Departing OpenAI Leaders Warn about AGI

has an informative article titled Departing Open AI. Leader says no company is ready for AGI, and that's a really good start for our open AI discussion. So, first of all, AGI stands for artificial general intelligence. So it's an example of a term that on the surface sounds like it's fairly straightforward, but when you start to get into the weeds, you find out it's actually really vague and bor defined.

So generally AGI references a computer system that's capable of processing information in a way that's similar to how we humans think. That's the most vague way of defining it. Now. AGI does not necessarily mean superhuman intelligence, though that's typically where the conversation ends up going. An AGI system could theoretically be pretty darn stupid in fact, but still process

information in a way that mimics how we humans do it. Anyway, It's hard to define AGI because it's hard to justifine regular old intelligence, like how do you do that before you even get into artificial intelligence. But recently OpenAI dissolved its AGI readiness team, and that in turn follows a decision that the company had made earlier this year to nix its super alignment team. And both of these teams focused on ways to develop AI recis responsibly and to

identify and mitigate potential risks relating to AI. Turns out that's not necessarily a revenue driver, right. Putting limitations on your development process can be a real drag. Sure, I mean those limitations might be there in order to prevent AI implementations from doing really bad stuff, But come on, what are the odds that's going to happen? Oh, we don't know the odds. We actually we actually don't know

how likely it is that AI could mess things up. Well, then that's just as good as knowing there's no chance that things can go wrong? Am I right? Just a note, I am not right anyway, snark aside. This move is yet another nail in the coffin of the original mission

for open ai. Now, you might recall that way back when it was first founded, open ai was just a simple, humble, little country non profit company with the mission to conduct AI research in a way that stood to benefit human humanity the most with the least likely potential for bad consequences. I don't know why I decided to go to simple country lawyer mode for that, but anyway, later on, open ai spun off a so called capped for profit company, and this company would in turn still be governed by

the larger nonprofit organization. It's just that the for profit company would be operated for profit to help fund the research. And it didn't take very long for the for profit arm to eclipse the nonprofit side, and the for profit side essentially shed itself of the shackles of restrictions and caution. Why. Well, the big reason is that AI, as I have said many times, is wicked expensive. Open ai churns through money at a rate that is hard for me to conceive.

We're talking billions of dollars per year spent just to run operations, and I think they were on track to lose like five billion dollars this year before they ended up having a massive fundraising round. Nonprofits are just incapable of raising enough money fast enough to meet that kind of demand, Like, no one's going to pour money into a drain like that over and over, you know, perpetually,

so again and again. Open Ai has chosen the pathways that are most likely to lead to heaps of cash while downplaying the concerns of people within and outside the company about the potential dangerous decisions being made. In Robison's Peace in the Verge, she covers how the former senior advisor for AGI Readiness, Miles Brundage, warns that no company

in the world, including open Ai, is prepared for AGI. Further, he warned that governments around the world really need access to unbiased experts who are in the field of AI while they are actually formulating policies and regulations, because as it stands, the people who are extremely passionate about guiding policy tend to be folks like Sam Altman, the CEO of open Ai, and as you might imagine, these people

are not unbiased. It really creates a conflict of interest, right, It seems like a pretty safe bet that Altman's guidance to leaders around the world will focus on policies that will help or at the very least not hinder open Ai while potentially having a larger impact on the company's competition.

So Brundage is essentially saying like, the fact that open ai is making these decisions should cause some concern, and leaders around the world need to be savvy when they are consulting with various experts and making sure that the advice they receive isn't being guided by self interest. Another good read is an opinion piece from John Herman of Intelligencer.

I guess you could argue that it's not really an opinion piece, but I feel like Herman injects a lot of his own opinions in it, and I am pretty much on the same page as Herman. Well, anyway, the article is titled Microsoft has an open Ai problem. Now. I mentioned in an earlier tech Stuff News episode that Microsoft has seemed to have kind of a rethink when it comes to its relationship with open AI. In Facebook terms, I think we would say that the current relationship status

would be at it's complicated. So Microsoft has invested billions of dollars, like nearly fourteen billion bucks in open Ai so far, and Herman points out that the agreements between the two companies happened relatively quickly and without a whole lot of thought as to how that process of funding would actually work, which seems like pretty loosey goosey to me. And now Microsoft and o AI's partnership hinges on if

open ai develops AGI. So essentially the agreement is that this partnership between the two companies will dissolve if open Ai creates an AGI product of some sort, allegedly because open ai has a real, deep concern that a company like Microsoft might misuse such a powerful and potentially destructive tool as artificial general intelligence. And sure, I think that's possible, Like I think if Microsoft had access to AGI, there

could be some pretty negative consequences. But it's not like I have an enormous amount of faith in open Ai either. It's not like I look at them and think, oh no, I want them to have the keys. I don't want anyone to have the keys. I don't want it to be a thing. But I guess that's not an option. But anyway, when you're talking about a company like open ai,

which has actively been dismantling the systems. But in place to ensure safe development or artificial intelligence, you can't really be advocating that they're the ones who should be, you know, the stewards of AGI. Herman cites a Times article that points out that open AI's board of directors actually has the authority to determine what AGI is and when, if ever, open Ai achieves it. So, at least hypothetically, even if outside parties would disagree with the definition that open AI's

board comes up with, that wouldn't matter. If open Ai said, oh, we did it, we created AGI, and if literally no one else in the world said that's definitely AGI, it would be enough for open ai to sever its relationship with Microsoft. Now, the implication here is that this isn't really some sort of protection for open ai to make sure that AGI is not unleashed upon a world and

like the next build of Windows or something. It's really more about creating kind of a switch, a kill switch, a bargaining tool with Microsoft, so that when it comes to renegotiating with Microsoft, open Ai is able to essentially say, hey, if you don't give us more money, we'll say we made AGI, and then that's it. We walk and you can't do anything about it. So that's a possibility. I don't know if it's a reality, but it's possible. It's almost enough to make one jaded, isn't it. Anyway? The

article's well worth a read again. It's Microsoft has an open AI problem, and it's an intelligencer by John Herman. Okay, I got a lot more news to go through, but first, let's take a quick break to thank our sponsors. All Right, we're back, and David E. Sanger of The New York Times has a piece titled Biden Administration outlines government guardrails for AI tools, and that brings us back around to those government backed policies relating to the development and deployment

of artificial intelligence. So essentially, these guardrails lay out the scenarios in which using AI would be appropriate and allowed, and the scenarios in which it really absolutely should not be allowed. So, for example, using AI to detect and deter cybersecurity threats, that's pretty okay. Using AI to develop a new generation of fully autonomous weaponry, that one falls

into the no no category. So the guardrails put up a lot of responsibility on individual departments and agencies to conduct their own reviews and determine in which use cases AI might be appropriate and in which ones it would not be. There's a lot more to the guardrails. There's like thirty eight pages of material that's available for the public to read. It includes a call to have the US attract more AI experts to the United States rather than have them work for a rival nation like China

or Russia. But at least some of the information about these guardrails is classified, so I have no idea what could be in those Also, as Anger points out, the efficacy of these decisions is somewhat unknown since the United States is currently in an election year and the next president might just make all these guardrails go away, so we don't really know if this is going to matter at all in the long run. Fun times. The Attorney General for the state of Montana has filed a new

lawsuit against TikTok. It's a new lawsuit, but it's an old tune. They're accusing the company of purposefully promoting addictive and harmful content to young users through the recommendation algorithm. The Attorney General alleges that TikTok essentially has lied about the nature of promoted content in violation of the Montana

Consumer Protection Act. So essentially, the attorney general says that TikTok knew that users, including young users, would encounter content that's just unsuitable for people who could be as young as thirteen years old, but pretended that it has safeguards in place to prevent such young users from encountering, you know, mature or disturbing content. An unnamed spokesperson for TikTok disputes the allegations and says they are quote inaccurate and misleading.

In the quote, you might recall that previously, the state of Montana actually banned TikTok entirely, but then a judge later overturned that ban on the grounds that it violated the First Amendment, that's the right to free speech. Several other states have similarly filed charges against TikTok with similar accusations. Essentially, it boils down to TikTok's algorithm promoting inappropriate and sometimes

outright dangerous content to impressionable young users. Meanwhile, of course, TikTok is facing down a well a ticking clock situation here in the United States. Currently, the app faces the possibility of a nationwide ban unless its Chinese parent company Bite Dance, divests itself of TikTok. So it might be that by this time next year, lawsuits against TikTok, at least here in the United States, will no longer really

be relevant, or at least not as straightforward. Speaking of bans and social media, Norway is increasing the minimum age allowed for folks to use social media. So previously the minimum age in Norway was thirteen, but now it's going to be increased to fifteen years old out of a concern that social media can have a harmful impact on young users. Of course, making that law is one thing,

but enforcing it is entirely something else. I mean, as Miranda Bryant of The Guardian reports, the Norwegian Media Authority found that more than half of nine year olds in Norway are already on social media, and as you go up and age the percentage increases, so already there are tons of kids in Norway who are actually using social media when they're underage, like they're below the age of thirteen,

even though the law says they ain't supposed to. So apparently part of the plan is to change the way that social media companies verify the age of users in Norway, so that it's not a trivial task. Just fib about one's own age admidly, that's a pretty low bar. I mean, if I were a kid, then I was trying to make an account on a social media platform and the age verification was just asking me, Hey, when when's your birthday? I could probably do some pretty quick math and fudge

that date enough so that it would let me in. Anyway, while I have my doubts about how effective this new policy is going to be, I absolutely do feel empathy for those who want to shield kids from the ravages of the algorithm, because there are days when the various recommendation algorithms that I encounter make me want to go

and hide in the woods, possibly in Norway. Alison Morrow of CNN has a piece titled almost anything goes on social media as long as it doesn't make billionaires feel even a little bit unsafe, And yeah, billionaires live by different rules than everybody else. In fact, they often are

the ones who decide what the rules are. So while people like Elon Musk are extremely vocal about the importance and sanctity of free speech, they're also quick to say no, not like that if the free speech includes stuff that is potentially harmful to them. So the case Morrow is focusing on that of Jack Sweeney. He's a college student who has created numerous social media profiles on different platforms that are dedicated to tracking specific billionaire private jet routes, which,

by the way, is publicly available data. Sweeney is not hacking into the mainframe or anything like that. All he's doing is just scraping data from public sources and posting it to social Anyone could do this, but various platforms have removed Sweeney's accounts, often without giving him any forewarning or explanation, and when asked, they usually give a pretty hand wavy explanation that the information poses a risk to

certain individuals. But I mean, if you happen to know that Elon Musk flew into your town, you might track him down an assault him or something, or maybe you know, just yell at him because your cyber truck don't go no, mo I guess that's the fear, Like, how do you know where the person specifically is? Who just know where they're jet landed? Anyway, Marrow points out that the various platforms seem to have very little problem with hosting stuff that could be extremely harmful to the rest of us,

whether that's misinformation or hate speech or whatever. That stuff, well, that stuff's not likely to touch the billionaires, so there's no reason to worry about it. Let that stay up on those profiles. We all just have to fend for ourselves. But publicly available information about billionaires, now you're talking about dangerous information. Anyway, read the article. You already can tell

what my opinion is about all this. I'll just Quotemorrow directly for this little bit, because you know we're aligned on this quote. It's just not clear that Meta cares as much about its users' privacy and well being as it does about Zuckerberg's end quote Preachmorrow. The US Federal Trade Commission created new rules this past summer that gives the agency the power to find people who post, buy, or sell fake reviews online. I'm sure you're aware that

fake reviews are a real problem. Once upon a time, I think the average person could go to an online marketplace like Amazon and lean on user reviews when making purchasing decisions. But over time, companies have found ways to incentivize folks to write good reviews. Sometimes it's fairly innocent. You know, you might have a little card in the box that your product came in saying something like, hey,

if you like this, please consider writing a review. That seems pretty innocuous, but it can quickly get more murky, like you might get one that says, hey, hey, if you write a five star review, we'll give you a coupon for some other product or service. And sometimes it just gets downright overtly wrong, like write a positive review for this thing that you might not even use or own,

and we will pay you money. Well, the new rules that the FTC passed are now in effect, and essentially, if you are caught trading in fake reviews, you could face a fine of up to fifty one, seven hundred and forty four bucks, and that includes AI generated reviews. If they find that you've been doing that, you could be paying some pretty hefty cash. Now, on top of that, companies are not allowed to suppress negative reviews. They're also

not allowed to promote positive reviews that they suspect are fake. Now, how the FTC plans to go about proving that a company was aware that a review was fake, I'm sure that'll be challenging. On a case by case basis. But the rule is a hefty piece of writing. It is one hundred and sixty three pages long, and no, I have not read the whole thing, so I'm not prepared to give my full opinion as to whether or not this is a good idea or a bad idea, or

a good response or an ineffective response. I need to be able to read the whole thing before I come up with that. So it looks like I've got my reading material for my upcoming flights. Yay, okay. I got a couple of articles I want to recommend for your reading pleasure. One is a John Broadcin article on Ours Technica and it's titled Cable companies ask Fifth Circuit to block FTC's click to cancel rule. I think that's the

least surprising headline I've ever read. Obviously, cable companies don't want there to be a straight forward click to cancel route for users. But yeah, read the article because it kind of explains what the situation is and how the specific courts that were chosen could end up having a massive impact on this potential rule. And finally, I have one other reading recommendation. It's an article by John Walker of Kotaku. It is titled Roadblocks's new child safety changes

reveal how dangerous it's been for years. So it's an eye opening piece that applauds these safety changes, but it does ask, hey, how is it that this wasn't already a thing? So check that out as well. I hope all you out there are doing well, and I will talk to you again really soon. Tech Stuff is an iHeartRadio production. For more podcasts from iHeartRadio, visit the iHeartRadio app Apple Podcasts, where whoever you listen to your favorite shows. M