Back to episode — Episode 2297 Scott Adams - CWSA 11/19/23, Extra Tasty News Today, Find Out Why
Context —
t day I ever had, which I've never had, I'm just saying, I'm trying to imagine it. The best day you could ever have was Sam Altman's last day or two. Imagine getting fired from the most storied company in modern history, $86 billion worth of value, changing the world in every way, and you get fired like a dog. And within 24 hours your entire company says no, whoever fired you, you guys got to go a…
← Previous segment →ly she's highly qualified. She's an engineer for the most important job probably in the world at the moment, so she must be crazy qualified as well. And I can guarantee that she's smarter than the average person. I mean I can guarantee that personally because she's been following me on X. And wouldn't you agree people who follow me on X are way above average in intelligence? I mean it's a trend. It's a thing.
All right. So there was some deadline that the staff set for the board resigning, but the deadline has passed, so we don't know what happened. There might be some waffling going on. But also there's a rumor that the real problem here was that maybe Sam Altman wanted to go a little faster with the development of AI, and maybe the board wanted more caution. I don't think we know that, but that's the current speculative best idea of what the difference was. But we see that GPT-5 is coming, and the early indications are it's going to be a wow update. So right now we're at GPT-4. So I've told you I've been using GPT-4. And here's what GPT-4 doesn't do. So I don't know that GPT-5 will do these things, but I'll tell you what can't be done with the current one.
Currently GPT-4 will let you upload a file. So if a file is uploaded into what they call a GPT, it's like an app, then you can ask questions about that file and it can keep referring to it. But it won't take a big file. So the limit on file size is so small that you can't even upload something like a book. Like you can't even upload a book. A book seems like the minimum amount of a file that you should be able to handle, right? A book. One unit. One book. It can't even do that. It also doesn't talk to you. I don't think. Maybe somebody fact-check me on this. But you have to literally type when you're talking to it, during a time when AI can listen to language better than any language processor ever has. So it makes no sense at all that you have to type to it when you should be talking to it. All right, so that's a limit of the current version.
The current version also cannot return an image. Is that amazing? You can ask it for some information, but it can't show you a picture. What? That's crazy. Then the other thing you can't do is it can't remember your last conversation, and you can't train it by just telling it stuff. It won't remember it for the next session. It'll remember it during the current session, but as soon as you log off and back on it won't remember it. So just think about the extremeness of those limitations. It also can't go off and do something on a website for you and come back. Like you'd have to add some other architecture to do that. So there's a whole bunch of stuff that GPT-4 doesn't do that are super obvious that you would need for it to really be useful in your life.
Like GPT-4 is still, in my hyperbolic opinion, so this is not exactly true but this is how it feels, it feels like AI is still demo-ware. Meaning all the work that I did to try to make an app that works called a GPT, in the end all I could do is show a demo of what it should be able to do if it worked. And everything I've ever seen has been a demo of what it should be able to do if it were better. Right? If it were better it could definitely do this, and watch me demo it. But you can't do it because it's got some limitation built in. So now just imagine if the only difference is that GPT-5 doesn't have those limitations, then all bets are off. Because you can't do a damn thing with GPT-4 really except play around. But GPT-5, maybe. We'll just see what they free up there.
Some people say that it might be AGI, which would be the super intelligent version compared to what we have, which is just basically pattern recognition. I don't think super AGI is coming, not really. I think that people might call it AGI because it'll be cooler, it'll seem a little smarter. I don't think so. But suppose we get to something close to it. It's going to open up a lot of questions. Sam Altman said before he got fired in some event, he asked this question: Is it a tool we built or a creature? In other words, there are real questions of whether you've created some kind of a creature, a living creature. You can argue what living means in this context. But as I posted on X, this will raise some interesting questions. And if it's not GPT-5 that raises these questions, you can guarantee these will be raised in some subsequent version, probably in a year.
And here are the questions I think will be raised. Are we, meaning people and the universe in general, God's debris? Is everything in the universe just God's debris? And is AI the beginning of God? God's reassembly? Is God our past or our future? Because it looks like we're creating God. Were we always meant to create God? Now I don't mean that GPT-5 will be God, but there's some version after that that's going to be pretty indistinguishable because it'll just be able to do anything basically. So that's an interesti
Context —
ng question. Also I have a book called God's Debris that gets into that question. Number two, wait until you find out the implications for free will when AI tells you it doesn't have free will and it acts just like people. What are you going to say? Are you going to say, oh look, but you're a machine. People have free will. And then you can't tell any difference between the machine and a person t…
Next segment → →