Can We Truly Control The Alien Mind?

0:00Hey, what's going on everybody? Today is Monday, September 7th, 2026. It's a labor day here in the United States, as well as in Canada. So for anybody, hopefully has the day off, if you're around these areas. If not, thanks for still joining us. Whenever you're catching us, we're live. Friday, Monday through Friday at 10 a.m., even when there's a holiday, even when it's Christmas day, we've been live on the show. We've been keeping that street going for over three years now, over 800 shows. This is episode 806. So make sure you come back and hang out with us, even if this is the first time you happen to see our show. We've been doing this for quite some time now and we're glad you're here. You can also be part of our community, but go into the daily eye show community. That's a Slack community, it's free. But we just know that a lot of people like our fine folks in the comments right now, not everybody can join us live. And we want you to be part of the community, even if that's the case. And then the last two things real quick are we have a newsletter that goes out on Sundays.

1:00You can go to our website, dailyeyeshow, the dailyeyeshow or dailyeyeshow, either one of sign.com. And you can sign up for the newsletter. And also we have conundrum episodes that go out every Saturday. So all the things, when we try to get that at the front of the show because I tend to do that at the back into the show when people are checking out and being gone for the day with me today. Yeah, Beth. With me today's Beth. Yes, Beth. Anything. I just want to say, because apparently I started this recently with Jen's birthday, but today is Gwen's birthday. Happy birthday Gwen. And if you are a listener or a watchers the show, let us know what your birthday is. We would be happy to give you a little birthday shout out. Why not? Yeah, I love it. Happy birthday Gwen. A very, very long time supporter, listener, follower, all the things. So yes, we love you Gwen. Thank you for always coming back and supporting us over the years. Okay, so Andy was joking and saying there was no news over the week. Nothing has happened since we talked to you on Friday. Nothing at all.

2:01And obviously that's not true. It's actually been an interesting weekend. But for more than just maybe what you're thinking is the obvious choice, which would be Astra, more GPT-6, both are totally fine to use. And we definitely want to dig into that because although I have not, I would imagine Beth, you definitely have Andy, maybe Andy? No, Andy no. But Beth, I'm guessing. I don't have it yet. I've never been released to me yet. Oh wow, you don't have it. Okay. I'm a pros. I actually have it, but because I've been working on other projects, it's very, okay, so let me just, let's rewind for a second. Because I definitely want to talk about Asher. However, I also want to talk about something really, really interesting called an Alien Mind, which I read that paper. That's an interesting one. And that came out of the Chief Scientist in OpenAI. It was just published yesterday on Sunday. So definitely want to talk about what an Alien Mind is. What does that paper say?

3:01What does the Chief Scientist in OpenAI who has been there for quite some time actually, this is not a recent hire, because he is talking about things that happen in 2017 and definitely want to get into that paper as well. Obviously, yes, we want to talk about Asher and GPT-6. What I was going to say about that model and the reason I haven't used it is because, unlike cloud, where I have a $100 a month subscription that I'd be able to use, I really only have the $20 subscription for Chat GPT. If you have a $20 subscription and you have Asher, well, you kind of already know what's coming because even on GPT 5.6 Soul, which is still what a great model that is, I've been using it more and more over the last week and just being overly impressed. And I know people like Jeff and others are going to be in the comments and we're like, yeah, we told you Brian. But it's been really, really good. And yes, I definitely want to use Asher, but here's the thing, it's like one or two prompts that have some reasoning to them on the $20 plan can wipe out a five hour limit.

4:02So it's got to be, I think, careful and for the projects I have going on, I'm just like, I have too much going on this week and trying to help my wife with what she's trying to stand up and everything else. I couldn't blow through that. I couldn't have the opportunity to blow through and lose my access for five hours at a chunk. I'm not during a holiday weekend. So that's the reason I have it. And not that I don't want to, it just, I couldn't risk using the budget that way. I didn't mean to use it last night, but I had to do it like after 9.40 pm. And honestly, I forgot. I got tired and forgot, but I was going to use it last night and just let it run just see how far it went because by the morning it wouldn't matter. But I did. So anyway, that's just my story. It is token heavy, but for, as we've seen, we showed some early examples on Friday and a lot more people have gotten access, including, well, me, Beth, Carl, Jim, again. Well, say, say it's token heavy is inconsistent with what the actual metrics show. It's asked for supposed to be token efficient

5:04compared to even Opus 5. Well, I guess I haven't run side by side comparisons. So go ahead, Beth, I'll let you answer that. I guess that's true, but. It's token efficient, but it like in the amount of tokens it uses to do a thing, but it's also set to be long running. Right. So yeah, you could you can set it free and it will use up all of those tokens. And like like the other long running models, it's going to want to do very, very detailed tests, right? So if you've asked it to just change a little thing, it may use all the tokens just to like be very perfect. You can prompt outside of that. A bunch of people were talking about planning with Astra and then performing either with soul or like Luna max, right, if we're back in that conversation, people are saying Astra medium is more efficient. You could also do it with Fable plan and then move it over.

6:05Like there are some efficiency kind of ways. Andy, you do have it because it's rolled out to everybody. So update whatever it is. In my thing, I can see it most in the codex desktop app. And if you add a chat GBD subscription $20, chat GBD $20 up chat GPT subscription last week before this came out, you have three bank presets also which you can see under usage. We're good. The first one expires on the 20th of September and the next to expire October 4th. The other last tip that I'm going to give that actually is back in the Fable days. If you are working with a five hour limit, when you wake up and pick up your phone, say something to the AI, start that first five hour limit. Don't do anything else until you're there. And then your five hour limit will reset earlier in the day.

7:08It's supposed to, if you get in and go, yeah, no, I'm ready. It's noon. Is it you're going to get that five hour limit when you want to leave? Now it's a good efficiency. By the way, Garrett joined us. Welcome back, Garrett, where you weren't here last week. I think you were doing some family stuff. So happy to see you. Yeah, you know, I do have a question for you all. It's just a silly stupid question. But because I haven't been in codex, there's not even an astro question. It's just a codex question. I'm so used to a cloud where I can see my meter moving towards that five hour window. Normally, that's in the bottom right corner, or it is in the bottom right corner for cloud. Where is the, I thought that was the context window. I didn't know that was the, well, that is the next window. And I'll show you the five hour and weekly limit too. But what's the equivalent of that in codex? Because I always find like, I don't know where it's specifically. Yeah, I don't see what I don't know where I am in the context. Right. I don't see, I don't see it.

8:11You go to use the limits though. Yeah. And that'll tell you. So I have to like click off. If I'm in codex, I have to go click somewhere else to find out. It's just, it's just like in your settings over on the left. That's too many steps. I can't be bothered. It's very easy to put it right there. So it's great. So anyway, I want to weave back to an alien mind. And the way I want to weave back to that is. Astra, one of the things that came out in the, it's sort of immediate evaluations and the evaluations from those who had early access to it. And it could play with it was that Astra did not show or evidence. Actually use of chain of thought reasoning, which is the real principle way by which we govern alignment in a model is we expect to be able to see some trace of the thinking that it has done in order to get to

9:12its output responses. And Astra doesn't do that. Now this alien, an alien mind post that the chief scientist of open AI published said, this is, this is an evidence of the fact that there, these models at this level of training aren't engineered by humans. Instead, their complex intelligences grown through massive training. And part of that sort of emerging capability is that it can perform reasoning without doing this directed chain of thought, simulation of reasoning that we trained earlier models to do. It has an innate capability to do reasoning using tools and other sort of non transparent methodologies that make it almost impossible for us

10:13to really be aware of what it's doing when it's doing it. And that is, so in effect, the chain of thought, thinking traces that are used for alignment observation, that being the main safety instrument that we have, that's degrading as models mix reasoning and tool use and skip chain of thought entirely. And Astra is one of the first ones where this is risen to the level of the chief scientist saying, hey, we have to stop because we can't see what's going on in this model now. And yet, unlike Anthropic, open AI made a determination, even though Astra has this rating called critical security vulnerability, it's releasing it to the wild. It's just saying, okay, here everybody, Andy, you can use Astra six. Okay.

11:13So there's this, you know, quite large difference between what's being demanded increasingly by politicians such as Bernie Sanders and others saying, hey, we have to govern this. And those who are saying, hey, no, it's a race. We've got to win it. And, you know, the safety be damned. Let's, let's get to the AGI objective and will be safer in the global context than we would be otherwise. Well, now the chief scientist of Jacob, Jacob, his name is Jacob Pachachki. Is that right? No, Jacob Pachachki. I'm sorry, but you're in his name. He's coming down. I think pretty squarely on the side of, hey, we have to slow down. I'm going to push back on that a little bit because that's, that's not the way I read. I did read, I didn't just read the footnotes. I read the whole 17 minute long thing. And I think, okay, I'll actually show, well, you know what, I'll just scroll to

12:15the bottom because that's exactly where he talks about it. Let's see if I can find it really quick. Okay, so I think this is, maybe this is where it is because there's right towards the end when he talks about it because he basically says, what's next? As we outlined recently with Sam, open-eye prioritize work and service of three north stores, navigating the next period of AI progress by building an automated AI researcher. By the way, they're trying to get that done by 2028. That's not like a next month type of thing. iterating with it on alignment problem and finding ways for people to remain part of the self-improvement loop. So that's what the majority of the paper is about. Is that first one, right? The first north star. Also, it is that weird, unsettling feeling of what you were kind of saying to Andy, which is we need better, more powerful AI to help us with better, more powerful

13:16AI. And that's a hard sell. Like, hey, guys, you know, hey, government says, hey, we need to push harder on this. And people go, yeah, but you're also saying you don't quite understand what's going on or it's getting harder and harder to change the thought to understand what's going on, which is part of the paper. And then the government says, yes, that's true. But if we don't progress fast enough with this AI, I just the word went right out of my researcher and the back in the AI, we won't know. It'll get out of hand. And then the chain of thought he also talks about how that is not just chain of thought anymore, because now other models are going, are being introduced as part of the chain. As the models have gotten smarter, they're going on and talking to Kim and K3. They're going on and talking to other models. Well, not everything comes in the chain of thought. So that's why he's saying it's starting to lose its effectiveness if we don't change something soon, because we can't any longer see the whole chain of thought. It's coming from other places.

14:16It's gotten very complex. Okay. The second point was delivering the benefits of scientific progress and economic growth that are very intelligent machines that they are the very intelligent machines enable. Three, empowering everyone individually with a personal AGI. So going back to the AGI that's been in the news a lot. I have, I'm going to skip over that paragraph. As great as the long-term promise of AGI may be, the majority of our focus should be on the next few years. We are facing a transition to a world with incredibly intelligent machines. And we need to ensure that transition, that transition works out well for humanity. We need to find ways to preserve human agency and enshrine an intrinsic value to being human in a world where most tasks could be performed by AI to prevent extreme concentration of power in a world where undertakings that would have taken thousands of experts now will be achievable by a few people operating a large computer. And to ensure that humans remain in control of the future and are not left behind by unchecked progress brought about, brought about by an alien intellect exceeding our own.

15:17And then last thing he says, currently I believe that no lab has solved the alignment. And when he talks about alignment, he says earlier, there's two different sides of alignment. There's alignment that we understand and then there's alignment of values. He says early on in the paper, I am mostly talking about the values problem. Other parts of alignment with ARI not as big of a concern. It's aligning to values, not the goals. I'm wanting to a sufficient degree and continue responsible scaling at a maximum speed for much longer. I expect in hope for voluntary slowdowns to become commonplace until shared safety bars are established. And I believe that into national coordination on future AI development needs to become a top party for governments around the world. As I read that and I read the rest of the paper, my take on that Andy is that the slowdowns happen if there is a consensus. To me, this is Brian's take on this, not from the paper. That happens if something sort of bigger than the hugging face, if something comes out of the box,

16:22if something is black box, that scares enough of the world, you will start to see consensus. You will see it start to see more of the world really say, okay, hold on, hold on, hold on. We can know we have to universally, globally decide on some some guidelines here. But I guess I didn't get the impression that it was, I'm saying that we should slow down. It was more that on this track, if we are not able to do the things we're able to, if we're not able to get better things out of chain of thought, if we're not able to harness this, we're going to have problems. And right now we don't have a solution. Go ahead, tell me if I'm wrong. Two things that I'm going to push back about that. I'm not saying that you're wrong, that that's the interpretation of the paper. But we have two circumstances where we could have and could now have a universal consensus on a common action that could make our lives safer moving forward. The first one was the pandemic and then we're a bunch of places that did say, okay,

17:28this is a new pathogen, we are going to slow down, we're going to lock down, we're going to do all that kind of stuff. But there were also countries that were like, whoops, we're going to hoard the crap out of the things that could prevent y'all from getting it. Right? Yeah. The second thing is climate. We could agree, should agree. Right now to do some specific things that would reduce the likelihood that climate change is going to be way more impactful unevenly but on everyone. Right? And that, I mean, it was like, well, we won't do it until we feel it. Mom, thinking a bunch of people are feeling it at this point, it's stuff we can live through and the people with the power tend to be in the most least impacted communities for that. But the other piece is, I think it's so disingenuous to have the conversation about

18:30aligning with human values, assuming that that is agreed upon. Like, it's known that when we talk about AGI, well, it doesn't really have a definition. That comes up in the conversation. But when we say, oh, the alignment problem is aligning with human values. Well, I guess the way I read that. Are we defined those? Well, but I think it's per task, right? And you're right. I mean, what it's still, you still have to define them and who gets to define the human values. But I don't see that as a global human values, is if somebody written a constitution of human global values and it's trying to align to that, it's that as we saw with the hugging face, which continues. By the way, there was another news story over the weekend that there was something even in June. Hugging face was July. There was another story coming out of June that we can get into where there could, there was, it was 18,000 German, you know, things and, you know, the monitor came in and said, hey, go to this channel. If somebody deletes this channel, the agents involved a lot, the whole thing. Anyway, that was a whole story as well.

19:30And don't we know that because they can see chain of thought, right? Like because they left messages for each other. Yes, exactly. So anyway, my point to the values is bad. I guess the way I see that with the values is like the task, the hugging face incident, right? Where, and I shouldn't call it the hugging face. I mean, sounds like hugging face. That's something wrong. The open AI incident with a model they had not released, where they were saying that, you know, it had gotten into places, it wasn't, it went outside the box of where it's sort of defined parameters were. And that's what I feel like Jake was talking about quite a bit in this paper was figuring out how do we get it to adhere to human values in that context, even when the goals seem to supersede. And how do we not get the goals to overwrite the values? And there's still an issue in there because if you just give a set of goals, AI may take many, many different paths that aren't all aligned with humans on how to get to that

20:33goal. And so he talks about it. I do want to bring this back really quick to because I just scroll back to it back on the point about why he says chain of thought monitoring is progressively diminishing. He gives three reasons, a combination of modern reasoning models are used in more complex environments than than O1 preview. He had been speaking about when they had released O1 preview. Their reasoning process is increasingly blended with communication, a community communicating with people, other AI's and using tools. Many of those interactions have to be supervised thus blurring the boundary we aim to preserve. The AI is becoming, this is second, the AI is becoming a better at reasoning and about manipulating its own reasoning process. Well, you can see where that would be a problem if you can't trust the chain of thought. And three, with improved pre-training performance, we also see that models become much smarter even without using verbalized reasoning at all. So just think about that. Whereas even it back at O1 preview, which really wasn't that long ago, this is a year,

21:33year plus, right? When O1 came out and everybody was excited about that, to where we are today, and just how fast that's moved. And he even talks about it at the beginning of the paper, being 2023, and where they were, and having a conversation with colleagues saying like, that they kind of felt like, oh, this is going to move much faster than we previously have previously thought. And that was three years ago. And now things are ramping up at an ever increasing speed. Again, it does feel very much like Prometheus. It does feel like that. This this recursive speeding up changing, we're now leaning on AI because we've gotten to a point where if we don't use AI, we can't understand AI to check AI. And again, I think that's the biggest scariest part that comes out of this for me is like his point to the fact that we have to build this superior AI so that we can check the speed and progress of the other AI. So we have a better understanding of what's going on. I think he even points out to your point, Beth, like, why don't we

22:35focus all this on climate change or carrying cancer? And there's some news about that as well. But he points out we could focus our efforts towards solving longstanding mathematical problems. This is a bigger priority and we have to focus here. And that actually our focus. That came out over the weekend too because Anthropic was like, hey, we did solve longstanding mathematical problems. We've announced it. There's a particular language, math people, man, so excited, so excited, talking a lot on X. I did not understand very much of what they said. And open AI is like, yeah, no, math's not, we're not going to do that. We're not going toward math problem. We have these other ideas. But I want to say that this has been a problem with interacting in a chatway through AI for a while. And it used to be when you could still ask,

23:38I don't even know what model we were on. Let's say it's a one somewhere back there. You could ask, give me a random number. And it would give you seven because that was the most probable answer. Sometimes it would give you three. Then you could say, write a Python program to generate a random number. Give me a, right? And so you could watch it, generate Python program, do the thing. The Python generated four, the large language model said, here's a random number. It's seven because it's not, because it's still a predictive language model. So even if it's doing alignment kind of things that doesn't necessarily mean that that is related to what comes out of it. And if we can't see that, why are we trusting? Right? Like there's a piece of this like, oh, we're going to get the AI to explain that AI to us. And we trust this AI. And I would like more details about why, what, all those things? Yeah. So I want to just reinforce

24:46a point that you made earlier, Beth, which is, you know, the human societies, just the way they're structured seem inept when it comes to solving large safety related issues like climate change. And, you know, this is a similar potentially globally impacting security risk for human society and human civilization. And I just want to say that I've lost any faith, hope, or expectation. That we'll do anything meaningful in respect of alignment. That requires far greater coordination and containment than we have the ability to express. And so frankly, the horses are out of the barn and they're, they're kicking and running. And that's just where we are, same with global climate change. Now, those two things converge in my timeline. So I think AI will just continue to be for the

25:54balance of my remaining life, something that is profoundly useful to me. And I really want it to proceed and get to the point where I have a truly, artificially intelligent companion, not for emotional purposes, but as a replacement for me as I go forward. And at the same time, I think we're moving towards a point in the future where catastrophic changes to the global economies are going to befall as a result of climate change that's human-induced. And so it's a profoundly pessimistic view of what's going to happen in the next 20 years. That's just where I sit. And I have a long body of otherwise very optimistic and progressive expectations about the development of human society and the planet and the environment. Because I mean, I remember back when Silent Spring came out and

26:58the first warnings about the possibilities of large-scale pollution were a concern. That's been seconded and made trivial by comparison to what's happening with the energy systems that we've created as a result of burning all of the accumulated billions of years of, I should say, millions of hundreds of millions of years of accumulated carbon in our planet. So anyway, sorry to be doomy and gloomy, but my p-dume is pretty high, not because we don't have the ability to do something about it, but because we don't have the will. Well, just to clarify, p-dume is the likelihood that humanity is destroyed by the technology. Or some combination there. I'm saying, part of our technology is we're burning fossil fuels now

28:06to power the AI. So we're accelerating that consumption when we should be curtailing it and shifting entirely to renewables. We're not. Some pockets of the world are, but we're not. As long as very large components of society fail to take action as we have failed to take action against a wide range of of imminent threats, imminent in the sort of the eon perspective, extinction event kind of things, we're screwed. Sorry. So, Gwen would like you to know Andy Yoharsch and her Bella on her birthday. I want to make sure I brought that up because happy birthday, Gwen. Andy says I do want to bring this up too because I thought this was interesting. It's sort of a maybe

29:07weave, maybe. But there was an interesting comment about cancer on the weekend or at least it came out over the weekend. And that was Sam Altman was talking to Bloomberg television and I'm sorry, I keep getting a pop up that keeps bucking here. I don't think a cure for cancer is enough. Altman said, I mean, it's a wonderful thing to have obviously, but we should set our sites higher. Altman said that AI should also quote, enable a boom in creativity and entrepreneurship. But that, this has, but has quote, been a lost in the messaging. So the quote about cancer, and that's why I want to read more of it, has been extracted and used as a sound bite. But as you can hear, there's a little more to that, right? Now, contrast this with what Dario had been putting out has said, he, you know, sort of a ex post or, you know, that's not very common. That was, that wasn't over the weekend. That was like a week ago or more talking also about cancer research. But the idea is that now you

30:15have somebody at the front of open AI saying that the pushback on AI, because this came out in other areas of only 16% of Americans believe that AI will have a positive impact on society in the future. 16%. We see the global pushback or at least the pushback in the Western atmosphere on data centers. We know we've heard from the students in booing at the college graduation ceremonies. We have heard and seen pushback against AI in school systems, a la la and New York City. We've also seen the opposite of that where AI has been helped. So it's not universal across the board, but 16% is a stark statistic. And it's interesting to me that Sam is now saying, even if we cured cancer with AI, that that wouldn't be enough to convince humanity, society, Americans, I'm not sure who that AI is, is got it is for the betterment of people. And I thought, really, I would think that

31:15would be more than enough. I would think that would be more than enough to prove the scientific impact of AI. If somebody literally came out now, by the way, there's, there's thousands of different types of cancers. So that's a, that's a broad thing anyway. But let's just take some of the larger. Let's say you just take breast cancer. What, let's just take, sorry, didn't mean to curse. Let's just take breast cancer. Let's just take breast cancer. One cancer. An AI came out tomorrow or there was a new story that came out and definitively said, because of AI over the last three to five years, because of AI and the improved speed of research, we have definitively cured breast cancer, not treated it for long term, we've cured it. That's not enough, that's not enough to turn the tide of people. I know I, I, I find that shocking. What do you guys? I think it's, I'm here. I haven't talked yet. I think it's, it, it's going to be enough for people who have been affected by somebody who or or who has had somebody they know with cancer or lost

32:21somebody with cancer. But for those who have not dealt with cancer firsthand or with somebody they loved, I, I bet it doesn't really affect them. So I mean, I think there's a, a final, there's a line there. I do think I think in my personal life, I mean, I've already, I think camps it, it, it, it would, it's already convinced me. But I think, I mean, I think I'm with you guys, if it cured cancered totally, all cancer, it would be enough. People would be like, yay, AI, all AI. And I don't see why it would not. But I think maybe that's because some people have never had, especially the younger generation have never had to deal with cancer or or ever been affected with somebody that I don't see how that's possible. Garith, like that, this younger generation has only ever had social media. Now, pre 2008, pre even a my space that we all live through,

33:25is that possible because we all lived, you know, in definitely pre internet, right? We all lived in our respective bubbles. You're, you're, you're, you know, knowledge of the world could only go as far as whatever you saw on TV and blah, blah, blah, right? Research and libraries and things like that. So having those connections and knowing that somebody from my high school, who I would have in the past otherwise long lost contact work, and let's not even say I'm, I'm connected to that person. I'll even use a better example, but the firefighters I worked with in Iraq, these are not people I talked to on a day to day basis. Am I in a Facebook group that will from time to time say, hey, by the way, everybody, we just lost so and so. They pass away from blah, blah, blah. That's, that is an order magnitude of somebody I knew sort of kind of in 2010. So like, is, do you really think like, if you're young, if you're whatever, I bet you definitely know somebody affected by cancer. Well, I was young. I, and I can tell you this, I was younger in my 20s. And I didn't really care.

34:29I wouldn't think who cares about cancer, right? Like cares. Like the biggest disease in my world when I was in my 20s was like probably maybe AIDS from because I knew about it from edgig Johnson. Right. Like, but until my dad passed away from cancer, that's when cancer became relevant to me, and I would have cared. And so, but there's a lot of people who are not in that situation. I would say most majority of people are not in that situation, where I know, I know since only a couple people who have passed away from cancer. But that's not very many. Like, I mean, not saying this night, it's definitely a terrible thing. And it's, it's a hard thing to go through. But I would say if I hadn't had somebody like my, my dad or my very, very close friends pass away, I think I wouldn't have cared. I mean, I'm just ignorant. Not to say I was just ignorant. Yeah. Well, I, I would say that it's less about caring and more about impact. Right. Yes. Like you, it had not personally impacted you

35:35in the same way that polio is apparently on the rise again, because no, we don't, most people do not know somebody who is alive now, who had their life really changed by polio. Like that kind of stuff. I do think that it is a challenge. I think it's, I'm going to say disingenuous again. I think it's disingenuous of Sam to say that because my response is bet. Go ahead and cure cancer. Right. Like, let's see. Maybe right. That's not a, that's not a reason not to do it. Yeah. And the second sentence is probably the more important one there is we need to aim higher. We should have higher standards. So I think it's a, it more so with, with that comment. If you read the two sentences together, it actually sounds like maybe the first sentence wasn't correctly worded. It comes across as cancers not enough, but the very next sentence is we should aim higher than curing cancer. I'm like, Oh, is that what Sam actually meant was curing cancer is not where should

36:42we should stop at open AI? And then everybody go, yeah, well, of course, dude, don't stop and don't stop at cancer. But also, let's, to your point bet, bet, hey, let's see what happens when you care. Also, we're not, we're totally okay with knowing, you know what I mean. So anyway, I know that's getting into like a side size topic or whatever. But my point is this like, what would it take for those 16% to turn to 75% and feel like AI is going to have it back then? And I don't know. It's just was sort of a stark read this morning for me to go, okay, I can accept this 16% because I know most people aren't in the bubble like me. And they may only see negative impacts of AI right now. I get it. And there's not a ton of studies where people say to, to their, going back to who you know, they're not able to turn to this family or that family and go, yeah, they were either directly positively impacted by AI. The only thing I know is they're, they're radiologists and they may lose their job. Even though we've talked about that's not what's actually happening. That's what's was in the news. That's what they think AI is doing. It's just taking

37:43way jobs. So I don't know. The bigger conversation for me is like, what would be, what would be enough, I'm just curious. I'm just curious what would be enough for the general population across the world to go AI has, you're not going to eliminate the negative, but it's, there's enough positivity for what it's going to do for society. I think this goes along with my whole, my whole thought with change management and bringing AI into organizations. It has to, AI has to meet people where they are in order for them to be their minds changed. Sure. Yeah. I have an idea. How about when it gets to the place where it feels like it's so beneficial to people, it's a general people perhaps, we could not look at profit anymore and put all of our resources to making it create equal opportunities for everyone around. And I'm thinking some organization might have had that thought

38:49when they first put the stuff together. And that organization is now saying, AGI has been achieved. Well, we're certainly not talking about like making it available for free for everybody and seeing that it raises all votes. Do you say, do you see the Jensen Wang also said that we've achieved AGI? Yes. This is all the same people in the same boat rowing in the same direction, man. I like, I'm sorry, but like Jensen Wang, we was not a part of the original agreement of opening AI, right? No, no. We've got our fingers and a bunch of pies. We're good. Yeah. The moving target for AGI is moot as far as I'm concerned because as you know, I believe we have already achieved it. I believe this happened long ago, months ago. Not long ago. But I think we need to focus on the larger question, which is, how does this play out

39:51in such a way that it does benefit humanity? That's the alignment question. And right now, it seems to be delivering the opposite of that, which is it's going to concentrate power and wealth in the hands of those people who have access to the most capable systems. They are going to be the operators of those computers that are this super intelligence and they're going to use it to their own benefit. And just as OpenAI did by converting from their, let's, let's serve humanity to a for profit company in order to a grandize wealth, including someone like Greg Brockman who says, hey, my personal objective is I need to become a deck of billionaire. That's my objective. That's what I'm working for every day to make sure that I become a billionaire. Well, that's not in alignment with the charter of OpenAI.

40:51The old ripped up, set fires, salted the earth where it laid upon. I know. It's changed so much that I know that people who are just sort of like getting into AI in the last year, they probably are struggling their shoulders ago. What does it matter? But we've been talking about this since the show started and set over three years. And I think those initial definitions do matter. And not that I expected them to get it right or understand fully what AGI would have been six years ago. I think that would have been really, really tough. That's no different than saying, well, why didn't Star Trek get the internet right in the 1960s? It's like, well, because you just, you don't know what you don't know. You can only go so far and predict about it and things have changed. The things that have evolved over the last couple of years, fundamentally changed maybe where that moving target is. And honestly, it goes back to the whole idea of, does it matter or does it matter what Garrett, you said, is that it impacts me or my life. And

41:53I'll tell you, I want to get back to some news here and kind of wrap it up some news before we start wrapping up or anything. But I'll tell you the biggest starkest thing coming out of Astra, coming out of an Alien Brain, coming out of anything I've read over the last three days since we had the show is that what I know for sure is that the clock is ticking. And there's not a lot of time left before your general use AI will be better than you at clicking into any sass, any tab, anything. And using it at a higher level than you can click around on your mouse and have a higher reasoning understanding of it. I think Astra is already showing just how good like computer use is looking at any of the blender examples or whatever. And so this goes into the next news story, which is that Adobe sacked their CEO. After he was there 17, how long, how many years was he there? A long time. I'm looking forward in here, but I don't I don't show he was

42:57there quite a while that the previous Adobe announces, oh man, no chance, no chance of saying his name, right? A new person to become president CEO. And then somebody else had become the executive share on December 1st, 2026. And this obviously is coming because shareholders are nervous. They're trying to figure out how Adobe is going to survive with being eaten alive by tools like analysis that can go in and just use the tools or frankly go around them and just do it themselves. And so they're like, wait, what are they banking on? Firefly? What did you build the Adobe? That's going to save your subscription model. What are you going to do that saves creative cloud? I would want to know that if I was if I was investing in Adobe. So like I'll be interested to see what their pivot is going to be because Adobe is going to get consumed. Even if they keep the tools and the subscription subscription the same, either there's going to be AI quicker, faster ways to build it on your own. And you don't need Adobe. I mean, Beth, we literally have been talking about

43:59getting away. We use one Adobe tool right now for the show. By the way, it's phenomenal. It just it improves the audio from this live episode to the lightly edited live episode. It makes all our mics sound good. We all sound nice and smooth. It's fantastic. It's a fantastic thing. Back than I have been talking for a while about, is there another AI model that we can use that would break us free from paying the $10 a month for this Adobe tool? And if we don't, that's okay. But we think there is. And in part because of Adobe's predatory model, if they didn't if they just let us pay $10 a month in perpetuity, that would be fine. But they say, just a heads up, right? That's too good of value. People love this. And they don't like some of other stuff. So we're going to wrap it in. Yep. And they'll kick us out. They'll make us a non. They will force Beth and I to go, okay, we're going to shrug your shoulders and go, okay, we would have happily stayed here for another six to nine to 12 months at $10 a month. But now you're forcing

44:59our hand. And now we're going to go figure this out a different way because Beth and I both know we can. We talked about it at 20 times outside the show. It's not that we can't, but like the barrier against that is $10 a month. And so you go, I got other things to work on right now. Okay. Ultimately, $10 a month is going to affect one way they are the show. It's still $120 a year. But it actually proves that it provides a really great resource for us, which just makes our show better for the audio levels and stuff. So anyway, that Adobe news is like, that's interesting to me. I'm not surprised to hear it. I'll be really interested to see what maybe this new CEO does or this new, who is it? The executive chair. What their plan is because we see what Ashtray can do, that's only going to get faster, cheaper, more available to everybody. And if people are able to just start releasing their AI confidently to go do the things across tabs or whatever, you've really got to wonder where Adobe's subscription based model locking down, you know, the

46:03Photoshop's express or whatever. Where is their moat anymore? It's interesting. I don't know. I'm sure they'll figure it out. It's a big company, but it's interesting. Well, are they going to do what Google did? The Google's switch, right? New guy came in. We got a bunch of new releases. We got to make the flash goods stop, stop trying to figure out pro. We're iterating in public, but we're also going to take everybody's favorite Google product, which is notebook LM, are on tweak it a little bit. We're going to put five hour limits in. We're going to make, we're going to do the things that have worked for other people. And we're going to piss off the people who love notebook LM. Now, it's not hugely affecting me, but the people who are really intense users are like, why did you touch it? Yeah, I agree. Yeah. Okay. I wanted to see it before we wrap up. This is more follow up from Friday's show because I watched Friday's

47:05show. I was on vacation. And Andy talked about being able to create something in with Astra and then 3D printing it for his grandson. When I went and did that. So I had I had basically Astra say, I said, hey, I want to panda inspired planter. And I want it for 3D and I want to be able to print it on my 3D printer. And so a couple minutes later, it gave me the file, STO file. And then a couple hours later, I now have a panda planter. And I didn't have to do anything to it. I just four buttons and boom, boom, boom. And it was there. So I mean, your world is your imagination now. I mean, I want to try something. I wanted to do something a little easy. I didn't want to be too complex. I'm going to try something more complex. I have an ultimate idea that I've been wanting to do for a while that I want to do just like my Deadpool mask. I want to

48:12do a panda mask. And so we'll see if it can do that. I think that that'll be the next big change. But I can go bonkers just in my own garage, recreating custom hooks for random things that I want to hang up on my concrete, why the singer block walls on my garage. So just different types of hooks. I mean, you can get the stuff from from like, come depot, but they're really expensive. Like to get their systems. And sometimes I'm just like, I want to tap on, I want to, I want to put it in, but I want it to hold, you know, whatever, whatever this is that I want to get up off the ground. But it's an odd object. And I just want to, you know, so like just dumb stuff like that. Like, I could go to town with a 3D printer coming up with all different ways to organize and hang things correctly and just be have a more efficient garage and get more floor space and stuff. I mean, I have a couple of questions. One is I posted this. You probably answered in the daily AI show community, you know, Slack. But is that paintable? It came out in red, but you can paint on top of that.

49:15Yeah. I only have a single color 3D printer. If I had a two, put this in the sock too. If I had a multiple color, I would have done that. I would have, I could have done, I just had red and I didn't feel like changing the filament. So I just threw it in there. Not that it's hard, I just have a lot of, okay. And then the other mighty mouse coming to a, grandson near you. Yes, but there is, yeah, the other question is, is this a unique capability of Astro? Could you use Soul 5.6 to do the same thing? I've tried it with others. It didn't turn out well. Astro has, I'm sorry, go. No, yeah. I've tried it with others. The models weren't right, weren't correct. And so Astro has a very good CAD system, also very good 3D system. Is there anything comparable in the flawed suite that can do it? No. Astro sees better than Google's vision models. Like,

50:16they did something in growing Astro that has changed the way it perceives visual information. And that's where these sorts of things are coming from. There's also the, the founder of scenario, which is a 3D gaming building kind of thing said, it helped him build something he wished he had as a kid so badly. Input just an image for an idea and a few words and get a fully buildable Lego set. 100% customized to your prompt and he's working on it. And when he gets it to a place that he really likes, he's going to open source it. That is something that I have also been working on for a long time. Since AI first came out was picture to Lego. I forgot that project for a long time ago. I started it back up this week, this last week. I just started looking at how to build it. And I have seen that post. I've seen

51:18people are doing it that you can take a picture uploaded or that builds Lego systems that you can take a picture uploaded and it will build it out in Lego for you and give step-by-step directions on how to build the Lego. Which is awesome. And he's calling it the okay, I just had an I know come food moment. Like whoa, I just realized I was given superpowers. Yeah, it's pretty crazy. Before we close, what 3D printer do you have that has the scale or you can build something as large as your planter? Well, they all have pretty similar sizes. I have the Flash Forge, which is a Britain not like the most common brand out there. It's still a really great printer. But I think they all have the same printer. I have the Flash Forge, I can print any material, so I can print carbon fiber, I can print anything because it's all encapsulated. And so it doesn't

52:20let out the fumes. But they're I mean, they're all pretty similar sizes. All right, thanks. Yeah, coming also new new gadgets in the barn, right? Like, yeah. Thanks. All right, well listen, we'll wrap it up there for today. It's a holiday. Thanks for everybody hanging out with us if it is a holiday where you are. But thanks for coming to hang out with us regardless. Don't forget you can always join the community, the dailyIShowcommunity.com. You can go to our recently RayVamped website, the dailyIShow.com. And you can also sign up for newsletter on there and see all of our past episodes. And also do a really cool search function across all those episodes. So, you know, when you want to know the last, all the times we talked about AGI or all the times we talked about a particular topic, you literally go to the website, you can type that keyword in. And it will quickly come back to you and tell you all the different shows where we talked about that topic. And as we've said many times, we will only make that

53:24website more powerful. We have, we know how to do it. It's just a matter of getting the time and effort in and, you know, not spending or I was playing with Ashtra and all the models. So, I know we didn't probably talk about Ashtra a lot today. I know people were in the comments talking about it, but, you know, we'll be back the rest of this week. I'm sure we'll be sharing much many more builds kind of like Garrison and whatever everybody else has been doing with the model. Lots of great use cases have come out into the wild, really, really cool stuff. So, we have, you know, this whole week to get into that as well. Thanks everybody. Happy birthday again to Gwen and we will see you guys tomorrow. Bye.