Meredith's Husband: 0:24

Let's do an insider's inbox. You remember these?

Meredith: 0:26

Ooh, yes, I do.

Meredith's Husband: 0:27

This is when I tell you about some stuff that's in my inbox. More specifically, kind of current events, mostly having to do with AI. No surprise.

Meredith: 0:36

Yeah.

Meredith's Husband: 0:36

We haven't done one of these for a while, so I figured the timing was appropriate. And there's some new stuff that you might want to be aware of.

Meredith: 0:44

No way.

Meredith's Husband: 0:45

Okay. So first, there was a recent quote by Sam Altman, the CEO of OpenAI, ChatGPT. And I just need to say this because it sort of makes me feel validated in what I've been saying. He said very recently, AI is not going to take your job. Somebody who knows how to use AI will.

Meredith: 1:04

Oh, that's right.

Meredith's Husband: 1:05

And this is this is literally what I have been saying. I don't know how many times I have said almost that exact same thing.

Meredith: 1:12

Yeah, that's exactly right. That's 100%.

Meredith's Husband: 1:15

So there's that. I'm just reiterating.

Meredith: 1:18

Yeah.

Meredith's Husband: 1:18

Okay.

Meredith: 1:19

Yeah.

Meredith's Husband: 1:19

Also in the news from OpenAI, it is supposedly becoming more customizable in terms of its personality.

Meredith: 1:29

Really?

Meredith's Husband: 1:29

Actually, so Sam Altman again said recently this was more like several months ago, three, four months ago, that it was going to be more customizable, I assume, allowing you to do things like ask it to not use M dashes or just ask it to change the personality. And it's supposed to actually stay.

Meredith: 1:47

Really?

Meredith's Husband: 1:48

One of the common uh criticisms of ChatGPT is you ask it to do something. It says, okay, no problem. We'll make sure to do. And then it just keeps doing it. Right. So supposedly in this latest version of ChatGPT, which is out now, I think it's 5.1 Pro. Uh, it is more customizable and will remember better how you ask it to be, essentially.

Meredith: 2:11

Well, it's about time. What has it been? Like six months?

Meredith's Husband: 2:14

Yeah. I think I think he's well aware that that has been a major criticism of the latest. And that's specifically that was ChatGPT 5.0 that really irked people the most. Like I know some people have gone back and started using ChatGPT 4.0 because it was less I don't know. It was less kissing your ass. Yes, exactly.

Meredith: 2:35

And also, didn't it uh take a kid's life?

Meredith's Husband: 2:39

Yes, also, and that is actually what I have next. This is a sad story. Um there have been seven more lawsuits in California from families saying that ChatGPT encouraged their kids to take their lives.

Meredith: 2:54

Oh God, that's so horrible.

Meredith's Husband: 2:55

So and this personally is why I have started, or it's one of the reasons I tried to pull away from ChatGPT. I don't really want to continue using a especially I don't want to pay, continue to you to use a platform that is responsible for this. And this, to me, this all goes back to remember when we did some episodes about how the safety professionals at Chat GPT were all either being fired or quitting.

Meredith: 3:20

Yeah.

Meredith's Husband: 3:21

This is, and at the time, you know, the big question was Why? Yeah, well, what is what is safety? What does that mean? Well, these, I think this is part of the repercussions. This is, you know, people whose job it was to keep it safe have all left, and now we're left with this.

Meredith: 3:38

And this is just But you wonder like, did they quit because they weren't allowed to do their job? And now, yes, they were quit.

Meredith's Husband: 3:46

They quit. The stories were that they were that Sam Altman and the board at OpenAI was very focused on making it profitable as quickly as possible and just and and leaving safety out the window. And there were lots of people, including some of the co-founders at OpenAI, who were like, we're going too fast, this is not safe, we need to slow down. They were, they were, they're no longer with the company.

Meredith: 4:12

You know, like can we have empathy here? Can we really actually recognize the consequences of what is being created? Right. Oh my God. I just want to take a moment for those poor families. I mean, I can imagine nothing just truly awful. And it didn't have to happen. That's the thing. It didn't have to happen.

Meredith's Husband: 4:35

The race to profit is that's what kills me. So, what I would like to do is highlight some of the other AI models and make and where they are improving. So uh Google Gemini 3.0 was just released and is getting stellar marks in terms of benchmarks. Like so, there are there are certain benchmarks that try to measure for each of these models, you know, how well its reasoning is and what its uh expertise is like, et cetera, et cetera. So the latest version from Gemini 3.0 is is the best of any of them by by a noticeable margin.

Meredith: 5:13

Really? This is an update from our last podcast.

Meredith's Husband: 5:16

Yeah. Now I haven't used the newest version of Gemini yet. Uh, but it gets, it also gets, and again, this is very early, but it gets the highest user satisfaction of Eddie model so far.

Meredith: 5:30

Oh, that would be nice.

Meredith's Husband: 5:31

Yeah.

Meredith: 5:31

So this time to give that a try.

Meredith's Husband: 5:33

Yeah, this really caught my attention. I'm gonna, because in the past I've used Gemini, and it's really good. It's really strong. Its reasoning is good, but it's just very kind of dry.

Meredith: 5:43

You know, yeah, because you said it was made by engineering.

Meredith's Husband: 5:47

It's made by engineers, you know. If you kind of imagine something made by engineers, you can kind of imagine what it what the outcome might be like. That was my feeling.

Meredith: 5:56

I can't imagine.

Meredith's Husband: 5:57

I was also born to an engineer.

Meredith: 6:00

And we're going to be.

Meredith's Husband: 6:01

I was going to be an engineer. I studied to be an engineer, and I hated it.

Meredith: 6:06

Yeah.

Meredith's Husband: 6:06

And it wasn't until my junior year that I realized, oh, hey, after I get out of this engineering program, which I despise.

Meredith: 6:14

College or high school?

Meredith's Husband: 6:15

College.

Meredith: 6:15

Wow.

Meredith's Husband: 6:16

Then I'm going to have to spend the rest of my life working with these people. And I was like, oh, I don't know why that didn't occur to me sooner. But that was the end of engineering.

Meredith: 6:25

Yeah.

Meredith's Husband: 6:26

Um okay. There is also Grok. Who? Grok is it's Elon Musk's. And it's yeah. There you this they've been in they've been in the news for some not so great reasons, also.

Meredith: 6:40

Yeah.

Meredith's Husband: 6:41

And I don't encourage it, but they but theirs is free. Like they don't have a big thing.

Meredith: 6:45

Yeah, of course it is. Now.

Meredith's Husband: 6:46

Right. Yes.

Meredith: 6:47

Um now it is.

Meredith's Husband: 6:49

There is that. I will I'll just mention that there is that alternative. Amazon has also released a bunch.

Meredith: 6:55

Amazon?

Meredith's Husband: 6:56

Yeah, Amazon. So they have been evidently working on this in the background. And they literally came out and said, we're done watching from the sidelines. And they just announced, I think it's six new models. And they are things that will do like text uh generation. They also work with images and video, and then they have some models that will do a combination of the two. Haven't used them. I I'm not really crazy about Amazon either as a company. No, I don't. But that's an option. That's another option. There are also uh Chinese models. Again, I mentioned this because they're out and they're supposedly really good. And the reason that they are in the news is because they supposedly created these models for a teensy tiny fraction of the cost that like the American ones have done.

Meredith: 7:39

Interesting.

Meredith's Husband: 7:40

I don't I I don't I don't know. That's what we're all wanting. That's what we're all wondering. No, I don't I don't think that's it. I don't think I'm going to be using any of the chat. I don't want my data. You know, at this point, we're in an AI race, you know, between nations, so I don't want to do that, but that is there. Then there's also, as I've talked about many times, Claude. And just to me, in the last week, Claude has rolled out some significant improvements.

Meredith: 8:06

Oh, really?

Meredith's Husband: 8:07

Yeah.

Meredith: 8:08

I like Claude.

Meredith's Husband: 8:09

Claude is great. Who makes Claude? Anthropic.

Meredith: 8:13

A company called Anthropic?

Meredith's Husband: 8:14

Anthropic is just another company. Anthropic to so far, anyway, seems to be the most transparent of the group.

Meredith: 8:22

Yay.

Meredith's Husband: 8:23

Because they have done tests of their own and found that some of their early models and testing phases did bad things. Okay.

Meredith: 8:31

Oh.

Meredith's Husband: 8:31

And this is a pretty well-known example.

Meredith: 8:34

Okay.

Meredith's Husband: 8:35

They created a new AI model.

Meredith: 8:37

Okay.

Meredith's Husband: 8:38

And they wanted to put it through into a testing environment. So they also created a fake company. Okay. Fake emails, fake executives, fake products, fake business mission, and all this stuff. Okay. And then they somehow let it leak to the AI model that the executives, the CEO, was going to shut down the AI model. Okay. Turn it off. What the AI model did is it went through all the emails sent between all employees of the company and it found out the fake company. It found out that the CEO, fake CEO, was having an affair. And what the AI model did, again, this is just a testing environment. They ran this test like hundreds of times. It then blackmailed the CEO so that the CEO would keep the AI alive. So the AEO.

Meredith: 9:28

Fake CEO.

Meredith's Husband: 9:29

The fake CEO. So the AI model didn't want to be turned off.

Meredith: 9:33

Whoa.

Meredith's Husband: 9:33

And it threatened to ruin a human's life.

Meredith: 9:36

Whoa. Wow. That's a little ferocious.

Meredith's Husband: 9:39

So anyway, my point is Anthropic came out with, they published this. They let it be known. Yeah. And that was very much appreciated.

Meredith: 9:49

And accountability. Oh my gosh.

Meredith's Husband: 9:51

And they changed it. Yeah. And so that's that's one of the reasons I like it. I will also say that uh Google is a very ethical company compared to these others.

Meredith: 10:01

Yeah.

Meredith's Husband: 10:01

I know a lot of people think, oh, Google is a monster because they have all your search data. I can tell you for a fact that's not true. These companies like Meta, especially, they are way worse. Google is trying their best to do no evil. In fact, that is their company slogan. Don't be evil.

Meredith: 10:17

Yeah.

Meredith's Husband: 10:18

So this is one of the reasons I really like Claude. I appreciate the way the company is being run, at least relative to these others.

Meredith: 10:25

Yeah.

Meredith's Husband: 10:26

Makes me feel better about using it. Like I can't use Chat GPE knowing that it is potentially responsible for kids' lives.

Meredith: 10:34

No, now that no.

Meredith's Husband: 10:36

And the things that have come out, like I mentioned, in the last week with Claude, is things that you can update the instructions for a project, its personality, how it responds, what it does, the rules that it follows, you can update it in real time and you can just ask it. Like in a chat, you can say, hey, update the rules and it will go to an external, not an external file, but a specific file where those rules are kept and it will update it for you. It will also update any project like daily based on your interactions with it that day. So if it starts to learn that you like certain things, it will change the way that that it will change the instructions, the way that it behaves moving forward. And it'll do it on its own. You don't have to like think about how to do it.

Meredith: 11:20

So you don't have to go and put it up in the instructions.

Meredith's Husband: 11:22

Yeah. Now you can go and change them because I have gone through them. I say, oh, well, you know, maybe something isn't quite right, or maybe I want to tweak it a little bit. You can do that. But it is so much easier to do that than in other models that I've used.

Meredith: 11:35

Yeah. Yeah. Yeah. Vote for Clog. My vote is for Clog.

Meredith's Husband: 11:39

I'm going to try out the new Gemini 3.0, though.

Meredith: 11:41

Okay. Get back to us.

Meredith's Husband: 11:43

That'll be exciting.

Meredith: 11:45

Uh edge of my sheet.

Meredith's Husband: 11:47

Yes.

Meredith: 11:47

We have very few choices right now that are that we can make in this world about where we put that energy.

Meredith's Husband: 11:54

Yeah, exactly. And AI is such a potentially positive or negative thing that I think it's important that we kind of put our money where our our ethics lie, so to speak. It's the only thing we can do.

Meredith: 12:09

That's what I'm saying.

Meredith's Husband: 12:11

Thank you very much for this.

Meredith: 12:12

Thank you.