Tech News: Poisoning AI with Digital Art

Strip as Israel pursues a ground war against Hamas. In the past, companies that provide real time traffic updates have suspended operations in conflict areas in order to avoid turning into a method for gathering intelligence on things like a military presence and troop movements and that sort of thing. Bloomberg cited anonymous sources saying that the Israel Defense Forces or IDF, made this request to both Apple and to Google to temporarily disable live traffic notifications and Google Maps.

Spokesperson Carlone Bordeaux stated that quote anyone navigating to a specific place will still get routes and ETAs that take current traffic conditions into account end quote, so the navigation tools will still have access to this live traffic data. They're just not going to display the location of any traffic bottlenecks or anything like that. They'll tell you how long it'll take you to get to your destination, but not what points along the route you should expect slow traffic.

So that's just an update on things that are going on in the tech sphere in that particular conflict. Yesterday, the US government announced it has designated thirty one technology hubs across America that will focus on innovation and be eligible to compete for millions of dollars in grant money. The Economic Development Administration, which is a government office I had not even heard of before today, said that the goal is to encourage regions that are already pursuing innovation

in quote unquote critical technology ecosystems. The idea is that the government is encouraging research and development that will lead to technology and organizations that are all about job creation and improving national security, like removing our dependence upon other nations, for example, and also making the United States a leader in various technical fields. Areas of focus include some pretty

obvious candidates like artificial intelligence and clean energy. That also includes stuff like biotechnology, medical technology, and quantum computing, among other disciplines. The thirty one hubs span thirty two states and Puerto Rico, so it sounds like at least one of those hubs state lines. So I hope everybody can

get along. The International Energy Agency announced today that according to its estimates, renewable energy will make up nearly fifty percent of global electricity production by twenty thirty and that by that time there will be ten times as many electric vehicles on the road as there are today. Now, that is encouraging, but we should also remember that our overall demand for electricity grows every year. It's not like

it stays stable. So while renewable sources will take up a bigger piece of the pie, we're also talking about a much larger pie. To that end, analysts expect that we'll see peaks in the demand for fossil fuels this decade. Also, electricity is just one way that we actually use energy. If we take all of our energy needs into account, they predict that the share that fossil fuels represents will

go from eighty percent down to seventy three percent. So there's it's a very long way to go in order to truly eliminate our dependence on stuff like fossil fuels. The agency warns that despite the move toward renewable sources, are continuing use of fossil fuels currently has us on a trajectory to see a global temperature increase of around

two point four degrees celsius. That's nearly a full degree above the target that was aimed at in the Paris Agreement, and it at least implies that we're going to see more major shifts than things like climate and then more granularly weather patterns and severe weather events. Yesterday, Tesla disclosed in a financial filing that it is the subject of

investigations from the US Department of Justice. So the DOJ is looking into the company for a whole bunch of different reasons, including how the company markets its driver assists technologies, which Tesla refers to as autopilot and full self driving or FSD. The investigation is also looking into whether Tesla purposefully misled customers regarding how far their vehicles can travel on a full charge of the battery. It's also looking into the company practices, like the ones that relate to

quote unquote personal benefits and personnel decisions. So one of the personal benefits could be the rumored glass house for Elon Musk. I had totally forgotten about this story, y'all, And when I read that there was a glass house project, I thought, huh, I don't think I've ever heard the term glass house project. I wonder what that is. Is that like skunk works. No, we're talking about a literal house with glass walls, you know, the kind of house

where you don't want to throw stones. The question is if Elon Musk was actually using Tesla to fund and build this house for himself, that would violate laws regarding a company paying out benefits to an executive on top of their normal compensation. I'm sure the full investigation has a veritable laundry list of concerns, because Musk does have a reputation for, let us say, playing fast and loose with rules, as well as a history of campaigning to

have fewer rules in general. Analyst Ming chi Quo, whom I mentioned recently in another news episode, has some more juicy Apple news for us. Apple juice anyway. He wrote a piece on medium that says Apple is kicking into gear to get into the AI field. The company had lingered behind a bit when compared with competitors like Google and Microsoft, but that's pretty much par for the course

for Apple. If you look back on many of Apple's products, like the big revolutionary products that Apple released often, you will see they were not the first to market. The

iPod was not the first MP three player, for example. Instead, the company would spend time figuring out how to best meet customer expectations, or how to blow them out of the water, or how to even create expectations where the customer wasn't aware they had any Anyway, Apple has a lot of catching up to do, and Quo's analysis says that the company will be spending an enormous amount per year to get there. So on the conservative end of the spectrum, Apple would be spending around a billion dollars

a year to start working on AI implementations. Quo suspects it's going to be closer to five billion dollars by the end of twenty twenty four, so five billion split over two years, with the vast majority of that really spent in twenty twenty four. Even then, Apple will still be trailing well behind companies like Microsoft. Those companies have

been incredibly aggressive while pursuing and developing AI technologies. However, at least in the past, Apple has shown that its approach can have an appeal that its competitors just can't match, so maybe the same thing will be true with AI. And speaking of AI, Technology Review has an interesting article about how some digital artists are including elements that are invisible to human eye, but they're intended to mess with

generative AI. So generative artificial intelligence has to receive training in order to actually generate stuff. So with art, that includes feeding millions of images into the AI system so

that it, in turn can learn to create images. But this concerns artists because it's their work that becomes the fodder to train these AI systems, and you can end up with generative AI that will lift stylistic elements or sometimes complete actual imagery from an established artist's work and then pass it off as the AI's own creation or someone else if they don't even indicate that AI was used to make it. Meanwhile, you have artists concerned that no one asked them permission to use their art to

train these AI models. They certainly haven't been compensated for the use of their work. And then the artist's own portfolio can become the fuel that ultimately robs that artist of their livelihood or at the very least diminishes the value of the work they produce. So what's the solution. Well, some researchers have developed a tool they call night shade. Nightshade by the way, that's a family of plants, but I'm pretty sure they're specifically referencing deadly nightshade, a plant

that is in fact poisonous. And the tool makes changes to digital art pixels, and it's changes that we humans don't detect, Like it's invisible to us, we don't actually see that any changes have happened. However, the AI scraping these images totally can see that information, and the information is misleading the AI. It's giving the AI the wrong implication of how these various elements inside of art are created, and it would prompt generative AI to start making some

mistakes when trying to replicate certain types of stuff. And ultimately, the more the AI references works that have this kind of stuff in it, the more wrong they get. So, for a simple example, you might have some invisible AI information there that is instructing AI and how to make a picture of a dog, but it's all incorrect information. Maybe the thing that it creates ends up looking more like a cat than a dog, or maybe it looks like a lovecrafty and monster out of the depths of

horror and imagination. Who's to say. But the point is that if AI does start generating images like that just upon simple queries, it indicates two things. One, the AI company that created that tool trained it on art that has been put through this night shade tool, and that would indicate that, yeah, that AI company was doing this was actually using artists' work, presumably without permission or compensation. And the proof in the pudding is that, oh, oh,

your generative AI isn't creating pictures correctly. And of course number two is the generative AI won't be creating pictures correctly and thus will be less useful as a tool. You can't really use AI to generate art about something specific if it thinks that a house is actually a boat or something along those lines. By the way, you can actually read more about this in the MIT Technology Review article titled this new data poisoning tool lets artists

fight back against generative AI. Okay, I've got more news stories to go through today. But before we do that, let's take a quick break. We're back. The Verge reports that both AMD and Nvidia are gearing up to produce ARM based CPUs for Windows machines soon. So that's significant because right now Microsoft is partners with Qualcom when it

comes to CPUs for Windows eleven based machines. But it appears that in the not too distant future there will be a lot more competition in that space, and you may find yourself shopping for a new Windows PC and then you might have to compare different ARM based processors with each other. This could also put pressure on Intel to innovate in its own processor family or else risk losing market share to AMD and to Nvidia, as well

as to Qualcomm. I really like this news because healthy competition tends to mean better products and better prices, both of which are great if you happen to be a consumer. Gizmoto reports that Microsoft has finally addressed an issue in its spreadsheet platform, Excel, that caused a headache in the field of genetic research. So the problem is that Excel thinks it knows better than the user and will automatically make certain changes to data if that data is in

a particular format. So, for example, let's say you're a geneticist and you type in the characters sept and the numeral one sept one all together, and of course what you mean is septin one, which is a protein coding gene that plays a part in cytokinesis and the maintenance of cellular morphology. Excel would not know that you are creating a spreadsheet about genes, so it would assume that when you typed sept one, what you actually met was the first of September, and it might change that entry

so it says one dash SEP. The same is true if you happen to be typing about membrane associated ring CH type finger one. Geneticis would encode that as March one m a RCH one. That of course would turn into one dash mr. When your spreadsheet starts changing your gene codes into dates, that's a problem. And worse, Excel did not have a feature for you to turn that off, so you couldn't tell it, hey, stop converting my gene

code codes into dates. And it meant that when it got time for you to publish your work, you would have to go through and spend countless hours making corrections. Because the spreadsheets you were using had changed all your codes into dates, so you had to make sure which one of these should be dates, which of these are actually codes. You could easily make mistakes and miss stuff.

It was a huge headache. In fact, it was such a huge headache that a few years ago the field overall changed its encoding system to work around this problem. But now Microsoft has created an option in Excel that lets you turn off the autoconversion feature as long as you're not running macros on your spreadsheets, because then apparently it all breaks down. I'm not sure if this isn't a case of it being better late than never, but I do think it's an example of how an automated

process can actually make things harder rather than easier. When you extend that to AI implementations, it can raise some

really serious questions. Obviously, this is a much more simplified approach, right, This is a conversion for one tiny instance of data, But when you extend that out and you think about the sort of things like the conclusions that AI can come to based upon the information it has access to, you can understand how sometimes the conclusions are not really ideal all the time, because sometimes AI makes mistakes just like regular I does. And by IY, I mean intelligence,

not I. I wasn't making a grammatical error. I was making I guess joke is being too kind. Uh, let's just move on over In the UK, the Public Accounts Committee recently highlighted a looming challenge for the energy sector in that company. This is according to the Register, and

it all has to do with smart meters. Smart meters are supposed to monitor energy usage and maintain a communications channel between customer and the energy provider, and it's meant to keep an accurate tally on how much energy a customer is using, but also to tech stuff like interruptions and service and that kind of thing. Ideally, smart meters reduce the need to send actual people out to physical locations to take meter readings, and that ends up freeing

them up to do other stuff. Not too long ago, the National Audit Office in the UK found that energy companies had rolled out smart meters to around fifty seven percent of their customers, so that meant forty three percent still did not have smart meters. This, by the way, was way behind schedule of the initial plan. Also, about

nine percent of those smart meters were not working correctly. Now, the Public Accounts Committee built on top of this report and found that twenty percent of the smart meters that have been deployed are living on borrowed time, and that's because these meters are only compatible with aging cellular networks, primarily two G and three G. Meanwhile, telecommunications companies are planning on sunsetting those networks and shutting them down, and that means that the smart meters won't be so smart

after that. These twenty percent of are deployed smart meters, so this shows there are lots of challenges that still are in front of the energy sector in the UK. Nearly half of the businesses and homes in the UK have yet to receive a smart meter at all, some of those who have one have one that doesn't work, and then a whole bunch of them around seven million different locations are going to need a new smart meter before the cellular networks completely shut down, or else it'll

be like they have just a regular meter again. And considering that this whole project has had to push deadlines a few times already, I'm sure this comes as an extremely frustrating development. It also just reminds us that connected technologies are only useful for as long as the underlying connection exists, which I know that's obvious on the face

of it. We can take that for granted, right when we talk about things like the Internet of Things, there's a lot of future proofing that needs to happen for the Internet of Things to remain useful, because otherwise, if the Internet of Things is only compatible with older wireless technologies, for example, and we gradually sunset those because we've created better processes, better protocols for wireless communication, if we can't update those IoT devices, they just become inert and we

have all the stuff that isn't doing anything. So that's just a good reminder for us to have. And it's very difficult to future proof things because you never really know how far along they're going to go. Sometimes, if you're lucky, you have designed your product, not just lucky, but smart. You've designed your product in such a way where you can update it with things like firmware updates to keep it relevant longer. Ultimately, you're always going to

have to swap stuff out. Your goal is to make it as less frequent as you possibly can and to maximize the life cycle of the devices you've deployed. Anyway, this whole story was giving me flashbacks to the United States back when television broadcasts switched from analog to digital.

It caused no end of confusion. Only a small number of people were actually affected by that small in the relative sense, because they needed to get a converter to convert digital signals to an analog signal to feed into an old television in order to still pick up broadcast TV. For a while, everyone was thinking that they might need one of those converters, and very few people actually did. Dating in the modern world is hard, or so I'm led to understand. I personally do not date because my

wife frowns upon it. Jokes. I've been happily married for more than twenty five years at this point, so I am totally ignorant when it comes to navigating the dating world these days. To be honest, I wasn't exactly an expert back when I was single. Anyway, What if I told you that Tender is introducing a feature that will let you turn friends and family into your own personal Yenta.

Yenta was the matchmaker and fiddler on the roof, to be clear, So Tender's introducing this feature where you can list up to fifteen people on your account who will then be able to review potential matches based upon your tender profile, and then they can weigh in on which

ones they think you should swipe left on or right on. Now, they won't actually be able to do the swiping for you, which is a good thing because can you just imagine giving your parents access to your tender profile to be able to vote on whether or not you should ask that nice young person out on a date or not. What a nightmare scenario that is. I apologize if that

major anxiety spike just now. They will also will not be able to message the other profile, so you don't have to worry about them reaching out on your behalf. What they can do is give their opinion about whether or not a person looks like a catch or if you should just nope out of that potential relationship, and

they call this feature Matchmaker again. Getting back to Filler on the roof, I'm just having the song matchmaker Matchmaker make me a match play through my head, so hopefully you are not aware of that song so that it doesn't do the same to you. But I know that the rest of the day it's going to be playing in my head, partly because I was in a production of Filler on the Roof. Anyway, I think this sounds like Tinder is setting up the stage for a future

reality television program. Just imagine it's a reality TV program where people can only go on a date with someone who matches on their tender profile and gets the most votes from that person's can unity of matchmakers. So like your group of friends and family, I'll say this person is right for you, and then you are obligated to go on a date with that person. That sounds like a reality TV show, Ready to go. I'm just giving away ideas for free again, aren't I. You know what,

I'm gonna do some soul searching on that. When we come back, I've got a few more stories to talk about. Okay, we're back, so let me tell y'all a story. Way back in twenty twelve, a company called Cloud Imperium Games held itself a tiny little kickstarter was for a proposed

science fiction space game that they called Star Citizen. The game's scope was to be astoundingly huge, and it would include both multiplayer elements and elements that would be sort of a single player campaign, which became known as Squadron forty two now as Star Citizen grew and grew and grew in development and sold more and more in game products for a game that did not yet exist, and made more than half a billion dollars in funding, no joke,

like they they something like six hundred million dollars I think at this point. Anyway, the company decided to spin off Squadron forty two to be its own standalone single player campaign. But that campaign, just like the larger Star Citizen game, kind of just stayed in development year after year. And while supporters for Star Citizen have enjoyed releases of some content, like there's been limited releases where you've got

some gameplay elements that people can actually play. I don't want to I don't want to suggest that nothing has come out. That's not the case. But they were still waiting for the full promised title, and for Squadron forty two, that weight might and I stress might be getting close to the end because the company has announced it is now feature complete, so that means that the developers are done adding in features into Squadron forty two. They're not

adding in any more gameplay elements or anything like that. However, it's not quite ready to ship yet. The developer set it's working toward a beta and ultimately a release once they are done finalizing the game, polishing the content that kind of thing. They did not give a window for when that might be. They did release a new trailer, and I hope that the stars Citizen community that's been waiting for this content for more than a decade at this point are happy with the result, and that the

rest of Star Citizen also ultimately comes out. I do not have a stake in this. I never supported the campaign, I didn't pre order anything, and I have some very strong feelings about how this whole project has progressed over the years. Like I have been very critical of Star Citizen in general, but I genuinely hope that the people who have poured money and their own hopes into this

project end up being very happy. I would hate for this to continue on as just a never ending development, and I worry about people being disappointed with the final product simply because they've been waiting so long and have put so much money into it. I'm reminded a little bit of Bethesda, where they had their own Starfield product come out, and there were a lot of people who felt that it did not live up to their expectations. Other people really like that game. I have yet to

play it, so I haven't formed my own opinion. I know people who think it's a great game, and I know people who feel that it was a total disappointment. So I'm sure most people fall somewhere in the middle. But yeah, here's hoping that the folks who support Star Citizen, when they do get access to Squadron forty two, are not left disappointed. Now, imagine that you have to provide tech support to a spacecraft that has left your Solar System.

That would be quite the long distance service call. That's what NASA engineers are doing with Voyager one and Voyager two. So both of those spacecraft are now in interstellar space, they have left the Solar System, and NASA has sent some instructions to each of them in order to extend their ability to communicate with us all the way back here on Earth for as long as possible. So last year, Voyager one started to send some pretty weird communications about

the spacecraft's telemetry. Voyager sends telemetry data through a system called the Attitude Articulation and Control System or AACS. But it was passing this data through a computer system that wasn't working properly, and it was odd because it wasn't supposed to be passing the data through that specific computer system, and so sending back garbled data nothing exciting. It's not like the garbled data said, Hey, Earth, send more I love Lucy episodes, which would have been cool because then

aliens would be looking at us. Now it's just like just gibberish, really, And so the engineers have now beamed a patch to fix this issue with Voyager one. It's a software patch. There is the lingering question as to why Voyager one was routing the data through this computer system in the first place. We still don't know that answer, but hopefully this will at least be a temporary fix

for the problem. The team also came up with a new schedule for firing spacecraft thrusters for both Voyager one and Voyager two in order to reduce the frequency of those thrustering ignitions for a couple of reasons. One, obviously, there's a limited amount of fuel, right like, you don't have infinite fuel on the spacecraft. But the other one is that each time you fire the thrusters, it builds up a little bit of residue on the inlet tubes

that feed into those thrusters. So the concern is that if they kept up with a more frequent schedule, that build up would reach a point where it would interfere with the thrusters. So now they'll do it less frequently and hopefully extend the useful life of the spacecraft. And they're obviously not going to work forever, but the hope is to extend their lifespans a little bit more as

they continue on into interstellar space. Now I have a couple of articles that I'm going to recommend, but before I do, I do have one final story, and that is Martin Getz has passed away at the age of ninety three. Gets was the first person to secure a US patent for software. He did that back in nineteen sixty eight. Now, I've talked about this before when I did an episode about the Patent Office, because the idea of patenting software was one that had some major hurdles

to overcome. And that's because the US Patent Office has rules on what you can and cannot patent. One thing that you cannot patent is a mathematical process, and there are arguments to be made that software ultimately is really just a series of mathematical processes, and as such, software programs, etc.

Are not viable candidates for patents. That was the argument. Ultimately, the Patent Office rejected this idea, at least within certain parameters, and Gets his work in data sorting software became the first to receive a patent from the US Patent Office. Gets really campaigned for this in large part because independent

software developers had no protection for their work. So let's say that someone came up with a really useful program like sorting data in a mainframe computer system, Well, nothing would stop a massive company with practically unlimited resources from just copying or reverse engineering that software and using it for themselves and never paying the developer anything. So Getz

pointed out that would be unfair. It's not fair to just stand by while very powerful tech companies could steal the work from independent programmers just because there are no protective measures in place. So his patent set a precedent, and it meant that programmers had a chance to secure their work and protect their livelihoods. And I think that's a good thing, So rest in peace, Martin gets Now

that wraps up our stories. But there are a couple of articles that I would like to recommend you check out if you have the time. Again, I have no connection to these media outlets. I do not know the writers personally or professionally. I just thought these were good articles and you should check them out if you've got time and you're interested. So first up is a piece by Jennifer Pattinson Twoey of The Verge. It's titled Matter one point two is a big move for the smart

home standard. Now I'll have to do a full episode about Matter in the future and how the goal is to create a foundation for smart home technologies, but this article gives a great overview of where we're at currently in that effort. Secondly, I recommend Ashley Bellinger's article in Ours Technica titled will chat GPT's hallucinations be allowed to

ruin Your Life? This article covers how AI companies are trying to insulate themselves against stuff like defamation lawsuits, so that if their AI chatbot claims that I don't know that you go around kicking puppies or something, you wouldn't

be able to sue them for ruining your reputation. I think it's valuable to read that as well because it very much plays upon the concepts of accountability and artificial intelligence agents very important, really, and it highlights something that is kind of in this gray area when it comes to the law. You know, we're starting to define AI in a legal sense through various court cases. For example, the court case that determined that AI is incapable of

holding a copyright. You cannot copyright a work that was created by AI because human authorship is a key component of copyright as it is defined today. So that's kind of how you know, we have to define things in the court system. It has to be done by the decisions of judges in various trials. So a very important part of the adoption of artificial intelligence inside the United States. And that wraps up the tech news for Tuesday, October two, twenty forth twenty twenty three. I hope you are all

well and I'll talk to you again really soon. Tech Stuff is an iHeartRadio production. For more podcasts from iHeartRadio, visit the iHeartRadio app, Apple Podcasts, or wherever you listen to your favorite shows.