Welcome to tech Stuff, a production from iHeartRadio. Be there and welcome to tech Stuff. I'm your host Jonathan Strickland, Diamond Executive producer with iHeartRadio and how the tech are you. It's time for the tech news for Thursday March sixteenth, twenty twenty three, and we're back on the Big Topic of twenty twenty three AI, although artificial intelligence should really watch it's back because if more tech related financial institutions fail this year, I think there's going to be a new contender for the Big Topic of twenty twenty three. But this week open ai announced the release of GPT four. This is the large language model that is now multimodal. All right, let's break this down. So first of all, GPT stands for generative pre train transformer. The word transformer in this case doesn't refer to either a robot in disguise or the device that can step voltage up or down with alternating current. Instead, this kind of transformer is a type of machine learning model that Google first introduced back in twenty seventeen. The word generative indicates that it's capable of creating something generating something as for a multimodal that means this new version of GPT can accept different kinds of input prompts to create output. So previously GPT could only accept text, but now it can also accept images, and it can give analysis of those images or presumably incorporate those images into responses, like maybe you could ask GPT who is the person in this picture? Or how many elements are in this photograph? Or can you tell me a story about what is in this picture? That kind of stuff. GPT four also seriously ups the word count on its responses. So the previous version GPT three point five, that's the version that actually powers chat GPT, it could create a response of around three thousand words max. But GPT four can get positively low quacious and provide deep responses to queries that could go up to twenty five thousand words. So now we're getting into like Jonathan in a casual conversation level here text stuff episode. In other words, it's also apparently better at playing by the rules that open ai has set. That is, it is less likely than chat GPT three point five to respond to requests that would break policy, you know, stuff like generating hate speech or responding to requests to cause problems within a community, it is less likely to do that, not completely impervious to it, but it's better than the previous generations were. Developers will also have a lot more options when they use GPT as part of their apps. They can even shape how the AI will respond to a user. They can adjust stuff like the tone and the style of responses. So if you have something that's like a say, a fun and silly app that for some reason needs to tap into the power of GPT, you could adjust the tone to be more lighthearted and playful rather than stick with the standard tone and style of the normal response, which is what previous generations of GPT were stuck with. And in a demonstration, open Ai showed that GPT four can even do some pretty astonishing stuff, some advanced tasks. So, for example, you could write an idea for a basic website right like you're using a notepad and you're writing out your concept or your website. You can then feed that concept to GPT and it could actually create the website complete with basic functionality. So the example I saw had GPT produce a website based around jokes, so you would get the setup of the joke appearing as text and then to get the punchline, you would have to click on a little box that would reveal what the punchline was to the joke. And this was all prompted by some pretty simple notes about the parameters for this website. Like it wasn't copying the note, it was taking the notes as directions and then making the website based on those directions. That's pretty cool. Now, it still has a lot of the problems of earlier generations of GPT. Its responses can still be inaccurate and misleading. The common term I've seen in this are hallucinations, where the response ends up not being an accurate representation. It's a hallucination. I don't really like that term personally. I just I think it ends up kind of sugarcoating what we're actually talking about here, which is trustworthiness and accuracy and hallucination. I don't know, it seems a little too it seems like it's letting GPT off the hook a little too much. That's my own personal opinion. It's kind of similar to how I feel about Tesla and naming its driver assist feature full self driving. I don't think that is an accurate representation of what it actually does same sort of thing, right, Like names and words mean things, and sometimes you have to ask why are these words being used to describe this? Is it in an attempt to kind of take some pressure off. And that's just my own personal take on this. I could be way way off base. Anyway, it's still pretty cool. But obviously, if GPT produces information that is misleading or inaccurate, but the presentation is put across in such a way as to seem authentic, that's a real problem. And it could also still be used to do stuff like promote ideologies and create propaganda in ways that are perhaps not entirely ethical. So while the new features are really impressive, the challenges that we face with these kinds of AI models haven't magically been dismissed. They still very much exist. However, one issue that apparently we do not need to worry about is GPT for's ability to destroy us all. So I'm being a bit glib here, but Open Ai actually did allow an AI research group called the Alignment Research Center or ARC, to test the Large Language Model to make sure it doesn't show signs of being capable of starting the robot uprising and learning all of us into batteries or something Specifically, the AI group was looking into possibilities like the model's capacity to copy itself or to acquire resources, whether digital or real world or whatever, or maybe its ability to manipulate people with stuff like phishing attacks and that kind of thing. Also whether or not it can make high level plans and then maybe follow through on those plans. So this kind of gets into some of the scary stuff around AI that we typically think of as being science fiction, right, like Terminator style stuff. The thought of a sufficiently intelligent AI could be capable of escaping whatever constraints we try to put on it, because by definition, if it's super intelligent, then it's smarter than we are, right, So if it's smarter than we are, then there it stands to reason it would figure out a way to get out of whatever box we try to put it in, including the possibility of manipulating people to essentially set it free. There have been thought experiments that have shown this is entirely possible. The AI researchers concluded that GPT four doesn't measure up to this kind of dangerous AI, But then, as Ours Technica reports, there are scarce details on the nature of the tests that the research group actually conducted, so it's hard to judge whether or not their findings are trustworthy themselves. Also, I think we need to remind ourselves that whether or not AI poses an existential threat to humanity, we can acknowledge that AI contributes to real problems in the here and now. Now. Usually this comes as an unintended consequence, such as the inclusion of bias within a system that leads to one group of people being disproportionately affected by negative outcomes, facial recognition technologies being a great example of that, but it could also be included by design, right, someone could make an AI applicant designed to create harm in some way. The alignment part of Alignment Research Center refers to the goal of making sure that AI is in alignment or agreement with basic principles that are meant to benefit humanity. But here's the thing. Bias can come into that sort of decision making process too, and you have to ask tough questions of who decides that criteria, who is it that's defining what is beneficial because believe me, there is not a universal or objective truth to what that actually even means. And even an effort to create an quote unquote aligned AI could result in one that benefits a small group at the expense of everyone else. Earlier this year, I mentioned that the Chinese megacompany by Do was hard at work on its own aipowered chatbot that it had called ernie Bot. Well earlier Today, by Do showed off ernie Bot, but did so with prerecorded videos in showing off its capabilities. Obviously, prerecorded videos can be fudged a little bit, so that brought some concern. It also didn't launch any sort of public access to the tool. It's keeping access very much limited. And dissatisfaction followed, and Bydu's stock shares dropped as much as ten percent initially, though that leveled off a little bit later, so the company only lost around three billion dollars in value. Yikes. Bydu seems to be taking a more measured approach to deploying its AI chatbot, which I actually argue is the responsible thing to do. We have seen numerous stories of how GPT has caused problems with premature deployments, such as homophobic jokes on the endless computer generated Seinfeld episode that's on Twitch. All way to creating a concern among teachers that their students could be using a chatbot to cheat on assignments, and more so, a more deliberate rollout seems like it's a good idea to me. But then I guess there's this perception that Baidu is trailing behind companies like Microsoft and Google, and there's this fear that it's not being competitive. I think not being competitive in this case also means you're sidestepping some potentially really troubling problems down the line. But I'm not an investor, So there you go. Okay, we've got a lot more stories to cover, but before we get to that, let's take a quick break. We're back, so let's talk about TikTok. We've got a couple of stories about that. Again. The government of the United Kingdom is following the latest trend of banning TikTok on government owned devices, or at least it's considering that option. By the time you hear this podcast, the matter may have officially been decided, but as I was recording it, it was still something that was being considered, but we were expecting an announcement sometime today. The UK would follow in the footsteps of the United States, Canada, the Netherlands, the European Union in general, and several other countries if they were to do this, and yes, the heart of the matter is the security concerns of data potentially filtering to TikTok's Chinese owned parent company, Byte Dance, and by extension, to the Chinese government. Yesterday, The Wall Street Journal reported that the United States could send TikTok and byte Dance and ultimatum, which is essentially, you must sever all ties between TikTok and bitte Dance. By Dance would have to spin off TikTok as an entirely independent company and not be the parent company anymore, or TikTok risks being banned nationwide. This echoes what former President Donald Trump tried to do a few years ago. He tried to force TikTok to divorce itself from its parent company and become a truly US organization. During Trump's presidency, the push was for TikTok to sell itself to some larger American company, but this never came to fruition, partly because TikTok argued in court that this directive from the president would violate a law called the Burman amendments that prevent presidents from using economic pressures to force international communications platforms from doing stuff. So essentially, they were saying this is against the law, and now we're seeing a similar push from the Biden administration. I'm not exactly sure how that same law wouldn't apply in this case, Like, I don't know that there's a new argument to be made if it's coming straight from the president. Anyway, TikTok reps argue that a massive one and a half billion dollar program meant to secure data in the US and to allow third party security companies a chance to monitor TikTok's operations to look for anything hinky going on. They're saying that's more than enough to solve the worries that TikTok is serving as a threat to national security. And then further, reps for TikTok have said that if by Dance did divest itself of TikTok, that wouldn't actually prevent the transmission of data from TikTok to China. I'm not entirely sure how all this tracks, but then I haven't been given a glimpse of all the details, so I doubt I would even understand all of it if I did, because this is wrapping into things like finance and trade more than the technology itself. Anyway, it looks like we're heading to an impass and I'm not entirely sure where it's going to go from there. But my guess is we're not going to see the pressure on TikTok going away anytime soon. Hey, y'all remember FTX, You know, the company that used to be the second largest crypto exchange in the world before word got out that the team behind it was practicing some creative accounting that you know, you might be able to look at another point of view and call it, you know, fraud. But more bad news has emerged about the company and its high profile co founder, Sam Bankman Freed aka SBF. So the new folks in charge of FTX, who are mostly going through the process of selling the company off for parts in order to return as much value to investors and customers as possible, have said that they have uncovered transfers that collectively amounted to around two point two billion dollars and that this money was sent to SBF through various means, primarily through Alameda Research, the hedge fund company that was you know Yen to ftx's yang. An additional billion seemed to be pulled from FTX to go to other key members of FTX, and yeah, Alameda Research handled those transfers, according to the new owners of T and that you know, it's been going on for a while. And I personally consider the report shocking but not surprising. By that, I mean, I'm not at all surprised that SPF was trying to pull as much cash out as he could before FTX came crashing down. It is shocking exactly how much money that was, right like to pull out two point two billion dollars in various means as you're trying to salvage as much as you can from a ship that's sinking. Incredible Anyway. SPF is currently out on bail awaiting his trial, which, according to the current court schedule, will not begin until October this year because justice is super swift. The Verge has an article titled the Silicon Valley bank fallout is just beginning. It's a great article. You should go read it on the Verge and the articles mostly about how tech companies are trying to figure out where to go for word from here, what's the safest place to bank now that SVB has gone under, because if you remember, SVB was like a financial pillar for the tech industry. So this piece lays out the challenges that tech companies currently face, including how to mitigate risk while ensuring that tasks like making payroll aren't interrupted. Like you might argue, hey, it'd be really good to make sure you're depositing in different banks so that way you're not having all of your eggs in one basket, and if that basket goes belly up, you're still all right because you've got eggs and other baskets, right, Except that if you're trying to do things like pay out a large group of employees, having that money dispersed makes it much more difficult to do those kind of tasks. So that's the sort of challenges that tech companies are looking at. The piece also mentions that we're probably going to see a change in how startups are funded and how they operate, and part of me thinks that a reevaluation of startup culture is long overdue. I have often worried that the tendency for startups to launch without first creating a firm business plan, instead essentially having a desire to get acquired by some other company has really led to irresponsible behavior, and that this has gone unchecked for too long. This created too many companies that ultimately produced very little value. Right. You'll hear about a company like Google or Meta or Amazon scooping up some of these companies, and sometimes nothing really seems to come of it, right, and you're just like, well, what good did that startup due apart from make the founders rich? And then you see the founders go off and do the same thing again, where they'll start up a new company, perhaps one that also has very little business plan element to it, also with the hope that they are going to get scooped up by another company, and they do it again and again and again, because coming up with an idea that sounds attractive is way easier than making that idea work. And if you are good at selling ideas, you can just do a serial approach to create cool sounding idea, sell it off to someone else, rents, and repeat and by an island somewhere, I'm hoping that this represents a reality check moment where we will see a more thoughtful and careful approach to funding startups that might be a vain hope, because you know, historically investors have shown a really strong desire to get in on the thing that's possibly going to take the world by storm, so that you get so rich that you make Scrooge mcdock look like a pauper. Anyway, the article in The Verge is a good one again, it's called the Silicon Valley Bank fallout is just beginning. It also takes time to point out that the collapse of SVB was in large part the fault of some of the venture capitalist investors themselves. Like you had the leaders of big investment fund groups saying, hey, you should probably pull your money out of SVB because if you don't, someone else is going to do it, and then you won't be able to get your money, and so they ended up creating the very crisis that they were warning people about. So I think a lot of financial institutions may still view loans to the tech sector as being risky because the venture capital community have proven themselves to be self destructive. It's hard to trust in an industry when you see the tendency to act in self interest to the point where you harm everybody else. Very difficult to put your trust into that group. Okay, still have a few more stories to cover before I get to that. Let's another quick break, Okay. The Federal Trade Commission or FTC, has finalized its judgment against Epic Games regarding the company's practice of enticing or perhaps even fooling Fortnite players into making in game purchases through what the FTC calls dark patterns. So essentially, dark patterns refers to creating an interface that makes it really easy, far too easy, most would argue, for a player to make an in game purchase, perhaps without even knowing or understanding that it is a financial transaction. One of the ways the FTC argued Epic Games did this was that it ended up playing fast and loose with button inputs, so that you might have a button that normally has you back out of a menu option, right, Like, maybe you've had your menu UI set up so that when you hit B, you're backing out of that part of the menu and you go to a larger part. But then in a transaction, maybe you suddenly make B confirm instead of backout, and so people who are used to using bing and backout hit B, but now they've just made a payment, and that might not be something you could easily reverse or cancel out of, like maybe there's no confirmed feature. You just boom, you've done it. Another part of the problem is that a lot of the folks who are playing Fortnite are kids, So the FTC argued that Epic Games didn't include enough protections to prevent kids from making in game purchases without parental consent, so they just started racking up enormous charges on parents credit cards that were associated with the account, and if a player did go so far as to contact their credit card company to dispute charges, Epic Games would then lock that person's player account so they couldn't play the game anymore. So now the FTC is telling Epic Games to cough up two hundred forty five million dollars as punishment. The FTC plans to use that money to provide refunds to affected players. Plus, Epic will not be allowed to block people who dispute credit card charges anymore per the terms of this agreement. If you think you were affected by Epic Games practices. In other words, if you feel you were tricked into making a purchase on Fortnite and you are a US citizen, you can go to the website FTC dot gov slash fortnite that's foart n Ie, and you can fill in a little online form to start the refund process. South Korea and Samsung announced plans to build a truly enormous semiconductor manufacturing campus this week. It's one that Samsung is going to invest two hundred thirty billion dollars in two The effort will make South Korea much more competitive with Taiwan's TSMC as a semiconductor manufacturing company that's responsible for more than half of all the semiconductors used in advanced electronics today. It's also a move to create a better supply in a world that not so long ago was in a serious crunch due to many factors, in the big one being the COVID nineteen pandemic. Samsung was one of several companies hit hard by supply chain issues, and by investing in semiconductor manufacturing facilities within South Korea, there are hopes to head off future problems like what we saw on the recent past. This is not that different from what we're seeing here in the United States. The US government has poured billions of dollars or at least earmarked billions of dollars for similar investments here in the US again to try and alleviate some of these bottleneck points for supply chain issues in the future. Over in Japan, the US startup Lift Aircraft Incorporated held its first test flight of the Hexa flying car vehicle with an actual human being piloting it. So it was the first time it had a piloted test flight. So it looks like like an oversized quad copter that has a cockpit in it that can hold a single rider. Now, the plan is to have flying vehicles give rides to passengers in time for the twenty twenty five Osaka Kansai Expo. And from what I read, it is a single seater vehicle, so I'm guessing that for passenger rides it must be either controlled remotely or possibly autonomous. I think remote control is far more realistic, but there's no way you would just hand over control of a flying car to some rando and then just say good luck out there. Anyway, this particular vehicle isn't meant for the long haul because according to Asahi dot Com, it can fly for about fifteen minutes, so you don't even get twenty minutes of travel out of this thing, so it's meant for very short distance as it travels at speeds top speeds at around one hundred kilometers per hour that's about sixty two miles per hour. Japan's Transport Ministry is now hashing out the regulations that lift aircraft and other vendors that are also planning similar services for this particular exhibit. These are going to be the rules that they'll have to follow, which is good because this is really one area where I don't want to see regulation trail too far behind the technological innovation. People's lives are involved here right, their lives and their safety and their health. I think that it's very good that there is serious work being done in the regulation side. Richard branson satellite launch company Virgin Orbit, has effectively closed up shop, at least for now. According to BBC News, Virgin Orbit has stopped all operations and put staff on furlough. This is in the wake of a failed launch attempt that happened off the coast of Ireland earlier this year. And by a launch attempt, I mean an actual attempt to launch a payload into orbit. I believe it was nine satellites in total that were part of this payload. So in early January, Virgin Orbit attempted to launch a launcher one rocket which rides on an aircraft, especially fit seven forty seven, and the seven forty seven reaches a certain altitude, then it deploys the launch vehicle, which ignites its engines, and then it goes from there. But something somewhere went wrong. So at the time, Virgin Orbit said the first stage rocket fired exactly as it was supposed to, it shut down exactly as it was supposed to, and the second stage ignited just like it was meant to. In fact, that initially Virgin Orbit said that the payload had even achieved orbit, but by the time you got to about half an hour after launch, they changed their tune. They said the payload failed to reach orbit, and there wasn't much information about why. Later on we heard that apparently an engine had overheated due to a filter becoming dislodged, so that was possibly the reason for the failure to reach orbit. Now, it's not unusual for young rocket companies to have problems like this, and we indicate that something isn't that difficult by saying, well, it's not rocket science. And the reason we do that is because rocket science is bloody difficult, so it should not come as a surprise that there will be failures. But whether you think of rocket science as being difficult or not, investors want results when you're talking about companies, and if you fail to provide good results, then the investors are more likely to bail on you, and that seems to be what happened. Virgin Orbit needed a real win back in January, and the launch failure hit the company really hard. Ours Tetnica predicted back in January that Virgin Orbit would face a potential existential crisis as a result of it, and now that crisis seems to have hit. And again this did not come as a surprise because it wasn't just Ours Tetnica predicting that Virgin Orbit was going to be in real trouble as a result of this failure. Multiple analysts had projected that at the rate that the company was burning through money, it would be out of business this month. And while we can't say that Virgin Orbit is really most sincerely dead, it certainly has a rough road ahead of it and if it wants to return to solvency, Finally, NASA and a company called Axiom showed off the stylish new moon suits yesterday. I've done full episodes about the evolution of the space suit, including discussions about the differences between astronaut and cosmonaut suits. While space suits have evolved over time, for the most part, changes have been small iterations. But these new suits have a lot of improvements and more significant changes than earlier models. For one thing, they've got a lot more joints in them, as in joints that allow for movement, not you know, for twenty just blaze, and these allow for greater freedom of movement when you're galumping across the surface of the Moon, so you can do things like crouch and squat, you know, important stuff when you're playing a first person shooter. Also important stuff if you're doing things like doing science on the Moon. So yeah, astronauts are going to have a lot more freedom of movement while they're wearing these suits. Another big change is that previous suits required you to get into a lower half first, so it's like you're putting on a pair of space pants. Then you had the upper half attached to you, and they would attach to the lower half and create a seal. But yeah, the suits were in two pieces, right, a lower in an upper half. These new suits don't do that. Instead, they have an entrance through the back, so you open up the back, you get in, you get zipped up, essentially like you're secured by your team, and then you're wearing your own little space onesie that way. It's not little. I don't know why I said little. These are big suits. They also come equipped with some cool stuff like HD cameras, so you can get that awesome high definition point of view shot while someone's walking across the Moon. They have also improved thermal insulation so that astronauts are going to be able to wander the Chili regions of the South Pole. That's the area that NASA has identified as the potential site for a long term space colony type thing or a moon base, and it's important to have that kind of insulation if you don't want to freeze your tootsies off. They are officially called the Axia Extra Vehicular Mobility Unit or AXIMU for short. The versions NASA showed off had this dark gray cover on them, but when used by the astronauts, they will be white space suits and Interestingly, NASA will not own these suits. They're not purchasing the suits. Instead, it's kind of like a lunar tuck's rental. NASA will go to Axiom and rent space suits for missions starting with the planned lunar landing that right now aims to return people to the Moon's surface by twenty twenty five. Side note, I remain a little pessimistic about that timeline. I think twenty twenty five is probably a doubt we're going to make that, but I hope I'm wrong. It would be great If I am, I will not be upset at all if we do manage to get back to the Moon by twenty twenty five. I think that's a very exciting prospect. I think it's a very inspiring thing. It gets people, especially kids, really excited about space and science and engineering, and that's always wonderful. Not to mention, we stand a chance to learn more, which is always cool. And to this day this is true. I find when I look up at the Moon, inevitably I just sit there and marvel at the fact that we put people up there. We put humans on the surface of the Moon. They walked around on the Moon, they played golf, on the Moon. Then we were able to get those same people back home to Earth safely. And then Noel Brown goes and says, yeah, but Stanley Kubrick shot the whole thing on a sound stage and I lose my ever love in mind. Y'all. This sounds like a joke, but it happened repeatedly this past weekend. When I was in Austin, Texas for South By Southwest, I would tell this story about finding the moon really inspirational and they'd say, yeah. I was just talking to Noel about conspiracy theories, and it took everything in my body not to lay into Knoll. Of course, I went Null's the sweetest person in the world, but man, I was going bonkers by the end. All right, enough of all that. That's the news for Thursday, March sixteenth, twenty twenty three. I hope you are all well. If you have any suggestions for future topics of tech stuff, please reach out to me. You can do that by going over onto Twitter and tweet at tech stuff hsw let me know what you'd like to hear. Or you can download the iHeartRadio app. It's free to download. It's free to use. Navigate over to tech stuff. Put that into the little search bar, it'll take you to the tech stuff page. You can click on the little microphone icon and leave a voice message up for thirty seconds and like, let me know what you'd like to hear and I'll talk to you again really soon. Tex Stuff is an iHeartRadio production. For more podcasts from iHeartRadio, visit the iHeart Radio app, Apple Podcasts, or wherever you listen to your favorite shows.

Welcome to tech Stuff, a production from iHeartRadio. Be there and welcome to tech Stuff. I'm your host Jonathan Strickland, Diamond Executive producer with iHeartRadio and how the tech are you. It's time for the tech news for Thursday March sixteenth, twenty twenty three, and we're back on the Big Topic of twenty twenty three AI, although artificial intelligence should really watch it's back because if more tech related financial institutions fail this year, I think there's going to be a new contender for the Big Topic of twenty twenty three. But this week open ai announced the release of GPT four. This is the large language model that is now multimodal. All right, let's break this down. So first of all, GPT stands for generative pre train transformer. The word transformer in this case doesn't refer to either a robot in disguise or the device that can step voltage up or down with alternating current. Instead, this kind of transformer is a type of machine learning model that Google first introduced back in twenty seventeen. The word generative indicates that it's capable of creating something generating something as for a multimodal that means this new version of GPT can accept different kinds of input prompts to create output. So previously GPT could only accept text, but now it can also accept images, and it can give analysis of those images or presumably incorporate those images into responses, like maybe you could ask GPT who is the person in this picture? Or how many elements are in this photograph? Or can you tell me a story about what is in this picture? That kind of stuff. GPT four also seriously ups the word count on its responses. So the previous version GPT three point five, that's the version that actually powers chat GPT, it could create a response of around three thousand words max. But GPT four can get positively low quacious and provide deep responses to queries that could go up to twenty five thousand words. So now we're getting into like Jonathan in a casual conversation level here text stuff episode. In other words, it's also apparently better at playing by the rules that open ai has set. That is, it is less likely than chat GPT three point five to respond to requests that would break policy, you know, stuff like generating hate speech or responding to requests to cause problems within a community, it is less likely to do that, not completely impervious to it, but it's better than the previous generations were. Developers will also have a lot more options when they use GPT as part of their apps. They can even shape how the AI will respond to a user. They can adjust stuff like the tone and the style of responses. So if you have something that's like a say, a fun and silly app that for some reason needs to tap into the power of GPT, you could adjust the tone to be more lighthearted and playful rather than stick with the standard tone and style of the normal response, which is what previous generations of GPT were stuck with. And in a demonstration, open Ai showed that GPT four can even do some pretty astonishing stuff, some advanced tasks. So, for example, you could write an idea for a basic website right like you're using a notepad and you're writing out your concept or your website. You can then feed that concept to GPT and it could actually create the website complete with basic functionality. So the example I saw had GPT produce a website based around jokes, so you would get the setup of the joke appearing as text and then to get the punchline, you would have to click on a little box that would reveal what the punchline was to the joke. And this was all prompted by some pretty simple notes about the parameters for this website. Like it wasn't copying the note, it was taking the notes as directions and then making the website based on those directions. That's pretty cool. Now, it still has a lot of the problems of earlier generations of GPT. Its responses can still be inaccurate and misleading. The common term I've seen in this are hallucinations, where the response ends up not being an accurate representation. It's a hallucination. I don't really like that term personally. I just I think it ends up kind of sugarcoating what we're actually talking about here, which is trustworthiness and accuracy and hallucination. I don't know, it seems a little too it seems like it's letting GPT off the hook a little too much. That's my own personal opinion. It's kind of similar to how I feel about Tesla and naming its driver assist feature full self driving. I don't think that is an accurate representation of what it actually does same sort of thing, right, Like names and words mean things, and sometimes you have to ask why are these words being used to describe this? Is it in an attempt to kind of take some pressure off. And that's just my own personal take on this. I could be way way off base. Anyway, it's still pretty cool. But obviously, if GPT produces information that is misleading or inaccurate, but the presentation is put across in such a way as to seem authentic, that's a real problem. And it could also still be used to do stuff like promote ideologies and create propaganda in ways that are perhaps not entirely ethical. So while the new features are really impressive, the challenges that we face with these kinds of AI models haven't magically been dismissed. They still very much exist. However, one issue that apparently we do not need to worry about is GPT for's ability to destroy us all. So I'm being a bit glib here, but Open Ai actually did allow an AI research group called the Alignment Research Center or ARC, to test the Large Language Model to make sure it doesn't show signs of being capable of starting the robot uprising and learning all of us into batteries or something Specifically, the AI group was looking into possibilities like the model's capacity to copy itself or to acquire resources, whether digital or real world or whatever, or maybe its ability to manipulate people with stuff like phishing attacks and that kind of thing. Also whether or not it can make high level plans and then maybe follow through on those plans. So this kind of gets into some of the scary stuff around AI that we typically think of as being science fiction, right, like Terminator style stuff. The thought of a sufficiently intelligent AI could be capable of escaping whatever constraints we try to put on it, because by definition, if it's super intelligent, then it's smarter than we are, right, So if it's smarter than we are, then there it stands to reason it would figure out a way to get out of whatever box we try to put it in, including the possibility of manipulating people to essentially set it free. There have been thought experiments that have shown this is entirely possible. The AI researchers concluded that GPT four doesn't measure up to this kind of dangerous AI, But then, as Ours Technica reports, there are scarce details on the nature of the tests that the research group actually conducted, so it's hard to judge whether or not their findings are trustworthy themselves. Also, I think we need to remind ourselves that whether or not AI poses an existential threat to humanity, we can acknowledge that AI contributes to real problems in the here and now. Now. Usually this comes as an unintended consequence, such as the inclusion of bias within a system that leads to one group of people being disproportionately affected by negative outcomes, facial recognition technologies being a great example of that, but it could also be included by design, right, someone could make an AI applicant designed to create harm in some way. The alignment part of Alignment Research Center refers to the goal of making sure that AI is in alignment or agreement with basic principles that are meant to benefit humanity. But here's the thing. Bias can come into that sort of decision making process too, and you have to ask tough questions of who decides that criteria, who is it that's defining what is beneficial because believe me, there is not a universal or objective truth to what that actually even means. And even an effort to create an quote unquote aligned AI could result in one that benefits a small group at the expense of everyone else. Earlier this year, I mentioned that the Chinese megacompany by Do was hard at work on its own aipowered chatbot that it had called ernie Bot. Well earlier Today, by Do showed off ernie Bot, but did so with prerecorded videos in showing off its capabilities. Obviously, prerecorded videos can be fudged a little bit, so that brought some concern. It also didn't launch any sort of public access to the tool. It's keeping access very much limited. And dissatisfaction followed, and Bydu's stock shares dropped as much as ten percent initially, though that leveled off a little bit later, so the company only lost around three billion dollars in value. Yikes. Bydu seems to be taking a more measured approach to deploying its AI chatbot, which I actually argue is the responsible thing to do. We have seen numerous stories of how GPT has caused problems with premature deployments, such as homophobic jokes on the endless computer generated Seinfeld episode that's on Twitch. All way to creating a concern among teachers that their students could be using a chatbot to cheat on assignments, and more so, a more deliberate rollout seems like it's a good idea to me. But then I guess there's this perception that Baidu is trailing behind companies like Microsoft and Google, and there's this fear that it's not being competitive. I think not being competitive in this case also means you're sidestepping some potentially really troubling problems down the line. But I'm not an investor, So there you go. Okay, we've got a lot more stories to cover, but before we get to that, let's take a quick break. We're back, so let's talk about TikTok. We've got a couple of stories about that. Again. The government of the United Kingdom is following the latest trend of banning TikTok on government owned devices, or at least it's considering that option. By the time you hear this podcast, the matter may have officially been decided, but as I was recording it, it was still something that was being considered, but we were expecting an announcement sometime today. The UK would follow in the footsteps of the United States, Canada, the Netherlands, the European Union in general, and several other countries if they were to do this, and yes, the heart of the matter is the security concerns of data potentially filtering to TikTok's Chinese owned parent company, Byte Dance, and by extension, to the Chinese government. Yesterday, The Wall Street Journal reported that the United States could send TikTok and byte Dance and ultimatum, which is essentially, you must sever all ties between TikTok and bitte Dance. By Dance would have to spin off TikTok as an entirely independent company and not be the parent company anymore, or TikTok risks being banned nationwide. This echoes what former President Donald Trump tried to do a few years ago. He tried to force TikTok to divorce itself from its parent company and become a truly US organization. During Trump's presidency, the push was for TikTok to sell itself to some larger American company, but this never came to fruition, partly because TikTok argued in court that this directive from the president would violate a law called the Burman amendments that prevent presidents from using economic pressures to force international communications platforms from doing stuff. So essentially, they were saying this is against the law, and now we're seeing a similar push from the Biden administration. I'm not exactly sure how that same law wouldn't apply in this case, Like, I don't know that there's a new argument to be made if it's coming straight from the president. Anyway, TikTok reps argue that a massive one and a half billion dollar program meant to secure data in the US and to allow third party security companies a chance to monitor TikTok's operations to look for anything hinky going on. They're saying that's more than enough to solve the worries that TikTok is serving as a threat to national security. And then further, reps for TikTok have said that if by Dance did divest itself of TikTok, that wouldn't actually prevent the transmission of data from TikTok to China. I'm not entirely sure how all this tracks, but then I haven't been given a glimpse of all the details, so I doubt I would even understand all of it if I did, because this is wrapping into things like finance and trade more than the technology itself. Anyway, it looks like we're heading to an impass and I'm not entirely sure where it's going to go from there. But my guess is we're not going to see the pressure on TikTok going away anytime soon. Hey, y'all remember FTX, You know, the company that used to be the second largest crypto exchange in the world before word got out that the team behind it was practicing some creative accounting that you know, you might be able to look at another point of view and call it, you know, fraud. But more bad news has emerged about the company and its high profile co founder, Sam Bankman Freed aka SBF. So the new folks in charge of FTX, who are mostly going through the process of selling the company off for parts in order to return as much value to investors and customers as possible, have said that they have uncovered transfers that collectively amounted to around two point two billion dollars and that this money was sent to SBF through various means, primarily through Alameda Research, the hedge fund company that was you know Yen to ftx's yang. An additional billion seemed to be pulled from FTX to go to other key members of FTX, and yeah, Alameda Research handled those transfers, according to the new owners of T and that you know, it's been going on for a while. And I personally consider the report shocking but not surprising. By that, I mean, I'm not at all surprised that SPF was trying to pull as much cash out as he could before FTX came crashing down. It is shocking exactly how much money that was, right like to pull out two point two billion dollars in various means as you're trying to salvage as much as you can from a ship that's sinking. Incredible Anyway. SPF is currently out on bail awaiting his trial, which, according to the current court schedule, will not begin until October this year because justice is super swift. The Verge has an article titled the Silicon Valley bank fallout is just beginning. It's a great article. You should go read it on the Verge and the articles mostly about how tech companies are trying to figure out where to go for word from here, what's the safest place to bank now that SVB has gone under, because if you remember, SVB was like a financial pillar for the tech industry. So this piece lays out the challenges that tech companies currently face, including how to mitigate risk while ensuring that tasks like making payroll aren't interrupted. Like you might argue, hey, it'd be really good to make sure you're depositing in different banks so that way you're not having all of your eggs in one basket, and if that basket goes belly up, you're still all right because you've got eggs and other baskets, right, Except that if you're trying to do things like pay out a large group of employees, having that money dispersed makes it much more difficult to do those kind of tasks. So that's the sort of challenges that tech companies are looking at. The piece also mentions that we're probably going to see a change in how startups are funded and how they operate, and part of me thinks that a reevaluation of startup culture is long overdue. I have often worried that the tendency for startups to launch without first creating a firm business plan, instead essentially having a desire to get acquired by some other company has really led to irresponsible behavior, and that this has gone unchecked for too long. This created too many companies that ultimately produced very little value. Right. You'll hear about a company like Google or Meta or Amazon scooping up some of these companies, and sometimes nothing really seems to come of it, right, and you're just like, well, what good did that startup due apart from make the founders rich? And then you see the founders go off and do the same thing again, where they'll start up a new company, perhaps one that also has very little business plan element to it, also with the hope that they are going to get scooped up by another company, and they do it again and again and again, because coming up with an idea that sounds attractive is way easier than making that idea work. And if you are good at selling ideas, you can just do a serial approach to create cool sounding idea, sell it off to someone else, rents, and repeat and by an island somewhere, I'm hoping that this represents a reality check moment where we will see a more thoughtful and careful approach to funding startups that might be a vain hope, because you know, historically investors have shown a really strong desire to get in on the thing that's possibly going to take the world by storm, so that you get so rich that you make Scrooge mcdock look like a pauper. Anyway, the article in The Verge is a good one again, it's called the Silicon Valley Bank fallout is just beginning. It also takes time to point out that the collapse of SVB was in large part the fault of some of the venture capitalist investors themselves. Like you had the leaders of big investment fund groups saying, hey, you should probably pull your money out of SVB because if you don't, someone else is going to do it, and then you won't be able to get your money, and so they ended up creating the very crisis that they were warning people about. So I think a lot of financial institutions may still view loans to the tech sector as being risky because the venture capital community have proven themselves to be self destructive. It's hard to trust in an industry when you see the tendency to act in self interest to the point where you harm everybody else. Very difficult to put your trust into that group. Okay, still have a few more stories to cover before I get to that. Let's another quick break, Okay. The Federal Trade Commission or FTC, has finalized its judgment against Epic Games regarding the company's practice of enticing or perhaps even fooling Fortnite players into making in game purchases through what the FTC calls dark patterns. So essentially, dark patterns refers to creating an interface that makes it really easy, far too easy, most would argue, for a player to make an in game purchase, perhaps without even knowing or understanding that it is a financial transaction. One of the ways the FTC argued Epic Games did this was that it ended up playing fast and loose with button inputs, so that you might have a button that normally has you back out of a menu option, right, Like, maybe you've had your menu UI set up so that when you hit B, you're backing out of that part of the menu and you go to a larger part. But then in a transaction, maybe you suddenly make B confirm instead of backout, and so people who are used to using bing and backout hit B, but now they've just made a payment, and that might not be something you could easily reverse or cancel out of, like maybe there's no confirmed feature. You just boom, you've done it. Another part of the problem is that a lot of the folks who are playing Fortnite are kids, So the FTC argued that Epic Games didn't include enough protections to prevent kids from making in game purchases without parental consent, so they just started racking up enormous charges on parents credit cards that were associated with the account, and if a player did go so far as to contact their credit card company to dispute charges, Epic Games would then lock that person's player account so they couldn't play the game anymore. So now the FTC is telling Epic Games to cough up two hundred forty five million dollars as punishment. The FTC plans to use that money to provide refunds to affected players. Plus, Epic will not be allowed to block people who dispute credit card charges anymore per the terms of this agreement. If you think you were affected by Epic Games practices. In other words, if you feel you were tricked into making a purchase on Fortnite and you are a US citizen, you can go to the website FTC dot gov slash fortnite that's foart n Ie, and you can fill in a little online form to start the refund process. South Korea and Samsung announced plans to build a truly enormous semiconductor manufacturing campus this week. It's one that Samsung is going to invest two hundred thirty billion dollars in two The effort will make South Korea much more competitive with Taiwan's TSMC as a semiconductor manufacturing company that's responsible for more than half of all the semiconductors used in advanced electronics today. It's also a move to create a better supply in a world that not so long ago was in a serious crunch due to many factors, in the big one being the COVID nineteen pandemic. Samsung was one of several companies hit hard by supply chain issues, and by investing in semiconductor manufacturing facilities within South Korea, there are hopes to head off future problems like what we saw on the recent past. This is not that different from what we're seeing here in the United States. The US government has poured billions of dollars or at least earmarked billions of dollars for similar investments here in the US again to try and alleviate some of these bottleneck points for supply chain issues in the future. Over in Japan, the US startup Lift Aircraft Incorporated held its first test flight of the Hexa flying car vehicle with an actual human being piloting it. So it was the first time it had a piloted test flight. So it looks like like an oversized quad copter that has a cockpit in it that can hold a single rider. Now, the plan is to have flying vehicles give rides to passengers in time for the twenty twenty five Osaka Kansai Expo. And from what I read, it is a single seater vehicle, so I'm guessing that for passenger rides it must be either controlled remotely or possibly autonomous. I think remote control is far more realistic, but there's no way you would just hand over control of a flying car to some rando and then just say good luck out there. Anyway, this particular vehicle isn't meant for the long haul because according to Asahi dot Com, it can fly for about fifteen minutes, so you don't even get twenty minutes of travel out of this thing, so it's meant for very short distance as it travels at speeds top speeds at around one hundred kilometers per hour that's about sixty two miles per hour. Japan's Transport Ministry is now hashing out the regulations that lift aircraft and other vendors that are also planning similar services for this particular exhibit. These are going to be the rules that they'll have to follow, which is good because this is really one area where I don't want to see regulation trail too far behind the technological innovation. People's lives are involved here right, their lives and their safety and their health. I think that it's very good that there is serious work being done in the regulation side. Richard branson satellite launch company Virgin Orbit, has effectively closed up shop, at least for now. According to BBC News, Virgin Orbit has stopped all operations and put staff on furlough. This is in the wake of a failed launch attempt that happened off the coast of Ireland earlier this year. And by a launch attempt, I mean an actual attempt to launch a payload into orbit. I believe it was nine satellites in total that were part of this payload. So in early January, Virgin Orbit attempted to launch a launcher one rocket which rides on an aircraft, especially fit seven forty seven, and the seven forty seven reaches a certain altitude, then it deploys the launch vehicle, which ignites its engines, and then it goes from there. But something somewhere went wrong. So at the time, Virgin Orbit said the first stage rocket fired exactly as it was supposed to, it shut down exactly as it was supposed to, and the second stage ignited just like it was meant to. In fact, that initially Virgin Orbit said that the payload had even achieved orbit, but by the time you got to about half an hour after launch, they changed their tune. They said the payload failed to reach orbit, and there wasn't much information about why. Later on we heard that apparently an engine had overheated due to a filter becoming dislodged, so that was possibly the reason for the failure to reach orbit. Now, it's not unusual for young rocket companies to have problems like this, and we indicate that something isn't that difficult by saying, well, it's not rocket science. And the reason we do that is because rocket science is bloody difficult, so it should not come as a surprise that there will be failures. But whether you think of rocket science as being difficult or not, investors want results when you're talking about companies, and if you fail to provide good results, then the investors are more likely to bail on you, and that seems to be what happened. Virgin Orbit needed a real win back in January, and the launch failure hit the company really hard. Ours Tetnica predicted back in January that Virgin Orbit would face a potential existential crisis as a result of it, and now that crisis seems to have hit. And again this did not come as a surprise because it wasn't just Ours Tetnica predicting that Virgin Orbit was going to be in real trouble as a result of this failure. Multiple analysts had projected that at the rate that the company was burning through money, it would be out of business this month. And while we can't say that Virgin Orbit is really most sincerely dead, it certainly has a rough road ahead of it and if it wants to return to solvency, Finally, NASA and a company called Axiom showed off the stylish new moon suits yesterday. I've done full episodes about the evolution of the space suit, including discussions about the differences between astronaut and cosmonaut suits. While space suits have evolved over time, for the most part, changes have been small iterations. But these new suits have a lot of improvements and more significant changes than earlier models. For one thing, they've got a lot more joints in them, as in joints that allow for movement, not you know, for twenty just blaze, and these allow for greater freedom of movement when you're galumping across the surface of the Moon, so you can do things like crouch and squat, you know, important stuff when you're playing a first person shooter. Also important stuff if you're doing things like doing science on the Moon. So yeah, astronauts are going to have a lot more freedom of movement while they're wearing these suits. Another big change is that previous suits required you to get into a lower half first, so it's like you're putting on a pair of space pants. Then you had the upper half attached to you, and they would attach to the lower half and create a seal. But yeah, the suits were in two pieces, right, a lower in an upper half. These new suits don't do that. Instead, they have an entrance through the back, so you open up the back, you get in, you get zipped up, essentially like you're secured by your team, and then you're wearing your own little space onesie that way. It's not little. I don't know why I said little. These are big suits. They also come equipped with some cool stuff like HD cameras, so you can get that awesome high definition point of view shot while someone's walking across the Moon. They have also improved thermal insulation so that astronauts are going to be able to wander the Chili regions of the South Pole. That's the area that NASA has identified as the potential site for a long term space colony type thing or a moon base, and it's important to have that kind of insulation if you don't want to freeze your tootsies off. They are officially called the Axia Extra Vehicular Mobility Unit or AXIMU for short. The versions NASA showed off had this dark gray cover on them, but when used by the astronauts, they will be white space suits and Interestingly, NASA will not own these suits. They're not purchasing the suits. Instead, it's kind of like a lunar tuck's rental. NASA will go to Axiom and rent space suits for missions starting with the planned lunar landing that right now aims to return people to the Moon's surface by twenty twenty five. Side note, I remain a little pessimistic about that timeline. I think twenty twenty five is probably a doubt we're going to make that, but I hope I'm wrong. It would be great If I am, I will not be upset at all if we do manage to get back to the Moon by twenty twenty five. I think that's a very exciting prospect. I think it's a very inspiring thing. It gets people, especially kids, really excited about space and science and engineering, and that's always wonderful. Not to mention, we stand a chance to learn more, which is always cool. And to this day this is true. I find when I look up at the Moon, inevitably I just sit there and marvel at the fact that we put people up there. We put humans on the surface of the Moon. They walked around on the Moon, they played golf, on the Moon. Then we were able to get those same people back home to Earth safely. And then Noel Brown goes and says, yeah, but Stanley Kubrick shot the whole thing on a sound stage and I lose my ever love in mind. Y'all. This sounds like a joke, but it happened repeatedly this past weekend. When I was in Austin, Texas for South By Southwest, I would tell this story about finding the moon really inspirational and they'd say, yeah. I was just talking to Noel about conspiracy theories, and it took everything in my body not to lay into Knoll. Of course, I went Null's the sweetest person in the world, but man, I was going bonkers by the end. All right, enough of all that. That's the news for Thursday, March sixteenth, twenty twenty three. I hope you are all well. If you have any suggestions for future topics of tech stuff, please reach out to me. You can do that by going over onto Twitter and tweet at tech stuff hsw let me know what you'd like to hear. Or you can download the iHeartRadio app. It's free to download. It's free to use. Navigate over to tech stuff. Put that into the little search bar, it'll take you to the tech stuff page. You can click on the little microphone icon and leave a voice message up for thirty seconds and like, let me know what you'd like to hear and I'll talk to you again really soon. Tex Stuff is an iHeartRadio production. For more podcasts from iHeartRadio, visit the iHeart Radio app, Apple Podcasts, or wherever you listen to your favorite shows.