AI Regulation

It seems now the whole world is constantly concerned with AI, and even now the governments are getting involved. Well, it's not like it was unexpected. The big tech companies who trained the models have been begging daddy government for regulation for a while now, and, well, guess what's happening. Money is a powerful drug.

I'm a little unsure of it. The basic attitude is that AI is a powerful new technology, and is disruptive, to use a word from the dullified modern lexicon, and so will need regulation to prevent bad things (unspecified) from happening with it.

Well, I'm not sure.

The Problem with Regulation

Isn't it odd that these companies, which so far have benefitted from a lack of regulation, are, now that they are entrenched as the powers which started the new AI boom, asking so fervently to be regulated? It seems odd. Surely, any company would benefit from a lack of regulation.

Well, not really. Laws are enforced ~it~fairly, which unfortunately means unfairly in this modern world. There is no separate set of laws for those with power as compared to those without, and so the massive players get the same punishment as the ones just starting out. A £100k fine for this-or-that violation is a massive penalisation for any new-comer with relatively little money; for the entrenched, multi-trillion corporations, it's sod all. They work it into their budgets as the ~~it~cost of doing business~it~.

This is why we repeatedly see pathetically tiny punishments for even huge violations committed by these mega-corporations. Really, what we need is a system which fines individuals and corporations according to a percentage of annual revenue, instead of a fixed sum, but we need to remember it's these self-same companies that write the laws. They aren't going to let it change to their detriment.

This means that any regulation that gets passed becomes an easy matter of, ~~it~well, we'll hire another compliance person~it~, for those companies which have the means for that, and a back-breaking expense, for those that don't. In essence, it just serves to raise the barrier to entry, further exacerbating the problem of monopolisation.

But it needs regulation!

Does it? I would reckon to say that we should not regulate ~~it~AI~it~ per se, but rather, we should regulate based on ~~it~the ramifications of AI~it~.

With any new technology (and realistically, AI is hardly "new" tech - we've had the concept of it since the 1960s, and the only reason it's caught on now is because we have more compute power) there will ~~it~always~it~ be changes. That is inevitable, and is the reason why we invent things in the first place. The problem we have, is that for something on this scale, we think, we don't know the ramifications of it. Shouldn't we step in before anything bad happens?

Well, firstly, this won't be the end of the world. This isn't some stamp collecting maximizer AGI, these models we have at the moment are LLMs. Now, they are big LLMs, but they won't be the end of the world. That won't happen for a fair while yet.

Secondly, surely we need to know what we are regulating? The entire reason why regulation thus far has generally been slow (when it is well-intentioned) is because we need to know the dangers first before we can regulate. And we don't know the dangers of AI. We have some ideas - it might cause problems in schools for instance, where the currently system is not adapted to cope with it, but really, we don't know what the good and the bad is. And there ~~it~will~it~ be good with the bad. When we realise that a particular thing it does is bad, then we can work to regulate that - but to throw the baby out with the bathwater here is just not a smart move, especially considering the consequences that these actions will have.

Thirdly, the cat's out the bag now. We can regulate corporately produced AI, the ChatGPTs of the world, but it's not hard to make your own model. Yes, it may not be as powerful (which again, will mainly be limited by compute costs, and not anything like skill or technology - ChatGPT costs 700k per day to run.), but it's far from difficult. We can't go back now, just like how we can't go back on the internet. And the internet has a lot of bad, but it also has a lot of good - I wouldn't have this website would the internet.

Why is the Government helping?

I understand the incentive for the companies. Of course, the only thing they are interested in is consolidating their own power, and increasing the number in their bank account. They could not care less about people, or morals, or anything. We know Meta doesn't care about people after they decided to keep giving teenage girls suicidal thoughts because it was making them money; we know Apple doesn't care about people given its products are made with slave labour; etc etc. It's obviously not a moral thing for them, and the only reason to believe they might actually be doing good is if you've fallen for the hype.

The question is: why is the government helping them? And particularly the UK government? (The US government may well have some vested interest, given that the companies belong to the US).

If I'm honest, I'm not sure. The topic needs more deliberation. I think partly, (and I think this with all the actions of the government), they do actually, genuinely buy the propaganda. Rishi is always talking about the ~~it~war on motorists~it~ as well, despite the damage that it does to our country. But I think he likely genuinely believes it. Same with all this AI stuff.

There's probably ulterior motives at play as well. ~~it~Follow the money*, I guess. I can't help but imagine (altogether without evidence) that at least some money is changing hands here.