The Underpants Gnomes of AI

There is this diffuse fear around that AI will take over, destroy humanity, and generally open a can of Terminator on all of us. This fear is very reminiscent of classic South Park’s underpants gnomes. These gnomes have a 3 part plan to make profit: 1. steal underpants 3. make profit. Step 2 is very obviously missing. Obvious to anyone but the gnomes.

The current fear of super AI goes along the lines 1. write software which can write (often grossly erroneous) text based on prompts and its training set from the internet; make over-saturated not-quite right images based on text prompts; and do a few other neat, fancy statistical tasks. 3. Software will take over the world.

Here are the original underpants gnomes:

And here is a another recent example of AI underpants gnomes:

In the VOX article, some managers ..

…announced their departures from OpenAI, the maker of ChatGPT, on Tuesday. They were the leaders of the company’s superalignment team — the team tasked with ensuring that AI stays aligned with the goals of its makers, rather than acting unpredictably and harming humanity.

thank you for keeping me and the rest of humanity safe from this grave threat. This feels like an old lady telling me she prays for me.

The aforementioned article goes into detail of some office intrigues at OpenAI, which I am sure have nothing to do with keeping the certainly well paid jobs there. There isn’t really any detail  about the supposed threats to humanity, and how the super AI/general AI will turn us all into slaves. Again.

Also: Not Enough Juice

In a recent paper by my mate Jay Coggan and myself, we show that there will not ever be enough energy available to build a “general artificial intelligence” or a super-AI. By orders of magnitude. I haven’t heard or read an even somewhat convincing counter-argument against our arguments. So even if there are not enough managers in an American software company with a title saying something with “safety”, the artificial superintelligence will not happen. In principle. Because of: physics.

So. Worry about anthropogenic global warming and biodiversity loss, war & poverty. Ideally, do something against these problems, too. A bunch of computer programs which can re-write texts they found on the internets? Or make shitty images from scratch? Probably safe, now and in the future.