To finish the year with something interesting, here's a short video about AI but well worth listening to.
I have been involved with transformers for over 10 years (It is not as new as people say it is. It is only the way we use them which is new.) and I truly do not understand why they allow for the kind of emerging capabilities that the video discusses. In fact, nobody knows except for the fact that they are necessary steps to become more efficient at predicting the "next" word. That's it? That's what "intelligence" is? The ability to predict accurately, and therefore modelize the world to achieve this goal?
This in the end is what is truly frightening. I, and many other people, were expecting that we would have to build a model of the brain to get anywhere close to AGI (Artificial General Intelligence). But in the end, it may not at all be the case... In which case we are much, much closer to AGI than we thought!
No comments:
Post a Comment