Repetition
The thing that bothers me most about AI is that it is not correct all the time. This might sound like a silly thing to say, but it's very important to me.
Prompting an LLM is very like talking to a person in some ways, you can ask a question and get a response. The major difference of course, is that nobody owns the response. If you point out a mistake with a person, it usually leads to a learning moment for one or both of the participants. Oh, we made a mistake in our communication or one of us had a misunderstanding of the topic. Perhaps a simple case of misremembering.
On the other hand, an LLM can and will immediately pivot, apologising, before giving you a differently incorrect answer or even the same one again. This "speaking", with no basis in beliefs or an identity or experiences, essentially acts as an incredible gaslighting machine. You can be gaslit but only in the direction your mind was vaguely already pointing. If you have a misunderstanding, it will help you justify this misunderstanding backing you up with evidence (real and imagined).
If you were to point out that even if this technology is wrong some of the time, it makes peoples lives easier and better (debatable). I would usually have to concede that I think making individual peoples lives better is good, but in this case it's not so simple. Because LLMs are not trained on all of human knowledge but instead a subset chosen by the people who trained it. This means that inaccuracy can occur not innocently but intentionally.
We need look no further than a certain social media app's LLM bot to grok (ha?) this. Its owner outright declared that it was too politically correct and was going to have it tweaked to be less "woke". Based on the responses by the bot after this announcement, we can see that the intentions of an LLMs "owners" do alter the information we receive. The most troubling part is because hallucinations are an inherent part of the system, it exists as an easy scapegoat for these tech companies and CEO's. The LLM didn't deceive you intentionally, it's just a feature of using an LLM.
These companies don't even face any significant consequences when we can see that their product outright encouraged, helped or pushed people into committing murders or suicide. I feel like I need to take a moment to point out that any financial penalty that isn't a significant chunk of a company's or corporation's profits or doesn't cause severe damage to their ability to continue doing whatever resulted in the consequences in the first place is not a significant consequence. The profit motive means that if the nett gain was greater than the loss, the action was not only justifiable but good.
It feels like there has never been less accountability for a company's actions. The vast power and wealth large corporations control means that they need to be more responsible and more accountable, and yet somehow they squeak by. (Don't even get me started on self-driving cars and the marketing shenanigans there)
The people who make and sell AI have changed the world. There is no doubt about this. But why were they allowed to? What gave them the right to fundamentally alter our lives without any clear or actual evidence that what they were doing was good or even beneficial? Governments did not stop them. Corporations bent over backwards to accommodate and support them. These already large, already rich, corporations basically just said this thing is good and beneficial and our entire economy has warped around it (Nothing worrying about that for sure).
Corporations have never misled society before right? They wouldn't do that, but surely if they tried the government would step in? I've never heard anything bad about tobacco or fossil fuels or pesticides or sugar. After all so many scientists came out saying that these things were actually great and the dangers they pose are fake or overblown.
You're telling me those scientists happen to be hired or financed by the large corporations?
I'm sure thats just a coincidence, surely they wouldn't mislead us for money...
Government and regulation have always lagged behind tech advancements. In some ways, this is inevitable. After all, how do you put rules on something that had not been conceived of? In practice, it's less inevitability and more a mix of "efficiency" and perhaps complicity. After all, AI and the idea of neural networks have been around for quite a while. However, governments and regulatory bodies aren't likely to regulate a hypothetical, if they did they would have to constantly be regulating everything.
However, I find it hard to imagine that all these major tech companies had been working towards and getting closer to "breakthroughs" of actually successful implementations without anyone hearing about it. Which leads me to think they convinced the right people to not push for anything to be done so they could push it to market and now we have to deal with the consequences. After all, now there are certainly people who would be extremely incensed should LLMs be made illegal until we could determine if them using the internet as training data or they were actually beneficial (Huh... did I mention tobacco before?). Once we have access to a thing, we are much more likely to feel strongly about it being taken away than if we never had it in the first place.
At my heart, I believe the internet should be a place of freedom. It started as a piece of technology that simply allowed people with computers to communicate with each other. Unfortunately, while governments weren't too aggressive in trying to control the internet, that is changing. To be honest though, computers and the internet have effectively been under the control of several major companies for some time.
While Linux exists, we have to be honest and admit it really was(is?) not user friendly enough to appeal to the average household computer user. Macs are fairly price restrictive and the majority of macOS users are from specific fields. So, our Windows OS overlords have been free to gradually make our PC using experience shittier and less in our control, the argument forever being "Oh, if you don't like it you can always switch" as if that is a real option. Blaming the consumer and asking them to leave in an effective monopoly is a pretty convenient cop-out.
Ditto with Google and Meta controlling most of the internet, serving us more and more ads, and stealing and "generating" content with AI. But of course, it's not stealing when they do it, you posted stuff on the internet, how can you blame them for using it. Ignoring, of course, the armada of lawyers they would shove down your throat (and several other even less pleasant places) if it were to happen in reverse. We aren't even worth giving a way to disable or opt-out of useless and shitty features these days. I don't want or need the WhatsApp AI button! But what am I going to do, everyone uses WhatsApp. My friends all use it. Businesses use it. The option does not exist to simply go to a competitor (RIP Telegram).
It's not about features, it's not about choice. They are part of the fabric of life now, and as a result there should be some measure of accountability. Instead, as always, they continue to take advantage, all the while pushing for regulation that would give them more power, more control. It's funny how they are always anti-regulation until they are the ones writing the regulation.
Tech companies have been engaging in shady business practices to really cement their control for years. There's no need for competition when you can simply buy them out or offer heavily subsidized rates to run them all out of business. They are allowed, enabled, to amass this power, but somehow don't have to take any responsibility.
Anyone can start a tech business after all, it's your fault that your Amazon competitor didn't take off. It has nothing to do with the exorbitant amounts of money they can burn on the vast network of warehouses, contract drivers and warehouse workers (peeing in bottles of course). We don't even need to mention the extremely beneficial deals they are able to strong-arm local postage services into.
Some of these AI bros or tech CEO types talk about the big apocalyptic potential of AI, which feels to me very much like a man holding a bomb saying "This bomb I financed and had built could kill everyone if I press the on switch" after having already pressed the on switch without hesitation. Never mind that AI can't even do most of the things they promised 3 years ago. AGI is always 5 minutes away if we take them at their word.