As per my understanding of how LLMs and human intelligence work, neural networks and enormous data sets are not gonna pave the pathway towards ASI. I mean, look at how children become intelligent. We don't pump them with petabytes of data. And look at PhD students for instance. At the start of a PhD, most students know very little about the topic. At the end of it, they come out as not only experts in the topic, but widened the horizon by adding something new to that topic. All the while reading not more than 1 or 2 books and a handful of research papers. It appears the AI researchers are missing a key link between neural network and human intelligence which, I strongly believe, will be very difficult to crack within our lifetimes. Correct me if I'm wrong.
[link] [comments]