The Hater’s Guide to Generative AI – Page 5

When you work in the tech space, you are guaranteed to be surrounded by AI and its fan club. While the technology itself has some wonderful applications, I see it as no more than a gimmick, at best, and a major societal challenge, at worst. A Poor Definition of AI Artificial intelligence (often just abbreviated to AI) has a broad spectrum of definitions and applications, probably as broad as the field of computer science at this point. That said, in my experience, AI is generally described in two forms: classical AI and modern AI. Classically, AI mostly refers to deterministic algorithms which solve complex problems quicker than humans, like the algorithm to compute the shortest path between two points or the algorithm for Tic-Tac-Toe. In general, these are tree-based and graph-based algorithms for solving real-world problems. Though, I think other algorithms were common like rule-based, clustering, and even statistical algorithms. More recently, AI has exploded in the realm of neural networks (which I believe is really just a subset of AI called ML). I assume this is due to the relatively low cost of computing power, specifically graphics cards, as neural networks are extremely computationally expensive. Regardless, now machines can "learn" in the sense that they can generate a statistical model by continually tweaking numbers until the model is "close enough." If you can sense my frustration already, it's because I don't really buy into the hype of AI (or rather ML if you want to be pedantic). In the remainder of this article, I'll talk about why. AI Is Everywhere And Not for the Better One of the main reasons I don't care for AI is that it has seemingly infiltrated every space, whether it's going to have positive outcomes or not. My first experience with this was coming back to academia to start a PhD in computer science. At the time, I was in a graphics lab that focused on data visualization. Research projects in that space came in two main varieties: data visualization for machine learning or machine learning for data visualization. In other words, people aren't doing data visualization work anymore. I assume that's going to have lasting consequences on the progress of the field, and I would not be surprised if most other fields of research are under the same threat. Then, when I step into my classroom, it seems like at least half of my students have some interest in AI. The marketing campaign is working, and young folks are jumping at the hype. To me, that's a bit concerning because machine learning is just one application of computer science. What happens to all the other subfields? https://twitter.com/nancytaughtyou/status/1764491684446900521?s=20 And this might seem out of left field, but let's just consider the stock market for a second. So many companies are going into AI that a bubble is forming that could rival the dot-com bubble, nearly 24 years to the day. Obviously, the internet is still here today, and AI will still be here after such a bubble. But, it does seem like there's way too much widespread adoption of the technology to the point of tech being created that otherwise shouldn't be. The Industry Is Running Out of Ethical Applications of AI Because AI has seemingly infiltrated every organization on the planet, we're starting to get applications of the technology that drift into dystopian territory for the sake of that sweet, sweet cash. For instance, you've probably all seen what Sora can do. Aside from its Euclidean geometry issues, the technology is extremely good. It becomes increasingly more difficult to tell the difference between reality and AI generated video, images, and even audio with each passing day. In an academic setting, I hear about desires for AI cheating detection tools regularly, which is not a great position to be in as an educator. I'm not interested in the AI arms race, where we develop AI tools to fight other AI tools. However, many of my peers are. As a result, there is a demand for tools that cannot be held accountable when they're used against students. https://twitter.com/MIT_CSAIL/status/1604884273789603842?s=20 And now, there are goofy applications like AI lawyers, which only serve to further distress our terrible legal systems. Thanks, tech bros! AI Is a Weapon in the War on Truth To make matters much, much worse, there is this hidden assumption that AI models contain truth. In reality, AI has to be trained on real data which is generated by real people. As a result, there's no surprise to me that AI always seems to become racist when given enough data. Then, there's this adversarial aspect of AI, where folks can inject models with bad data to break their effectiveness. To me, it wouldn't be out of the realm of possibility to flood the internet with racist imagery, so AI models pick it up. Adversarial techniques like this are already being used by the good guys to protect artists from generative AI. I suspect more of this will be needed in the future to avoid some of the more detrimental effects of AI. I can only imagine how cooked society will be once AI is used to cover the news. Truth will officially be dead. While on the subject of truth, one of the newer terms in the AI lexicon is hallucinations, which refers to an AI model's ability to generate lies. This comes mostly from chat-based AI (i.e., LLMs) tools like Chat-GPT, which will create studies that don't exist for evidence in an argument. In the tech space, I've also seen them hallucinate APIs that don't exist. If hallucinations seem like something only AIs and people on psychedelics can experience, you'd be wrong. You too can experience hallucinations any time you get on a social media platform now. Whenever you read a post, there's a good chance that there isn't a single person on the other side. Instead, the content is generated by an AI, and the comments are all generated by an AI. Welcome to dead internet theory. What a wonderful time to be alive. https://twitter.com/battleupsaber/status/1763488862104441014?s=20 Finally, since I watch a lot of Kitboga, I'm familiar with how AI is used or will soon be used in scams. Right now, it's concerningly easy to clone people's voices using AI. As a results, it's not out of the realm of possibility for a scammer to clone the voice of a family member and demand ransom. With Sora, the scammers could even appear as your family member, like some animotronic VTuber. And That's Just the Tip of the Iceberg Look, AI has been a fun toy in my experience, but that's about the extent of it. I've enjoyed chatting with bots and generating an image here and there. I've even enjoyed getting a bit of code completion. However, the reality is that AI is being used to do much more than provide fun little experiences for the average person. It's playing with people's lives, and we're in a situation where I only expect things to get worse, not better. In my dream world, we would use AI to solve world problems. For instance, we'd automate all of the jobs that people don't really want. Then, take that money and fund key government programs with it or convert it into a form of universal basic income. Unfortunately, there is no money in that, and the worker is just going to be continually alienated from their labor until there is a single oligarch running the show. Maybe, we'll have a revolution before then. Who knows. With all that said, I would usually sign off with some links for you, but I decided halfway through that I was going to paywall this. As a result, you're already my most active community member, so take care! I appreciate your support.