panel discussion

Friday, WFH. I did make it out of the house in the morning to see the hygienist and go to the market. Then a virtual meeting discussing another paper followed by a panel discussion Potentials, perspectives, and challenges in the age of AI. Panellists were Sam Altman (OpenAI), Nicole Büttner (Merantix Momentum), Prof. Dr. Volker Markl (BIFOLD and Head of Databases and Information Systems) and the moderator Prof. Dr. Fatma Deniz (Digitalization and Sustainability at TU Berlin).

Unsurprisingly, all panellists were excited about AI and extremely uncritical. Sam Altman asked the audience if anybody felt cleverer than ChatGPT. He was happy to admit that he was fine with not feeling cleverer than the machine. Bollocks: what's the measure of clever? Humans can do a lot of things beyond producing meaningless texts or images. LLMs are statistical models that remix what they have been trained with. There world is restricted to the training material. Our world is much bigger than the texts and images we have produced. I guess you could say, well just feed it more data. And maybe that is possible but we'd need to figure out how to measure certain things and tag them. The models would also become huge and we'd need even more compute to deal with them.

The next question turned to energy use. Sam Altman stated that AIs consumed less energy than a human. He said something about energy use per token. He may be right with that measure but I don't know what that means. What I do know is that a human brain consumes about 20W and that we cannot simulate a brain with the biggest computers available to us.  

Then he stated that AI will accelerate science production. An although energy intensive it will crack the energy problem and bring about fusion. I don't buy it. genAI remixes existing stuff and checks what sticks. This is all SciFi. More seriously I don't like the view of humanity this portrays: the need to optimise ourselves and to leave the machines to do the stuff for us.

Assuming the AI works as they promise. What is left for us? What is the point of our existence? Who will get access to this technology. Who is paying the energy bill?

Something that came up again and again, was using AI for improving medicines. We don't need new medicines to deal with the problems arising from poverty. Number 1 preventable cause of death are respiratory problems arising from air pollution. We already know what we need to do to improve the health of large proportions of the population.

There might be good applications for AI. I think using it for translations might be quite useful, although you'd need to be good enough in the other language to check it.

It might also be good when accuracy doesn't matter so much. Asking the stereo to play some Motorhead and getting Meat Loaf instead is not the end of the world. 

It will get used when it doesn't matter, like producing more images of kitten and scams. There will be enough people who fall for it. It becomes a lot more problematic when it is used for bureaucracy. It is already bad enough when you are in a minority group. All problems can be solved with the application of enough money. If you don't have any you are stuffed. 

Anyway, enough of a rambling rant. It was interesting to see one of the tech bros close-up.