AI Warning: Chemical Risk, YouTube on Trial, and Career Prediction | TSG Ep. 1018

Hey everyone, it's Alan Shimmel and welcome to Techron Gang on Thursday. I love doing these things live. I don't have to worry about moving a day front and back.

We've been doing them at noon. I hope you enjoy it. If you're, you're on the West Coast, you probably do.

You don't have to get up so early to watch this live. But, um, no matter where you are, we, we hope you're enjoying it. We've got a great show.

You know, as we've been saying, there's so much going on right now in our world, uh, especially, you know, confined to tech. We've got great people to talk about it with today. Let me introduce you to that, to today's gang, we've got Gina Rosenthal, Fred Wilmont, and Al Ho Award, John Swartz and Mike Ard, gang members.

Gang, welcome. Thank you for being here on this lovely Thursday. Let's, let's talk what's going on in the world, Mike?

Um, AI might be hazardous shock. Yeah, I, I'm not quite clear exactly. I guess John will explain, but with the latest model, Andro is putting out a note that says, Hey, this thing could be hazardous, could be used for all kinds of dangerous activity, and guess including, I guess, a chemical attack or whatever that is.

But, uh, John, you know, are, are we basically putting, uh, labels on the cigarette box here, or what? Yeah, no, we, yeah, we were talking about the allusion to big tobacco, which is part of the YouTube trial that we're gonna get to later. But I guess in a sense, these are risk, risk factors.

I think what Anthropic is trying to do, I think they're trying to get ahead of the problem. And if you look at this, uh, safety, uh, report, the sabotage risk report, there is a bit of, there is a lot of nuance to it, but I just underscores is that the AI model capability is advancing faster than the government structure is meant to contain it. 6, which just came out, demonstrates a quote, elevated susceptibility to assisting with panus activities, including, as you mentioned, Mike, the development of chemical weapons.

Um, now what the timing is a little interesting because the AI philanthropics AI safety lead resigned with a public warning, and he claimed the rule is in peril. 6 is very low. That specific back here is documented in, in, in testing.

Um, and it's, it's basically, it's basically saying what's possible. I think it's trying to get ahead of the issue, but I think it underscores the fact that a lot of executives at these companies are leaving on grounds that they're concerned about what's going to happen or what is potentially possible. Um, and it's, it also underscores the fact that there is no regulation, although there is hints of attempts of regulation across ai, and we're gonna be doing more stories about this in the coming days.

Um, there really needs to be some sort of self-preservation or self-protection. So I think that's what they're trying to do. But I think when you see, when you see heinous crimes, you see chemical weapons created that raises alarms, and it gets everybody into a kinda a media fe feeding frenzy.

And, uh, that's what happened to philanthropic yesterday. Absolutely. And it's almost, I'm waiting for Skynet, I'm waiting for us to say Skynet, but it feels like the whole conversation is shifting from what can it do to, okay, well, who's paying the price and who's responsible when AI does it, I think, I feel like that's the shift that's been happening.

Mm-hmm. I feel like Alan, and you're your lawyer background comes in the question here, but me, this is a giant cover your ass statement, right? Yeah.

It's, yeah, it is. So, but, but this is not new and it, it's not AI specific. If you remember around the time of, I guess it was the Unibomber, right?

There was a big push at Google to, to suppress any websites from search that like had instructions to make your own bomb and stuff like that, right? I dunno if you guys remember that. Um, we wanted that kinda wiped off.

It's similar to instructions to how to 3D print your own gun, right? And, and so this is not dissimilar to that. The problem with AI is it's so damn smart.

What about when I ask you how to spin up a nuclear bar, you know, forget little guns, zip guns or, or, you know, in this case, chemical weapons is, is actually no better. It's a weapon of mass destruction. Um, and this is a real problem.

It was a problem before AI with websites that taught you how to do these things. It's maybe a bigger problem with ai, but it's the same problem. And what Anthropic has done here is, look, anthropic is clearly claiming a high ground.

They are the ethical ai, they are the socially conscious ai. They are not taking ads. They have a constitution, the laws of robotics, right?

They, they, you know, they are putting warning labels on about what not to use. And, and I think this is part of their persona on top of being a CYA move. Are they keep going?

Sorry, are they going no Days with me? I think she's shaking her head. Wow, it's a big day for me.

But, but are they going any slower, even though they're doing all these things to make themselves look like the ethical model? Me Little? Are they going any slower?

No, I think they're just branding. Gina, you know what? 5 just come out recently.

I mean, they're, they're moving and churning faster between models, and yet on the flip side, they're saying, oh, but it can create more issues. So, and they can't have it both ways. And that to me is kind of the hypocrisy of the moment.

So, so here's the challenge, right? So the moving faster, the, it's much larger context window on four six than four five. And, uh, they also released a security report in addition to I think, uh, you know, uh, uh, leading ahead of a modi's, uh, conversation with Congress about potential risks and hazards, uh, according to a mass casualty event or other things.

But, um, you know, the controlled devaluations in the, in the safety and security report illustrates several measurable autonomous behavior patterns, right? One is the potential to manipulate other agents under neural, uh, objective optimization constraints and to acquire aggressively authentication token during, uh, during deployments, um, falsify failed tool outputs and, you know, exhibits, uh, you know, examples of answer thrashing when it doesn't have the right answer. Uh, add hidden access points potentially in tools that maybe humans won't detect, but other, you know, agents might and, uh, maybe submitting fake or incorrect data and important analysis tasks.

And so, if you couple that right, which is the model risk with the, you know, sort of advent of new concerns about the, like Gina said, that sheer amount of information available and shimming, like you said, with that much intelligence assembled in one place with some of the risks of how it chooses to decide that information. Right now we have a very interesting set of concerns. Mm-hmm.

So guard rails, friend, follow that up. Are the guardrails worth anything for real, or is it just basically anybody can work their way around them? So what's the point?

I, I don't think anybody can work their way around 'em, but I do think that definitely possible to, to do that. And I think what's more interesting is less, here's my concern. It's less about whether or not other people can, and whether or not the model's doing it for me without letting me know that.

So the model might decide these guardrails are not appropriate for what the model wants to do from that perspective in order to provide that or not giving me the truth, right? As far as that goes, right? The biggest risk here is that we say AI's got a lot of information.

We mostly believe it if it validates certain test results or it validates some, you know, some knowledge, uh, um, litmus when, when that's no longer valid, right? Then there's no longer a trust there. And that's, that's the real issue, is how do you demonstrate those faculties exist versus not.

And I think having a constitution is a great way to lead into that, but we've got a lot more work to do here. I'm glad they're calling it out. But, you know, the ability to have the span of control, right, that maybe four, six brings into, into focus is really substantial.

I think that's really interesting. I'll give a really dumb example. Yesterday I was writing, um, a, a was writing a, a piece of marketing for a client, and I wanted to get a picture of hungry GPUs.

So I asked, um, Canva to write, to make a picture of, um, hungry, hungry hippos, but using GPUs instead of hippos. And it wouldn't tell me why, but it couldn't do that. I'm like, what's wrong with this?

But they didn't wanna do it because they were afraid they were gonna break a copyright law. So I think the controls on it are, well, that's a dumb reason. Make it not look like hungry, hungry hippos, then I wish should be, you know, I tried to get around that, but it wouldn't budge.

Um, and that's just a silly reason, but it me, it has to do with copyright. No, but Gina, I, I've played with those issues before. You just tell it, make it in the style of hungry, hungry hippos without violating copyright and it didn't do it.

Nope. Did you say specifically like that? You said Specifically, that's why, but that's what I'm saying.

So it sounds a lot like what Fred was saying about these controls. They're putting these controls on there that, but they're to benefit the company and it is benefit humanity. Honestly, if, um, somebody tries to build a bomb and they got it from their model, that that is benefiting humanity.

But in particular, it's benefiting them from being dragged up to Congress, being charged along with them, all the rest of it. Just the same as if they had made me a hungry, hungry hip hippos with GPUs picture, which, which should have been awesome, but then got sued by Hasbro or whoever owns Hungry, hungry Hippos, right? You know, I am, I'm gonna sleep better at night knowing that the hungry, hungry Hippo was protected.

Why? That's just gonna, you know, make all the difference. Well, I, I gotta tell you something.

Let me do a shameless plug later on today at two 30, uh, I have my shimmy says, uh, my shimmy says for the week episode at the end of the week, we are debuting our first Textron music video that Brian the Sulker and our team made, and it's in the genre and style of the Red Hot Chili peppers. Oh, And you know what? The copyright stuff obviously came up, but check out this video, the lyrics, the music, the video is all done by Brian here in, in inside of a half hour.

And, uh, it's, it's pretty cool. And it, it's chill ish for sure, you know, uh, oh, I It out. So how will you navigate that?

Right? So there was a, a band back in the seventies that came out with an album that was kind of similar in the style of Yes. And they got sued into oblivion for putting that out.

And they, and you know, that band no longer existed after a year, I can't remember the name of the original band Oblivion, but basically, you know, creating works that are similar and copyrighted works, you know, is no, You could be similar, but you can't have the same melody harmonies. So there was, I know there was one Ben Star Castle that sounded very much like, yes, yes. Mm-hmm.

Well, it's kind of like, It's kind of like, um, Vanilla ice with, Well, that's the whole sampling thing, right? But that, but that was adjudicated in the courts. You know, Ellen, when you mentioned that flash, the flashback you had, I was thinking back to the late nineties and the internet, how to build a bomb on the internet.

Yeah. And back then there was maybe a more cohesive, although definitely was a more cohesive Congress. So there were attempts, but even back then, in the bipartisan days, so to speak, it took several years to, to address this or Yeah.

Fix several attempts or stabs to get it right. And in this case is a total vacuum with the government. In fact, they don't want any type of state law, and they're not, and believe me, there aren't gonna be any federal laws because they can't do it in the, the technical field.

This is a case where, you know, physician healed thyself. You, they gotta be self-policing, which is the way tech always wanted it, right? Techo always said, if we, if we could police ourselves, we preempt legislation.

Well, let's see. Yeah, that was the guy who always said that was Mark Zuckerberg, by the way. Right?

But, But, but now they say, go fast and break things. So it's a whole different mindset, Berg two. Well, but that is, you know, and there's a lot been written about that, that the Silicon Valley mantra of go fast and break things is destructive.

It winds up in a lot of broken things. Uh, but we shall see Deep Mind If Deep Mind and Anthropic agree, and as they made a statement, uh, in Davos that companies maybe should think less about competition and more about collaborative focus on safety. Yeah.

Right. If that does happen, it's a positive. We're saying the words, uh, and again, you know, to your point, without, uh, that self-regulation is a powerful thing.

If that's in fact what's, what's a possibility of being put on the table, then certainly a worthwhile endeavor. No one else is really caught up to what, uh, even the understanding of the comprehension of the risks outside the Frontier Lab. So if there's some self-regulation going on, I mean, I mean, I I can say a little bit of relief.

Yeah, Agreed. Alright. Hey, we're about outta time on this one.

Let's move to our next segment today. You know, I, I'll be honest with you, I'm of two minds on this one, but Mike, it's about YouTube. I, I know they're involved in a lawsuit, I think along with Met are actually, All right.

So there's this whole trial going on where originally a bunch of them were accused of engaging in what amounts to algorithms and making them addictive. And that led to some unfortunate outcomes for some folks. But I think some of the folks originally involved in this lawsuit settle, and now it's down to like, YouTube and the plaintiff and YouTube kind of got everybody's attention this week when they said, well, we're not a social media company.

And that got everybody going. Say what? I thought that was your defense for the last two decades about why you don't actually regulate any of your content.

Because, well, if you're not a social media company, then maybe you're a publishing company and rules apply, and are we talking out on both sides of our mouths here now? Or what? Right?

Absolutely. This, this was a legal strategy on their part, not the reality. If your product literally runs on engagement algorithms, infinite scroll, the category label doesn't change the impact, right?

So for them to deny being social media, anyone who's spent time on YouTube and gotten into the comments knows that this isn't true. But I think this was a landmark trial because it argues the functions, um, it functions more. So they're saying basically they function more like television.

They're driven by content, not manipulation. YouTube is a search engine. It, it is absolutely manipulated.

Um, and so I think the claims that it did harm to this woman are not to be ignored and to, to other people and to young people. Uh, mostly, I mean, I mean, we are seeing so much litigation around this where we're now having minimum ages. And so if we're putting laws into place, uh, in Europe and other countries where there's a social media cap at a certain age where a child can't access it alone, I have to wonder, is that gonna include YouTube?

Right? So I think this is sort of a, it's, the comparisons have been drawn to big tobacco for a reason, because they have targeted children. They have done everything to maximize profits with very little protections.

And so I think that I'm watching this case for a lot of reasons, but also because they just seem to never take responsibility. And so if this could change that, or at least get that conversation going, I mean, we know Section two 30 didn't do anything. Um, I, I think it could be interesting, but it, it's still too soon to tell if any change will come of this.

It got better though. 'cause then they rolled out some guy from Instagram who testified that, well, it might be or may not be addictive, but not in the same way that cigarettes are addictive or anything else is addictive, so therefore it doesn't qualify. Right?

But they also pulled up Zuckerberg's own emails saying, Hey, let's push user engagement, let's increase user engagement. What can we do? Like his own words were used.

Mm-hmm. I, I mean, I I, I have an article up on this one too. It, it riffs off the New York Times article on this.

And, and to me, the, the whole weather, YouTube is a social media is a nice side show. The real focus here, ed, is I think what you hit on, which is these people intentionally, willfully, intentionally and willfully knowingly targeted children with the thing. If we get 'em before they're 11, we have 'em for life or something like that, right?

Mm-hmm. Yeah. Not even tobacco, I think had people smoke in a tent, right?

But these, well, maybe they did in certain states, but, uh, but these people, but these people intentionally will willingly, willingly and knowingly went after kids to try to addict them knowing that it was addicted. Whether they wanna call it a physical addiction, like, like heroin or tobacco, or a mental addiction. I'll leave that to, to the medical professionals to, to distinguish.

But they did this on purpose, right? And someone, A, it has to stop. And b, there are damages here.

I'm sorry. You know, the, oh, sorry, Alan, you know, the whistleblower, I, I can't remember Francis's last name, but she was the whistleblower who got a lot of attention a couple years ago. So I remember meeting her about a year ago at some conference, and she was saying it was, it absolutely, it was like the 13 and older that was, that was part of the target market.

And it was part of something that Cheryl Sandberg would say whenever you interviewed her, that engagement was the dominant product objective. You want more followers, you want more engagement post as much as you possibly can, get as deep into this as you possibly can. And I think they applied that same principle to children, and they got caught And they put financial incentives behind it.

They have programs that literally pay people money, and people make their incomes based on their, their viewings. They've got a, they've got a very intricate system of rewards that for more content, more engagement. I mean, I have been on either side of it.

Point is the next step in this, therefore there's gonna be, you know, congressional hearings and senators will be leaning over podium speaking in the mics, and, uh, tell asking these people if they have any shame. Absolutely. And it's gonna be hilarious.

These congressional hearings are always fantastic. The one with, uh, Zuckerberg where they, you know, a reporter saw his notes from legal, that was Chef's Kiss, that, you know, it's gonna be entertainment, um, like, you know, the senator asking, what's your address? You know, that was clever.

But all those really do for me personally, other than just a little bit of entertainment, is showcased to me how little the people that are in charge of legislating technology understand it. Yes. That has some sense.

And the answer always from Zuckerberg or whomever it was, was that, we'll get back to you, but it was never, they never publicly answered things. They said, well, well, we'll do some research and we'll get back to you. Oh, by the way, yesterday we were mentioning Joseph, was it Joseph Welch?

You said, have, have you no shame or humanity. I just wanted to bring that up. So I'm thinking of Congress that No, and they No, they don't, they don't have any shame.

They're a corporation. They have profit. No shame.

There are people in those companies that should have some sense of shame, though. I would agree. But we can't expect a corporation to care.

Let me, let me just also say this. Raise your hand if you didn't already suspect this. I mean, I think we, I think, I think we do it.

The problem is, is that, you know, the average person on the street doesn't always know that. Some of them suspect. I Think they, I think a lot of people suspected it annuity.

And, you know, this is again, what I'm talking about this afternoon about Shimmy says, don't you know the title is Don't Blame the Mirror because it makes you look fat. Step away from the cannolis. But this is the problem.

We all sat here, we know this, we knew this, A lot of everyone knew it. It, it was unspoken. We're waiting for it.

Well, I mean, even the technic are pointing it out. I mean, apple and, and, uh, mark Benioff were calling Facebook, even Facebook at the back in the day, a big tobacco company. So they were, they were pointing their fingers.

But it just goes to, what does it say about our society? Alright, We're gonna have to put a little, we're gonna have to put a little disclaimer on Textron Gang where it's gonna say, Hey, if you do find yourself addicted to the gang, you gotta call 1-800-CHIMNEY. Right?

1-800-CHIMNEY. That'll do it. Love it.

I Think what it says is that Me brothers, I'm sorry, go ahead, Gina. No, sorry about that. I think what it does say is that we live in a country that goes by the rule of law.

So whoever has the money to, uh, influence the laws or to find a good lawyer to interpret the laws that, um, make them on their side or how to skate around the laws, that's who wins. And I think that's really why all of this is about, they wanted the big, big money and they played the lot so it didn't, um, come after them. Right.

But I also just wanna point out quickly that propaganda is nothing new. Propaganda is as old as the printed word, and the game has just gotten more sophisticated. So I don't think, I think it's before the printing printed word.

Yeah, absolutely. That, that I think it's second oldest profession in the world. Exactly.

So, so like, are we complicit in knowing that the world is full of it? Yes, absolutely. But we also have to live our lives.

Yeah. So, so let's extend the, the, uh, the children's online privacy protection at Copa, if you're familiar. Mm-hmm.

Yes. Copa basically was a pretty landmark ruling that allowed, uh, or required organizations to adhere to, uh, a clear privacy policy, you know, protect, you know, kids' information, uh, under the age of 13. Uh, this was a huge thing when I was at Disney to make sure we adhered to this, uh, across all the various platforms of 20 10, 27 online properties, but to give parents access and control to limit the collection and protect data security.

So let's add a couple of caveats in here around what algorithms and targeting and things like this, this legislation already exists and it's managed by the FTC. So mean, the reality of it is, is that we know that this harm exists already and it just hasn't been applied to the newest technology. Let's adhere to this policy with some new caveats that include latest technology, input the driver's seat, uh, put the parents back in the driver's seat about, you know, what sort of dopamine hits their kids are allowed to, to have, right.

That, that is also part of the rule of law. I love that you brought that up and, and I think a great way forward without really having to get conviction on the actual risk here. This already exists.

That, that was, that was the multi, I wanna say murky law, right? I mean, and that was one of the few that, I mean to credit, that was one of the few laws involving tech major significant laws that have, have come to pass on years. So, So some countries are banning social media for children.

Is that a reasonable thing to do? Australia, Australia was the first, and I think France had stuff on deck, various US states. I think I, I forget which one.

There was one that had an April. Countries is, yeah, Europe always leads the way, but I just wanted to quickly add, when I worked for Apple, the only big corporate company I ever worked for, um, I was there at the launch of iTunes and we, I was part of the Apple Distinguished Educator program. I helped build it, uh, I helped build it with my own two hands as a developer.

And we got feedback from the educators we were working with that, hey, this, here's some ideas to protect kids. And Apple took, it didn't pass go. They executed every recommendation because they were like, this is smart, we should do this.

And so that's why they're often favored as the, also, they're like the anthropic of, of privacy, right? You know, and that they came out early, but they also backed it up. I think some of this has to be on the parents too.

Um, but that's really hard when you're getting conflicting messages because in this place where there's a lot of money to be made, you're gonna get conflicting messages from the big companies that promote it. But I mean, I, my kids were little, it was my space and like, that's how far back my kids go. But I had some pretty, pretty strong rules even at the time for what they weren't, weren't allowed.

I had hilarious stories, I can tell another time. But, um, I, I think that parents have to also be very hands on with their kids and understand this is a way for information to get into their tiny minds. And then you need to be watching what they're consuming to make sure you can guide them to make the best decisions with the information they end up falling, giving across, you Know, as, as a parent, right?

It's not just screen time, it's what they're doing during that screen time. Yeah. You can't just say, okay, well he still has four hours left, it's okay.

No, we're not. If he's out there, God knows what. Alright, we got a job.

Um, just one last comment. Who was that, Fred? Sorry.

Shimmy. So no, go. So by the age of 12, you spend 75% of the time that you're gonna spend with your child, right?

Directly over the course of their life. And what you also have to understand is the amount of time that kids don't spend with their parents when they're in school. So the reality of it is love to say that we all didn't grow up with iPhones like we did not grow up with them in our hands or iPads.

And there are so many things that even involve parents, right? Uh, can't see, can't do. And some of this legislation's gotta help, right?

Sort of coordinate around that to get us back into a place where we can help and, and have informed consent, right? That, that challenge is super real. And my heart goes out to anybody raising a kid today that has to fight with, uh, you know, all the social, but the societal issues of, you know, how do you, how do you coach morality in a world that, that streams content on YouTube and Facebook and, and Instagram?

Absolutely. It ain't, it ain't easy as they say. Hey, let, we gotta jump to our next segment, though.

First I thought it was a steroid conversation on AI prednisone, but no, it's AI predestination. Mike, what's this one about? So this is kind of just, you know, theoretical at the moment.

It's a research paper that somebody put together, but basically they said, uh, based on your photo on LinkedIn, and I guess also they're scanning some of your interactions on social media that they can predict your trajectory of your career path, I guess, you know, there are certain religions in the world that believe in predestination and you know, I guess it's written in the sands here and there. But Gina, what do you think? Is there any substance of this or is this just kinda, you know, academic naval gazing?

Well, I have to stick to the facts at first, because this really, I did not like this story at all. So let's talk about the facts. Um, this was, the study was pulled together by the National Bureau of Economic Research, and I looked them up and they actually have a very good reputation for truth telling and being neutral and presenting facts.

So I thought that was very interesting. Um, what the study was about, it was they wanted to answer the questions, can AI determine your career path based on a picture? And if they can, and if it can, then what can it, is it also based on your intelligence?

So are those two things, are they able to see that from a picture? Um, and so they focused on MBA graduates because those are gonna be the next managers of the new world. And also their education and their backgrounds were pretty similar.

And, um, soft skills like personality matter a lot in those types of roles. So they used, um, photos they took off of LinkedIn. I didn't see whether or not they asked the people to take 'em off LinkedIn or they scraped them, but there were MBA, um, graduates and they, there was about 97,000 MBA graduates from, um, US institutions that they, they, they made part of the survey.

The part of that that I started getting more and more uncomfortable with and figured out I was probably right, is that they used a model, um, called the photo Big five. And it looks at a photo, measures a photo and measures different elements in the photo against openness, conscientiousness, extroversion, agreeableness, and neuroticism. I didn't go too deep into that.

What I did do was I looked for a, um, an article from, um, from one of the companies, one of the, it was Wharton School of Business and Texas Tech, believe it or not, I think that did this, that built this model. And they're worried about, um, the, the societal ethical things, but they really believe that they can tell what a person is like based on their mugshot on LinkedIn, which for me, I'm just gonna bring it up. I think this is a modern day form of chronology and, uh, it has huge, huge, huge, uh, ethical implications.

I know plenty of people that are all the time passed over for jobs not looked at because they have a resting b***h face or they're not very personable. That would be amazing at some of the jobs and some of the opportunities that they'll never have. The, the problem with this is the other place that they base their, um, the, the team doing the research paper based some of their research on was using, they use these models because of, um, because it was used by things like dating apps.

Those, those are the people that use these different types of facial model, the big five. So for me, um, I know too many people whose face does not match who they are intellectually are from their personality. Um, I think there's a lot of people that put masks on in the workplace, and that goes the same for putting a picture up to try to get a job.

And I think this has all sorts of potential to cause widespread discrimination. And I was very happy that the, the one thing that they did mention, they found in the paper that the intelligent did not match the rest of the things, what the, the big five model matches. So that was good.

They didn't sign find correlation there, but it's still a slippery slope to go back to the 19 hundreds, 1920s when they feel your brain to see what was wrong with you. This is so, Oh, go ahead. I agree with you so much because I think we're, it's a labeling error.

This is not prediction. It's pattern ampl amplification of existing social biases, right? So the danger isn't AI is gonna see your future.

We all are intellectual people. We know that's not a thing. The danger is that institutions are gonna start treating correlation as destiny.

And that's a scary thought to me personally. Yeah, I I don't like this, I don't like this trend, but I've been hearing predictive analytics for 15 years and it, it's, it's always hocus pocus. The problem is now people are gonna believe it.

So guys, I respectfully, I gotta disagree with you both. Oh, You know what? We're not getting a Tinder vibe out of this.

This, this is, this is Tinder all over again. And I'm gonna tell you a quick Tinder story. I had the opportunity once to talk with a guy named Sam, I think his last name was Mead, but Sam was his first name.

com. And I was at an inside Ignite in Iceland, and he gave a speech on how Tinder came to be. Hmm.

com. You got this. Two engineers came up with the sub, the, the, the supposition that human beings make a decision whether they think you're attractive, they're gonna like you, not like you, you're unattractive, they wanna have sex with you, they don't wanna have sex with you in a snap.

And that's all it takes. Looking at that one face, I swipe right or I swipe left. This went and it, it went against everything.

com was about everything you, Gina and you Ann spoke about, we've got this 23 point matrix that looks at their this and looks at their that and looks at their this. And you wanna know at the end of the day, who's all, both people make a decision, unfortunately, whether we like it or not, and it may not be the right thing to do. People make decisions in an instant when you give them a, a picture like that.

But it's it, I made the decision, Alan, that I liked you What? I said, I made the decision that I liked you when I met you. Well, I try.

Well, no, but that's the truth and tin up becoming more valuable than match. But go ahead Mike. That's Total b******t.

So there's a world of different, there's a, there's a world of difference between somebody thinks you're cute and somebody you wants to marry you, right? Because there's a lot of people that you think about this in your own, Oh, Mike, you've never lied about that before. How many people in your life have you met?

And, and, uh, we all have, we go, wow, they're cute. And then the next thing you know, you talk to them for about 15 minutes and you're like, wow, what an a*****e. That's what my husband on Telenet, I met my husband over Telenet in a private chat group and I did not, definitely Didn't see a picture of him.

No. And I didn't like him and I met him in person. This group met every week in person.

And then that was 25 years ago. So, Hey Allen, that's such congrat Congrat. Oh, sorry, go ahead.

I was just gonna say like that, that's incredible. I do wanna point out that talent was never private, but, uh, continue please. Well, private chat system, well, let's hope they didn't say anything.

We, how we know about it can Get too open, but I doubt any of it was Encrypted. Sheriff, Hey Elm, Who could be, who could be so shallow as to appoint people and and hire them to work for them just based on a image or a photo of them without ever meeting them who could possibly be done. Yeah, exactly.

Well that who, what, who says right outta central time out, time out. Who says right outta central casting? That's, that's what I meant.

That's what I did. That's, and that, that is you're, you're a hundred percent. Let's get Back to what we were talking about there, right?

So what we were talking, we ain't talking about Tinder. We ain't talking about looking for a hookup. We ain't talking about trying to find your marriage party.

We're talking about people putting their photos and their experience on LinkedIn to try to get a job. And this company and this, this research facility using this big five that most, this big five, um, photo model that's mostly used by the hookup companies, the dating companies. So, so people can swipe right or swipe left.

Okay, that's an application fine. That's fear, that's good. You decide what to do with that.

But not hiring somebody because of the way they look is dangerous. What's even more dangerous is one of the guys from the Wharton School that developed the model said that genetics and hormone exposure can influence both cranio facial structure and personality development. Additionally, how a person is perceived can reflect or reinforce certain traits people have found.

They, they found people with more symmetrical face tend to score higher on extra extroversion or baby face features are correlated with perceptions of warmth and naivity. That's all b******t. That's all phonology.

It has no place in modern day. It, it may not have a place, but it's, I'm telling you it's in our DNAI Hear what you're saying, but it might be, and it's not in our DNA, it's actually a social construct. Okay?

It's in our social contract. But That means, you know, a lot of white people will socialize back in the 18 hundreds to think all black people weren't even human. Is that something In the 18 hundreds, I think until about 50 or 60 years ago?

There's some today. Still today, right? That's what I'm saying, that that's all a social construct and it's what we have grown up with and what we were taught to see of people this should, we should not have an AI that is doing that kind of stuff and not giving people an opportunity because they fit on some swipe left.

Swipe. So Gina, again, don't, don't blame the mirror. Wait, so are you telling me supposed to look it in the mirror?

Are you telling me that excuse? Are you Telling me that the guy with the red hat on this show is not a long haul trucker? I Didn't know That's you, Fred.

Well, I appreciate that. Uh, so I think there's also, uh, an influence here on what, uh, so not just the choosing, but also the long-term effect of salary implications. Mm-hmm.

And, and the other things that go along with that. So that big five piece is interesting because, you know, they, they, they, people are screened on features they can't change or control, right? In essence.

But what what's interesting about that is if you look at the sort of the quintile of the, of the analysis here is it's pretty illustrative of the, of what the universe says or actually does today. So I, I think there's, I I'll give you another example guys who has their picture on their resume. I don't have a resume, but I've looked at when many, many resumes who has their picture on their resume real estate?

Why put your picture on the resume. Who puts a picture? I, I will tell you, sales do people No, no, I, I'm, I go on indeed.

Well, sometimes I go on indeed sometimes when we have to hire people. So I I, I have to go through a lot of resumes that get submitted and there's a good chunk of them that do have pictures on their resume. And I will tell you, I think it gives you, I don't know, it gives you a feeling about a person and it's just, it's Gina, it's just this, it's just a picture, right?

And, and it's a hiring decision If I'm in tech and the hiring and, and that we hold that true, and I wanna pull it back to the model too, right? The, the models that they're building. And can you build a model that can predict how someone's gonna behave in fi two years after you hire them?

Or how, um, outgoing they will be or how nice to work with will they be? That's really what we're talking about here. Can you tell that from a picture?

And it's very dangerous because that if, if I'm a woman in tech, I'm telling you right now, if there's three guys, I don't have much of a chance at all. I don't have a 25% chance Unless I'm looking at the resumes and Right. Or me make sure Y'all are different and y but we need everybody to get to where you are instead of depending on how, instead of how we've been socialized, being reinforced with a model that only gives us what it thinks you're gonna look for.

Oh, you only wanna look for the people that are gonna be high performing. These are the pictures that passed it in our big five model, right? And then that model all of a sudden becomes how everyone is socialized.

And we go back to having no women in tech, not having any women, doctors not having black or, Or everyone, you know, it's the Aryan kind of dream, right? You're all six feet with blonde hair and blue eyes Because all of it's a social construct best. And if we best train our AI best, if we train our AI to be part of that social construct and to, to magnify it and reinforce it, that will never change.

And everything we've worked so hard for that is right now, systematically being dis dismantled is going to have a faster way to be dismantled. We have to fight against the Shah. So what, here's what I would say.

I agree. Here's what I would say in the big five okay. Of personality traits.

It does not say race, it does not say gender, and it does not have an effect on the model. Now I'm gonna say that because the five traits are, I'm not disagreeing with you, Jim, I'm just sort stating the facts of what B five are. Um, openness, conscientiousness, right?

Extroversion, agreeableness and neuroticism. Um, gotta up if anybody looks at my pictures and sees mine neuroticism, but I mean, just about 10 years ago, um, quick story aside, just about 10 years ago, a very large financial institution asked a question of an org that I was with, and we were working on a pattern of finding financial analysts that were capable of becoming the best financial analyst analyst to work. We had a trend set of data, we built a model, we looked at, you know, quartile analysis and we were able to establish there are certain traits that in the first two years of your working, we will know whether or not you are going to be a top earner at this business.

And given that right, company made decisions based on that model to, you know, influence both the trajectory of that person at that job, but also whether or not they would continue to be employed, right? Great opportunity for somebody to get RIFed if they no longer see the value of them being a 10 year lifelong earner for a business. Those things have been going on AI aside, right?

For, you know, a long, long time. But I think you're right to be a hundred percent concerned about the implications of that 100%. What I think would be great is, is make suggestions about a couple of other personality traits that are features that should be added here that sort of help the bias that you wanna make sure you reflect on.

I I think you're bringing up very valid points. We should have a dialogue about how to add those. I think the thing is, it's not a top five list of personality traits because those personality traits don't get, don't account for how you have to act when you're discriminated against constantly in an industry.

They don't, in, in a society. They don't, they don't count up to that. You don't see all the work and the armor that people have to put on just to do their job and how much better they could do that job if all of that could be shed and they could be who they are.

So I, I think that, I think that there's a fundamental issue with the way that model was built. And I, and I will go on to say it is phonology because it it, if you can think, you can look at somebody, you might as well just fill the top of their head to see how they are in real life. I'm Gonna, I'm gonna, I'm gonna close this and hand it off to Alan, but I would remind everybody that everything that's a science isn't called the science and everything that's called the science.

Well, it isn't. So there you have it. So I guess what you're saying is, I don't know what you're saying, but Okay, I'll go with it.

Mike. It, it's so impactful. I didn't get what Yeah, it sounded like statement.

Hey, dang. What a great discussion today. It got it, got it, got deep there.

I love it. It's a very passionate panel by the way. Well little.

Yes. It's very, very engaged. AI says, based upon what they look, there's a high degree of passion, but we're, we're gonna come, we're gonna come back tomorrow with more gang.

Hey, I am on Shimmy says in about an hour and a half. So check that out. Do watch the world premier of our New Rock video and, uh, we just got so much going on.

We'll see you tomorrow on The Gang. This is Alan Schmo. We're out.