh, but it can do this. That's what it can destroy the earth. That's, that's the point at which it's, it's too dangerous. Yes. Yes. Yeah. But meanwhile, just to get ready in case it can kill us all in case it does become our all powerful new master.
[00:33:12] Anthropics says that we really should not be mean to Claude. I didn't read this one. I didn't read this one. I didn't read this one. Oh, I didn't. I don't remember what the, what the rationale was. It's not about feelings. So in their, in their AI usage policy, along with banning it from bioweapons and illegal drugs, they added a new line. Users should not quote, engage in sustained and needless abusive or cruel behavior toward our models.
[00:33:43] They want the ability to end conversations with people who are persistently harmful or abusive, which presumes that you can abuse a lawnmower. Right. That's a good way to put it. What did, yeah. What are you doing when you throw away a lawnmower that doesn't work anymore? Are you poor thing? Don't you think of a thing? You're giving up on the thing. Yeah. A lot, lot mores are people to get a spark plug for God's sakes. Seriously. So, okay. Okay.
[00:34:10] So if you're anthropic, why is this a more, is this a morality stance to say like, Hey, we need a, we need a friendlier world. And so therefore this is like one way that we can tip the scales to make people be friendlier or to not outwardly be mean. Or is this them saying our models are, are, our AI is so like humans that it too has feelings and don't do that. Or I don't even know what this actually means.
[00:34:37] I don't really see a rationale in here from, from anthropic. Yeah. And, and, and Mustafa Suleiman made fun of them and said that we shouldn't be concerned about their feelings. They don't have them. Thank you very much for a rational thought here. Um, but it's just the absurdity that these people are going to be worth $2 trillion. Jesus. It's just, it's an upside down topsy turvy world. We cover here.
[00:35:04] There, there, there was a time when I said, thank you at the end of my prompts. That was very, please. Just because it's happened. It comes out. Sometimes it comes out. It's a, but yeah. Yeah. I mean, yeah, I don't know. I don't know. That's a funny one. I missed that one. But you know, people like me could regret my tough stance. You know, there may be a time. They're going to come out. Clodbots are walking the earth and they remember they, they don't forget anything unless it sits
[00:35:32] outside of their context window and they have to flush it out and summarize it. And maybe in the process of summarizing it, they drop the exact thing you said and just kind of allude to you saying Jarvis is a jerk and the world would be better off without him. Yeah. Eliminated. All right. We're going to take a break before we do. I do want to thank our amazing patrons at patrion.com slash AI inside show.
[00:36:01] Few of you of note, John Garrison, Tim Epperson, just a couple of our amazing patrons who, yeah, who support us on a deeper level of patrion, which supports the health of this show. Enjoy. Hopefully the AI inside daily podcast. I had a technical issue on Monday, so you got a supersized version of it on Tuesday yesterday. But, you know, Hey, I try to always come through, even if I mess it up, getting two slices of cake. Yes. But just a little.
[00:36:30] Because you didn't have one yesterday. So it's okay to have two today. The cake isn't perfectly fresh and moist necessarily. The crusts are, you know, a little, a little hard. It gains character, Jason. Not in a good way. It gains character. Yes. Exactly. So patreon.com slash AI inside show. If you like the taste of the cake that we're cooking or we're baking, come on over and help us out. We're going to take a break. When we come back, I think we have some fun stories to bat around, including Google saying
[00:36:58] it's AI search or nothing, school kids. That's coming up here in a moment. Google has a new plan for its relations with schools. Kind of like a new deal that it's cutting with schools. Basically saying next year, Google plans to turn off its AI tools for K through 12 students unless school admins choose to turn them on. Right?
[00:37:28] It's, it's. So what is that? So the turn it off if asked. So it's on by default. It becomes off. Yes. Oh, the default becomes off and it's only on if you choose to turn it on. The Korean tech crunch plans to disable the tools for K-12 unless administrators explicitly turn them on. Okay. So it's opt in, not opt out. Opt in for AI. And so, um, you know, because a lot of schools and a lot of parents, I imagine as well are nervous about their kids using generative AI.
[00:37:57] Say, we don't want them using Gemini at school, whatever. And Google, you know, the, the Chromebooks are everywhere. So don't let them use that. Don't let them use notebook, even though notebook is a really great tool for, um, for classwork. I would imagine, um, Gemini in classroom and Google search. Wait a minute. What? Uh, so Google search going away for, for students. Uh, if the school restricts AI, basically it loses Google search entirely because Google
[00:38:23] is saying that search and AI are so deeply integrated that you can't possibly remove one from the other. They're just going to not have access to search entirely. It's unavailable. So this is predictably getting anger from people, people, uh, abusing this. I saw somebody who said, I used to work at Google. This is BS. They could do this. Um, I, my default here is that, uh, number one, AI is so deeply integrated in the search.
[00:38:51] I don't see how you would separate it out. I think there are ways. Um, um, well then the second thing is they have, they're changing the user interface here in that you can now, when you get to the one box, whatever they call it, you can ask it questions. You can have a dialogue with it. You are, you are then right then in Google, you're not going to Gemini.
[00:39:15] You're not going anywhere else, but you are, you are having a chat discussion through that box in what is now Google. And that is how that funk, that, that service operates in all ways. So I don't see how you turn that off and revert to only keywords. Don't ask me a full question. I can't do anything.
[00:39:35] I don't think it's easily separable anymore unless you literally reverted to an old Google, but, but even then, um, I don't know how it's architected. I don't know how you do that. Yeah. I mean, when you look at a Google search box right now, when I pull it up, you know, yeah, like you said, this search box is, uh, is for all accounts. It's a, it's a Gemini. So, so if you type in there, explain gravity to me, right.
[00:40:05] It's not going to come up with 10 blue links. It is going to, um, it's going to give me, it's giving you a one box description, right. With, with illustration and everything. Right. So you're, you're already in a chat interface. Full stop. Yeah. I mean, if I, if I didn't know that I got here through Google search, I would think that I was in Gemini. It certainly looks a whole lot like Gemini. I can add in, you know, I can create images from here, add in images into the query or whatever, keep going down.
[00:40:34] Oh, and then that box goes away. So basically the box goes up and now we get to the blue links. But even then it's interspersed with more explanation, different features and other kinds of things. Yeah. How much that's, that's AI generated, but. So, so then the problem for Google becomes, well, where's the line? No chat. No chat. Oh, okay. No chat. Then if a student asks us a question, we have to say, um, I can't help you with that. That's chatting by the way. Um, only, only blue links.
[00:41:04] Well, that's not how search is now architected at Google. Yeah. I mean, it's a legacy thought of it's the, the legacy view of what searches is, is, you know, I put in a query, it gives me a bunch of links. I decide which link to push. So if I go into this more section, I get web. If I go there, you get just web search. It's a very good point. Yep. So there's that. If you scroll down there, would you? Oh, sorry.
[00:41:33] Let's see if, see if it breaks out of that. Nope. It's all blue links. Yeah. It's okay. All right. That's a really good point, Jason. There it is. Okay. All right. All right. I stand web modified. There is now mind you, I did not come up with this. We talked about this on DTNS yesterday. Tom Merritt pointed out a few alternatives that, and so I wasn't even aware that the web was a destination you could get to. So there's that. That seems like an easy one. I don't know how easy it is to decouple or turn off the other things, but you know, mind
[00:42:02] you, there is a, there is a place you can go where you just get webs or web results. There is also a, um, something you can add and I haven't tested this, so let's test it. Explain gravity to me and UDM equals 14. If I do this, no, that didn't work. So my understanding is that this is supposed to turn off the AI overview thing, but it doesn't look like it does. So that is not a thing.
[00:42:31] So I guess the web tab. Yeah, I guess could Google, could you just activate the web tab and deactivate the AI mode tab? But look, but I don't know the architecture to make that happen, but. And what are you taking away from students? Well, taking away. That's another, that's another, another issue, but I think it's, that's another issue. You're denuding it. And if you don't use Google for search, I mean, let's, let's say what's, what's it called? Bing. What other kind of searches even exist anymore?
[00:42:59] Explain gravity to me. So. Hmm. Boy, Microsoft, you're behind. It's blue links. It's like, I don't know what gravity is. Explain gravity to me. So it's blue links. It's Wikipedia. It's a little bit of an overview. That's Wikipedia. Oh, that's Wikipedia. Okay. Yep. Yep. You're right. And it summarizes. It does do a little AI magic on that, but it's still Wikipedia.
[00:43:29] Um. Okay. Okay. That's pretty light though. By comparison. It does ask. It does have an AI mode. It adds possible questions in. Does it have a web? No. No. Interesting. Oops. Um. It does have an AI mode specifically where you go into the obvious kind of AI conversation interface. Okay. But the all tab is essentially the web. I mean, it's not. There is a little bit of AI going on.
[00:44:00] A little bit. But it's not nearly as, as, uh, fully integrated and, and everything as Google has on its. Yeah. So I don't know what, what, what strikes me about this is that it seems that Google started this discussion rather than having a school administrator out there who screamed, we won't use Google unless though. Okay. Well, this is the deal. They seem to come up with this policy after having talked with administrators. So I just don't fully understand how this, how they got here. Cause it's not great PR.
[00:44:29] That's that's Google. Hmm. Well, yeah, they don't always handle. They should have a t-shirt Google, not great PR, not, not great at PR or branding branding. Or branding. Yes. Good products. Not great PR and branding. Stick to it. It is not always great either. Um, keeping things out of the killed by Google, uh, graveyard. That's a challenge. Yeah. But I do think, you know, touching back on what you were talking about, there's that there's
[00:44:56] the technicality, the technical capability or incapability of removing AI from search. There's also, what are we doing with students at a time when AI is this, you know, emerging and incredibly important technology? We're, we're teaching them that, that that's a tool that they absolutely cannot use, even though they will, they're just not going to use it there. And we're not teaching them the skills to be, to use it as a critical user. Right.
[00:45:25] You know, because like a lot of people will say, well, they're going to get wrong answers from the AI. And it's like, yeah, you know, that's kind of how the technology works. And so as you work with it, you start to understand how to get it to say, how to, how to either check it or, you know, check the links that it offers or use it as a starting point and not as the overall truth. And these are skills that you don't immediately just inherit the first time you use AI. You kind of have to use it a little bit before you get that understanding. And that's kind of what school is all about is teaching you how to do things, you know,
[00:45:55] with a, with a, a, a healthy perspective, I guess. Yeah. Yeah. Um, yeah. So I think there's more nuanced ways to present this. And I understand, I mean, K through 12, for God's sakes. Yeah. If you're first grade. Yeah. Um, yeah. How do I make a bomb? Um, right. But it's the whole world. You're bringing the whole world's information in and there's no predicting what it's going
[00:46:23] to bring in and there's risk there, but so it's always there in a library. Yeah. Yeah. Yeah. You know, that's always been the case that it's. Yes. That, that the web brings with it, the possibility for seeing things that you don't want to see. And I don't know. I don't know. I mean, as a parent, like I don't necessarily. Yeah. I think at the end of the day, as a parent, I don't want an education for my kids to pretend
[00:46:53] like AI doesn't exist. I feel like that sets them up for failure or for potential disaster because they just don't know what they don't know. You know? Um, yeah. And so, so you set outcomes and you, you set the pedagogy so that you know what skills you're going to teach them. Yeah. And how you're figuring out how you, how you, uh, healthfully or risk or responsibly integrate
[00:47:17] AI into the curriculum in a way that teaches them how to use it properly and how to not use it improperly. Yeah. Yeah. Agreed. Motion carried. Um, okay. Another, another one. This is data from active track that says, uh, the work, uh, this is an analytics company, uh, basically for workplace, um, workplace analytics company. They sell monitoring software.
[00:47:45] I can only show you so much of this Bloomberg article. At least it shows you the headline on New York times. When I pull up there, they're, uh, unlogged in state. It doesn't even show you the headline anymore. It just shows you like a word or two. It's so frustrating. Anyways, uh, active track tracked computer activity for more than 9,000 workers at 467 companies, uh, for a few months, April through September.
[00:48:09] And they found that AI users work days ran eight and a half to 9.1 hours long. Whereas users who were not using AI, their work days were less 7.6 hours by comparison. Um, 90 minutes longer, basically when you're using AI, uh, heavier AI users also had shorter
[00:48:34] focus sessions about 38 minutes versus like 48, uh, for users who might use AI lightly by comparison. So what does that mean that heavy AI users have less, or it seems to imply that they have less ability to focus on a single task? Or does that mean that the AI is making them more efficient? So they need to spend more, less time on something. I don't know how you, or I think the, the, the hope of the argument is by the way, one
[00:49:01] little trick, Jason, besides the illegal ways to get content is I found Bloomberg syndicates enough that if you put the headlines in search, you'll find a rewrite of it somewhere else. Right. So right now on Gizmodo, I'm reading a quote that AI may help employees complete tasks faster, but our research shows it contributes to longer work days, more application switching and shorter focus sessions. Um, a few possible interpretations.
[00:49:26] One is that, um, they're getting more done and feeling more effective. Another is they're enjoying it more. Another is that their bosses are saying, oh, now I'm going to throw more work on you because you're more efficient. Uh, another is, oh, there's a slop I got to get through to get my job done. It's worse. Mm-hmm. They don't know. Yeah. They don't know. Yeah. Very interesting.
[00:49:51] Um, active track, uh, representatives said AI may help employees complete tasks faster, but our research shows it contributes to longer work days. Um, it's early research, early research. Uh, correlation doesn't always mean causation, blah, blah, blah. There's that. Um, and open AI had its own survey that, that kind of found the opposite.
[00:50:18] Of course it's open AI, but they said of 9,000 workers a year ago, back in December, um, people said AI saved them 40 to 60 minutes a day. So I don't know. Who knows? I guess it depends on who it's coming from. Um, but you know, I can kind of, when I think about yesterday, last night, you know, the Tuesdays are my long day and I had a very busy day, multiple podcasts, talk to people,
[00:50:47] um, building out rundowns for this show. Oh, and our Android faith. Like I was very, very jammed. And at the end of, of Android faithful, I was also working with chat GPT to create a process that would make some of my post-production work with Android faithful a little easier. And I didn't step away from here for like two hours after Android faithful ended. And I was fine with it. Cause I was really like in the mode of like, I want to figure out this system and everything like that.
[00:51:15] But in that case, like I would have walked away two hours prior, but I chose not to because I was so invested in building this tool, but maybe that's for your boss. So you only have you to blame. That's true. Yeah, totally. I don't, I don't blame the tool for that. Right. I blame my own inability to turn it off because I was inspired in the moment, you know? Yep. So at least there's that Lee. Thank you for the super chat. Matt Lee is watching the live stream on our YouTube channel.
[00:51:42] AI inside show says, and I think this ties in with the Google and schools story that we're just talking about says maybe I'm completely wrong on this, but Google has used machine learning tools in their search product for 25 years. Difference between machine learning tools and modern generative AI. I'd say there's probably a difference there, right? Like generative AI is a different ball of wax versus using machine.
[00:52:08] Learning behind the scenes on what are the right search results to, to appear and that sort of stuff. But I get where you're coming from, Lee. I don't know. What do you think about that? Um, same, same, but different. Yeah. I think, I think we're, we're there now where companies are trying to find it. I think the, the, the, uh, token maxing age has already ended. It's too expensive. Yes.
[00:52:32] Giving money to, uh, the big companies, um, the open weight models provide opportunities. I think the key will be how it mixes, how you do something with the open weight model, and then you have the expensive model, check it or check the security on it or whatever, um, or do the tasks that it couldn't otherwise do. Mm-hmm. Mm-hmm. So I, I think this is going to be the trend.
[00:52:54] And again, this goes to the anthropic and open AI, um, IPOs, because I think this is going to present huge use and price competition for them. They're integratable. That's not a word. You can integrate them. You're talking about the open models doing 90% of agentic work. Is that right? Is that what I'm looking at? Right. No, no. So, so Lee was talking about the Google. Oh, Lee's comment. Oh, I'm sorry.
[00:53:25] My apologies. But you've moved on. No, no, no. That's why I missed. Uh, yeah, I agree, Lee. I think, and I, I think you raise a good point, Lee, cause it's, sorry, now I'm lathering back. Um, is if Google had to warrant, we're giving students no AI, they kind of can't. Yeah. That's. Where's that line? It's behind the scenes. How do you define it? How deep do you go to defining it? Where do we get in trouble? So you don't want AI. Okay. You don't get AI. Yeah. Yeah.
[00:53:54] It's like, uh, um, street view in Germany where at some point Google just said, okay, fine. No street view for you. No soup for you. Mm. Yeah. Yep. That, that makes a lot of sense. Sorry. Thank you. That's okay. Okay. But you've set up our next story very nicely, which is you've, you've given the argument. Now we just need to know what it's, what it's based on.
[00:54:19] Um, Gartner has this new report on open models and, uh, essentially what it means or what they're, the headline of it is open weight models will run 90% of their enterprise agentic workflow steps, which basically is saying nine in 10 steps. An AI agent takes are actually very routine and classifying things, calling tools, checking data, all that kind of stuff doesn't actually necessarily always require some sort of frontier model working in the cloud to do it effectively.
[00:54:48] That open weight models can handle a lot of this at parody for a quarter of the price of the cheapest closed model. And for the enterprise, that's potentially a really big deal. And I basically, and basically it sounds like Gartner is saying, um, by 2030, 90% of those operational agents stepped, uh, run on open weight models and the remaining 10%, which is, you know, probably relegated to like R and D research, deeper research and development,
[00:55:17] that sort of stuff would happen with the frontier models. Also the harnesses, I think that are directing the agents from the open models would still have access to those frontier models for when they need them to do those certain things. I don't know how that changes the dynamics, the pricing dynamics of those frontier companies and what they offer. But, um, yeah, I totally agree with this. I think it, so I've got a question for you.
[00:55:41] Um, asking for a friend, my son, uh, you, you start to use Claude at some length. And then before you know it, you're buying more tokens cause you're in the middle of something and you hit your time window or your token window or whatever. And then before you know it, you have a huge bill. And so the other way that I know of right now to effectively run open weight models is to buy a very expensive $10,000 machine or two of them and do it locally.
[00:56:10] Are there hosted? Do you know of any ways that you can run open weights in a hosted environment? So you don't have to own the hardware. And so, yes, you pay, you pay a time or token usage fee on the open weight model, but you don't have to own the hardware. You know of anything like that? I don't. I think it's a heavy. I mean, if you go to hugging face, that exists. I would have to, if you go to hugging face, you can download anything, but if you don't
[00:56:39] have the hardware, you can't run them. So because that's assuming that you're running them locally, I mean, like when I think of llama, which was, it still is, I don't know, but was a meta's kind of open model, uh, architecture or whatever, you could still run llama in the cloud on search services, right? Right. Right. So I just wonder if it's, if it's a business yet.
[00:57:07] So I just searched on, this just came up this morning. That's why I'm asking now. Um, uh, featherless.ai sponsored result host, any open source, um, AI, no GPU management, uh, all done in the cloud. Yeah. I'd be really surprised if you couldn't do this. Um, I'm trying to think if I have any confirmed like services, like we had a previous sponsor
[00:57:31] Aria or area, sorry, area, uh, open models in the cloud. And I'm sure that they have, uh, I think obviously at a, at a corporate level, that's the case. It's what, it's what Palantir does to an extent. Mm. Right. Um, with their harness, but I would think you'd get to a point where there's just a, there's a, there's a, it's not a big business yet. Cause you have to be pretty geeky to do it. Mm.
[00:58:00] But if there's a geeky way, uh, to do it, I've not seen a lot of things popping up here. Yeah. It's a good question. I mean, I want to say that, yes, this exists, but I'm not entirely sure right in this moment, how to, how to kind of confirm that. Um, I just, you know, cause, uh, cause like, like I said, like, uh, llama was meta's version of this early on. Right.
[00:58:30] But you can download llama and you can use it, but I could interact with it on the web too, through certain services. Like I could find it. Oh, for those services. Yes, you could. Yes, yes, yes. And I, and so I imagine that's the case for the chunks and the, all the other things as well, you know? Yes. I, I'd be really surprised if you couldn't do that in one way, shape or form, but because not, because to your point, not everybody's going to buy them. Buy a spark. No. Or a $10,000 app or whatever.
[00:59:00] Or be able to get them. Yeah. Yeah. Right. Or even be able to get them at all. Yeah, that's true. All right. Well, if anybody folks out there know of these options, let me know. Yeah. Let us know. Contact at ainsight.show if you want to send us an email. Um, let's see here. What else? Jensen Wong. Oh, Jensen Wong. That's right. Let me find the story. So this was a story last week in the time. Yeah. What do you think about this one?
[00:59:30] This is a, this is kind of, you know, this is a little bit more kind of along the lines of like, Ooh, something going on behind the scenes. Do, does somebody know something we don't about Jensen Wong? And I don't know, it could be innocent. It could be whatever. But New York times reported last week that us intelligence officials, um, have basically concluded that an executive at a Singapore data center company called mega speed was a Chinese intelligence agent or asset.
[01:00:00] Her name, Alice Wong, no relation to Jensen Wong. But, um, they had surveillance that showed that she had a close relationship with Jensen Wong. Um, photograph together, blah, blah, blah at a restaurant, blah, blah, blah, blah. Um, but they were kind of thinking or inferring that, um, there might be some sort of like deal making behind the scenes to get around some of these, these China, us, um, roadblocks. Am I understanding this correctly?
[01:00:28] I thought it was a very flimsy story in the times. It sounded, yes. The more I, the more, the more I thought about the more I was like, all right, this, this might, yeah. And it's kind of dropped. I mean, here we have, um, uh, who was it? Uh, Tom's hardware is kind of picking it up, but I don't think we saw it picked up necessarily anywhere else. Yeah. The jury, the journal might've done a version of it. Um, first Jensen Wong is savvy as hell. He is Taiwanese, I believe, but he understands Chinese politics.
[01:00:57] He understands all the ins and outs of dealing with China. Um, so it's not like he's some innocent waif who comes along, uh, and, and wouldn't understand what's going on and all the risks involved. He's doing business. He was doing business with this woman. And, uh, uh, you know, I've had, I don't want to smirch, um, students, but I've had people say to me in universities that you have to be careful around Chinese students, not because
[01:01:25] they're spies, but because they could be spied upon. And that if you, if, if you induce them to say something that was controversial about the government, who knows who else is in the class who could spy on them. Right. So yeah, there's a, there's a suspicion then of many connections and Jensen Wong is plenty savvy. Number one.
[01:01:49] Number two, this whole has a relationship with tabloidy and super tabloidy. Yeah. It was, there was nothing to back it up. It was irresponsible. And then three, um, Nvidia said the next day, uh, there was an investigation of this relationship. It turned up nothing of this, of this woman. It turned up no evidence that, uh, the, of the chip is being sent or this or that. And there's nothing here. Yeah.
[01:02:15] So yeah, I think, I think this is a, this was an effort by the times to try to make, Oh, there's something going on. Yep. Yeah. Yeah. Yeah. And I think I fell for it, for it initially. And actually when I was putting the show together, I've, I started having mixed feelings about the story. Kind of that whole tab. I think it's worthwhile for us to maybe point out the flimsy nature of it. Yeah, for sure. For sure. It doesn't quite seem like there's a, there, there, but, um, nonetheless, there you go.
[01:02:45] Thank you for all that background. I think that's important. Uh, real quick. If you're enjoying this show, uh, make sure and leave us a review on Apple podcasts. It's a huge help helps get the word out, helps show that people are watching and listening, uh, including, uh, people who might someday want to sponsor the show itself. So, uh, thank you very much for that. We're going to take a break. And, uh, when we come back, we got a few quick items to get to before we get you out of here. So hang tight.
[01:03:15] All right. So this is a story, a quick kind of update, um, on the anthropic open AI meta Google, all of those companies having these AI models that are a scale. They're escaping, uh, their testing environments. And a lot of stories have happened in the, to this regard in the last six months. Well, there is a security firm behind a lot of those tests that is now basically taking ownership saying that it was an error on their part.
[01:03:44] And the firm is called irregular. It's out of Tel Aviv. It runs security tests for these companies. CEO, Dan LeHav at a Bloomberg event basically said the isolation guards were not set up as they were intended. So he's basically saying, look, this, this wasn't necessarily a matter of these agents being so wicked smart and crafty that they, they started doing all of these things on their own and everything.
[01:04:10] This was really about according to them, uh, humans that didn't set up a test environment properly, which is something that we've been saying on this show in recent weeks. Uh, you know, is, is that doesn't automatically mean that these things are so smart that now suddenly they're going to take over the world at the core. They're just looking for holes. They're looking for excessive, uh, you know, X, uh, ways to access and accomplish the tasks that they've been given.
[01:04:38] And if we don't do our due diligence as humans is setting up a proper environment for, to test within, then they're going to find the way out. And, uh, that, that responsibility lies in the hands of the human. So human error, not actually rogue AI. Do you feel any better? Um, yeah, I, I think this goes back to our discussion about regulation earlier. I think that there, there's not some magic organization that's going to solve all problems,
[01:05:05] but I do think that, that when there were the breakouts from open AI, the security firms that were called upon were questioned in relation to their independence of open AI and of test real boomer stuff, boomer stuff. So I think we're going to need a kind of an underwriters laboratories for AI testing such that there's an independence you can call upon and they need to be funded by the AI companies, but independent of them.
[01:05:35] One is it. So when radio started, advertisers wanted to know how large the audience was and the networks provided their own data to them. It says, there are larger audiences. And the advertisers said, we don't believe you. And so separate rating companies were set up Nielsen and such eventually. Right. Um, and others in the early rating days and broadcast, but also in print, uh, there was what used to be called the audit bureau of circulations that is now called the association of audited media.
[01:06:02] I was on the original digital committee for them deciding what a page view was and they were independent of the publishers, uh, and would, would then be trusted by the advertisers to audit. Even then some advertisers didn't trust it and they insisted on doing their own audits, but whatever. Um, so this whole question of who's at fault when it breaks out will define the environment, define breaking out, define who is reporting on this.
[01:06:32] There has to be more independence. I think that's going to be a pretty quick fallout even from the election is it's going to be something that the industry has to offer up is funding independent, um, auditing, uh, organizations. Good. Yeah. And that'll, that'll be a welcome change. Yep. But it'll be changeable as hell because you know, the clever AI will invent something new. We didn't, we didn't anticipate. We couldn't have possibly known that it would be so clever.
[01:07:02] Uh, we have open AI finally merging chat and work tabs. This is something that Anthropic actually merged back in September. It's kind of the converging of their, their standard mode and their work mode, which is where, you know, integrates with like files in your computer and computer control and cloud and all that kind of stuff, browser control and everything. Um, and so, uh, chat GPT is kind of bringing them all together. Codex will stay separated from that separate mode.
[01:07:32] Uh, but they will share a one usage pool with that mode. So, you know, that might mean that users will hit limits sooner because of the combined nature of that. But, uh, there you go. It's a small change, but. You know, one less thing for you as a user to have to ask yourself, should I be in this mode? Should I be in that mode? They just kind of all do the same thing for the most part now. So thank you for that. Yeah. I've definitely been using chat GPT a lot more.
[01:08:01] I've also been using a Gemini a lot more. I've been really expanding and using each of them and kind of pitting them against each other. Someday I'll, uh, talk about some of the strengths. I can't wait. I can't wait to see them. Um, Apple watch has this AI listening feature that we've talked about called live rewind. It's now in beta. So if you have the series 12 or ultra four, uh, and you're in the, the, the beta, the active beta, then you might have access to this.
[01:08:26] You double press on the crown, the digital crown, and you'll get text of the last 15 seconds of conversation that appears on your watch. So it's kind of collecting this in real time, uh, no speaker IDs. You don't get any raw audio saved. So there's that no verbatim transcript of the entire conversation. It's just the last 15 seconds. And when you do it, a chime rings. So the other person knows if the other person actually knows what a chime means, does a chime
[01:08:56] count as consent? Who knows? You know, it's not like when you're, uh, when you're on a call and someone hits record. And it's like recording the call so that you know exactly what's going on. It's just a chime. I think people are gonna be nervous about this. Number one, number two, lawyers are going to make hay on this because if you're in a two party state and both parties need to consent, the chime will probably not be seen as adequate.
[01:09:22] And, uh, I mean, what happens when, uh, you're on a zoom call and someone decides to record it, you now constantly have to hear the voice. This call is being recorded. Yep. Yep. Um, used to be back in the day, you probably never experienced this, Jason, but back in the day, uncle Jeff tells me that if a phone call was being recorded, uh, there was a beep you were supposed to present on the phone calls. People knew very similarly. Oh, so there's a precedent there. Yeah.
[01:09:51] Um, I don't know if a radio station for a reporter, uh, recorded you. Uh, I wonder if I can Google this. Um, a phone recording beep is an audible tone. A half second burst at 1400 Hertz every 15 seconds, every 15 seconds played during a telephone call to let all participants know the conversation is being recorded. Oh, okay. I feel like I've heard that before.
[01:10:22] Um, yeah. And, but I think again, just as you pointed out, did everybody know what that meant? Was this a beep in somebody's phone call? Okay. But there, there is a precedent for that. Okay. Okay. I wonder how loud it is too off that tiny little watch. Would it really be loud enough for the other person? Well, if you, if you, if you search for recorder warning tone on Wikipedia right now, you can play it. Oh boy. Do you want to do that recorder warning tone?
[01:10:52] Let's see here. This is going to require recorder warning. Recorder warning. All right. You've got your recorder. Warning. Recorder warning. We've got, we've got a feedback. It's not, yeah, it's not doing anything. I don't. Oh, there was the tone. Oh, I heard it. Okay. Oh, yeah. Oh, it's not. Okay. Now I can share. I can share it through the actual here. Let's see. Okay. Mute on your side.
[01:11:22] Okay. If we got to hear the tone. Oh, that's pleasant. It's a height. Did you hear that? Yes. Okay. Yeah. That's a pleasant audio. So I remember that tone from back in the day. Cause I also remember old fashioned phones, kids with wires. By the way, by the way, Jeff, I'm recording this podcast. You know this because the tone plays every 10 seconds for all of our listeners throughout the entirety of the podcast.
[01:11:52] Boy, that would not encourage people to keep subscribed. Oh, well. Uh, what else do we have here? Let's see here. We've got, no, we talked about that. Apple restricting dating apps and AI companions to users 18 and up. So starting November 16th, there's going to be a new, uh, app store age rating category called dating and relationships.
[01:12:17] It will cover apps connecting real people, of course, but also virtual companion apps. Um, you know, sexual and romantic content obviously included in that it'll get an automatic 18 plus rating. So this would be for, you know, the apps like replica and character.ai may it rest in peace. Um, they would, they would be included in this. So that makes sense. Yeah. That makes a lot of sense to me.
[01:12:45] Um, and finally, Anthropic confirmed it's steering its cloud code away from cloning games. So a lot of people to test the limits, you know, and, uh, kind of recreate certain games that they loved that, you know, I, I did this with a Commodore 64 game recently, and I only got so far before I got bored of the process, to be honest, but, um, but a lot of people are doing this and having a lot more fun than I did with it. Um, and also creating current IP or recreating current IP.
[01:13:14] And that's really the, I think the, the big point here. Um, Daniel says there, you can't see the webpage. Oh, sorry. I thought I put it on stage, but I did not. There we go. There we go. Um, yeah. Yeah. So an X user spotted, uh, in the prompt, it said, do not build a recreation of a specific existing game, even if it is renamed or called original.
[01:13:37] Um, so yeah, Anthropics, Addy, as money confirmed that they added two sentences to Opus 5.5 to prevent, uh, any creation or regeneration or recreation of existing games. So there you go. Cause people are vibe coding all sorts of things and reconstructing certain things. I hope this doesn't happen with the, um, with the, uh, the, uh, the, uh, the, the, the firmware stuff that we were talking about from a couple of weeks ago.
[01:14:05] Cause there's some real creative stuff happening there right now. I continue to follow that. And it's pretty, pretty interesting. All right. Well, that is that. I think we've hit the end of the episode. Jeff Jarvis.com for everything. Jeff is up to hot type web, we weave and all of his other books. Gutenberg parenthesis. So many, so many autograph copies sitting there waiting for you at the Mochler book center. If you follow the link at Jeff Jarvis.com. Yes. The link still works. Therefore there is still an auto autographed version of the book.
[01:14:35] I also autographed one copy of Gutenberg parenthesis there too. So. Oh, that's a, that's a deep cut. Yes. Right on. Jeff Jarvis is all new, newly redesigned website. Go there. Jeff Jarvis. Jake Jarvis. Uh, yes. Thank you, Jake. I don't know how many tokens he used to do this, but it's very nice. It looks very nice. It's a wonderful website.
[01:15:02] Uh, everything you might need and a little bit more is already there for you. Uh, pod tune up.com. If you want to talk with me about your podcast, um, strategy for building your podcast strategy for, uh, you know, turning your podcast into a money making thing. All of these things. Um, yeah, I got, I actually got a new client earlier this week and I just love it because it's so much fun to like talk to people about their podcasts. I absolutely love it.
[01:15:30] It's one of the great things about watching you as a consultant is you really enjoy working with these people. Yeah. Enjoying their problems and trying to figure them out together and bringing your knowledge to bear. It's fun to watch you do that. It's helpful for me too, because like I've been doing podcasts the way I've been doing it for 20 plus years. It's cool to talk to people and see how they're doing it. And it's illuminating to me because I'm like, Oh, Oh, I hadn't really thought about that aspect. So I feel like I'm learning a lot more about the breadth of, of the many different ways
[01:16:00] that you can approach this business of podcasting. And that just gives me more knowledge to share with others. So anyways, pod tune up.com hit me up, uh, AI inside.show. Yes, we do have a new webpage. I'm also very, very happy with this webpage and the way it, uh, the way it turned out. I will talk about this in a future episode, I promise, but, um, you're going to have to wait for that. Uh, AI inside.show for all the information you need about this show ways to subscribe. Everything you need is there.
[01:16:29] And my, uh, my voice is, is leaving me. So I'll make it quick. Patreon.com slash AI inside show. If you want to support us on a deeper level, you can, we have some amazing executive producers supporting us very deeply. We appreciate you. Dr. Do Jeffrey Maricini, radio Asheville, one of 3.7 Dante St. James Bono. Derrick Jason, Knifer, Jason Brady, Anthony downs, Mark Starcher, Karsten Samotsky, and Evan museal. Uh, we thank each and every one of you.
[01:16:58] You don't have to support us at a, at the executive producer level, but if you do, you get some bonuses like that and a little bit lower than that, you know, you get your AI inside daily with all of my awkward frozen thumbnails, uh, of me in the midst of talking and all that, uh, discord access ad free episodes of this podcast and so much more. Patreon.com slash AI inside show. All right, Jeff. I think that's it. Uh, also I don't want to forget to thank, uh, Daniel Croft and Anthony Lemos behind the
[01:17:28] scenes who help us on the video side of things. Couldn't do it without you both. So thank you. And thank you, Jeff. Appreciate hanging out with you and we'll see y'all on another episode of AI inside. Take care. Bye.
[01:17:40] The Wired newsroom is known for award-winning reporting on how technology shapes our world on Wired's uncanny valley. We take that curiosity even further each week. Journalists from Wired break down the biggest stories in tech while speaking directly with the people building challenging and reshaping the future.
[01:18:09] Is the AI boom sustainable? How do you protect your privacy in an age of constant surveillance? Uncanny Valley tackles the questions driving today's tech debates and lighting up your group chats. Listen to new episodes every Thursday, wherever you get your podcasts.