OpenAI Co-founder Andrej Karpathy predicts transformers may soon surpass the human brain in capabilities "Transformers have the potential to surpass the human brain in efficiency and memorization capabilities, despite current limitations in data availability and training. As AI technology advances, transformers may become more efficient learners than the human... brain, which operates under various limitations"show more

another very important point by Andrej on: 'why Open Source is essential for future tech'

Andrej compares self-driving car progress to AGI:

Computers have been more efficient than humans in many tasks for decades. Their memory is better, their computation capacity in certain contexts is many orders of magnitudes faster etc. etc. It's never been a question of efficiency but a question of effectiveness. Can computers actually get us to the Thing (capital T) by themselves faster than we can by using computers or other tools. So far the answer has always been "no" largely because computers don't actually have any desires. And without desires they are always tools no matter how efficient. And as long as they are just tools they will never be more useful than a brain, because they will always have to receive their commandments from a brain. Unless you believe that humans are not actually their brains and the brain is also a tool that humans (souls?) actually use, which is also a valid position. Then you could argue that being more efficient than the brain is not about replacing humans, but about empowering humans, which is also interesting.

okay, but even if we take that perspective, there's still a fundamental question about what drives those desires and goals. Where do they originate? Is it from some non-physical "soul" or is it simply a complex interplay of biological and chemical processes within the brain itself?

Facts but saying transformers are better than the brain is a partial truth. The brain is able to receive and analyze data in real time beyond transformers. That matters over memorization in most scenarios.

But the real world is messy and unpredictable. it throws curveballs that transformers haven't been trained on. And that's where the human brain shines. we can take in new information, analyze it in context, and make decisions on the fly, all in real-time...

Very true. But what if AI's progress depends more on *data* than slightly better algos?

ASI could most likely result in a World of near-zero ownership as it replaces personal responsibility with a highly-intelligent automated better one where everyone would benefit together without really doing anything.

The key lies in how we choose to implement and integrate ASI into our society. if we simply hand over all responsibility to ASI without careful consideration of the potential consequences, we might end up creating a world that's efficient but ultimately unsatisfying.

If you're Neurotypical, try playing this at 0.25 speed and you'll be fine.