I have said this before, but I was going to be ether a writer or a lawyer when I wren tot college. I decided on lawyer once it became clear to me that writer was a remunerative career. I had bene poor, and would continue ot be poor, poor enough that occasionally I didn’t know where my next meal would come from and even more occasionally that it came from nowhere. I had no interest in paying for college to continue that state of affairs, so lawyer it was.
Until they made me take an engineering course.
I attended an engineering school (I was working fulltime and paying for college out of pocket, so college meant the closest school I could drive to. Hence, a computer branch of a state engineering school), so that last statement wasn’t as odd as it might otherwise sound. But it was eye-opening. It turns out, I love solving problems. The process of taking what you have a hand, your own creativity, and turning that combination into something that materially affected the world? Fantastic — sign me up.
Eventually, that decision lead me to my current career. And at every turn, I have tried to step into more and more interesting problems. I have done pretty well as a self-taught programmer following that simple mantra —find the most interesting problems someone will pay you to solve. And that is why imitative AI is so irritating.
There are many, many problems that technologists could help solve, climate change among the most pressing. But even something as simple as improving real time machine learning translations would be a problem worth solving, never mind more focus on helping researchers parse through large data sets for answers. We could be doing all of that — but instead, the money pushes people to imitative AI. It is a massive failure of capitalism and society, that we let the failure of capitalism drive so much of our research time and dollars.
Imitative AI is not a success and unlikely to ever be a success. The people pushing it want to pretend that it is so that they can reap financial rewards for, essentially, replacing workers. The problem is that no matter which field you choose — art, or customer service, or coding — the results are underwhelming. Salespeople are being told to dial back their claims lest they over-promise. And these are not problems that are likely solvable — the evidence is pretty clear that hallucinations are inherent to the systems, not something that can be trained out of them. Never mind that they require so much training data that they are likely to require AI generated data, which will likely lead to worse models as they train on AI generated hallucinations. Oh, and AI images are about to swamp and make useless the system for finding and reporting child sex abuses images.
And I haven’t even gotten to the environmental issues.
Venture Capital capitalism has recently been a disaster. It spends all of its time investing in the hype of the day (first social media engagement, then crypto currencies, then NFTs, then the Metaverse, now imitative AI) in the hopes that one of the lottery tickets investments it makes will pay off massively. Instead, for about a decade, we have wasted time and money on fantasies. And we have structured our tax and economic policies in such a way that these companies are never disciplined by the market or by the government. They continue to set money on fire and drag the technology along with them.
It is frustrating. We could be and should be driving our technologist toward meaningful problems, not wasting time and money on society-harming fantasies of instant wealth for capital owners. We can and should do so much better than this. But in order to do so, we need to ensure that our choices are not driven entirely by fantasy economics. We need to spend government funds with less attention to the hype of the day, and more on the actual problems of the day.
Problems are for solving, but there is no point in pretending that we are solving the right problems right now. We clearly are not, and that has to change.