There is too much hype, or should I say ‘hopeium’ going on around the generative AI field discussed in social media. Seeing x100 performance with a combination of previous technology with new processing methods and better hardware is quite impressive as most of us would never have imagined this much power, BUT, these innovations will NOT solve all the problems in the world, and no, evil robots will not come after you anytime soon.

Yes, Sora and Bard are amazing, and tens of billions of parameters are also very impressive.

This will evolve, but there isn’t a guarantee that the same speed will continue in the development of these projects if the metric is the ‘originality’, or ‘new ways’ of producing gen-AI outcomes/results.

Many people try to relate to new AI models as if they are interacting with a human clone, try to find traces of advanced logic, human feelings, and such, which is currently very much impossible.

Choose your favorite large language model, use an excessive amount of processing power, and try to do a very long logical reasoning experiment. You will realize patterns and certain weighs in the model, that will make you go “ohh so that’s why it keeps doing this correct/wrong”.

Therefore, things will improve much more, but most probably never in my lifetime, we will see evil robots equipped with supercomputers try to control the milky way.
February 21, 2024