Michał Sapka's website

My Gripes With AI

AI may or may not be the future. It’s undoubtedly all the buzz recently. I don’t buy it yet as there are things I either don’t understand or am annoyed with. It’s getting to the point when it won’t be possible to ignore it, but we are still a few years from it if we’re lucky.

  1. no one understands it - it became common knowledge that we are unable to understand what an ML model does. They are too big, and all the values are provided by automatic learning. In regular programs, there is a simple flow any developer can follow and create a mind map. In the ML model, we are presented with a black box that somehow works.
  2. it took advantage of open society - do we know whose data was the model trained on? No, it’s a company secret. What we know is that the set was extracted from the general internet. People put stuff for free to promote themselves or to create some sense of connection with fellow people. The latter is true for this blog. But, at some point, it will be harvested to benefit The Algorithm.
  3. it will benefit very few - let’s be frank here: we don’t need automatic chatbots. Who will they benefit? Users will have some help, as they may not need to google an answer or write a boilerplate. But it will really benefit companies providing the models. The last for-profit technical revolution, the smartphone, made our world much worse, but Apple, Google, and a few others earned billions (vide #7).
  4. it’s an attack on an open internet - why do people put stuff on the internet? For the money, recently. But if you look at the longer timeframe, the primary reason was for others to check it. I applaud people who have web pages without any analytics. I use one to see if anyone reads something I write. When someone enters, it validates me and makes me happy. This is the idea of the open internet - anyone can create something for others. But what will happen when the visitors disappear? We already have a few closed silos (Facebook, Twitter, and so on), so fewer people surf the wider web. And when all the answers will show up in a single place? Who will continue to create when all that there exist is void?
  5. who takes responsibility? - a machine model is not a person (we’re not in GitS yet), but that won’t stop it from hurting humans. What if it provides wrong information? What if (when?) people start losing jobs due to it? If the pace of innovation continues, the social changes ML will bring will be incomparable to anything we’ve seen. (more in #7)
  6. what are the limits? - there were attempts to use ML to detect illnesses like cancer. We don’t do this on a broader scale as it provided too many false positives, and the cons were unacceptable. Will we do a similar thing again? Will anyone say, “stop, this is going too far”? Who would even be able to do that? Scientists are out of the equation now (again, #7), and governments still don’t know what is going on. Sam Altman maybe?
  7. it’s a closed revolution - we are talking about ML because of OpenAI. It was a non-profit, but now it’s just another multi-billion company. We are not looking at the work of scientists and enthusiasts trying to see where they can take us. Somehow, Bell Labs gave us transistors, Unix, C, and a million other improvements. Will it be the same as before? Bell had guard rails from anti-monopoly regulations, yet it was still hell of a show. I see no such thing for current champions of ML. Do I trust them to be people of conviction and value? No, I do not. 1
  8. it will benefit English-speaking countries - since we already speak primarily in English on the web, the models trained for English-speaking crowds will be best. Most of the world does not treat the English language as the primary one.

I see a huge potential, and we will see the revolution soon. But I’m too old to be an optimist. Silicon Valley is a hellhole full of cynics.

The way I see it, we created open internet for the betterment of humankind. Unfortunately, social media tried to destroy it and almost succeeded. But ML models are taking everything we made and using it against us for the benefit of very few.

I’d love to be wrong, but I don’t expect a good revolution. We will be lucky if this won’t destroy the entire society.

 This is not future we were promised, but about which we were warned about
This is not future we were promised, but about which we were warned about [source]