@MLStreetTalk Exactly, we are now seeing the true effects of AI as a tool that will enhance human capabilities to unforeseen levels and, as with similar tools in the past, will increase both working hours and work in general.
Now that many things became much easier we can do more.
@liminal_warmth PDFs of books require some type of good txt model representation. I use a page based with artifacts representation which produces relatively accurate results consistently
@Old_Samster All of these methods find local solutions at best, which can, and mostly are, very far from the true global solution. But its true that most problems are at their core optimization problems, equilibria in game theory, deep learning, minimum entropy problems etc
@hasantoxr Fail to see the connection between the curse of dimensionality and semantic retrieval; the embedding dimension is fixed, it does not increase with the number of documents.
Can see a case for overlapping clusters of vectors as the number of documents increase asymptotically
@NathanJRobinson AI like any other tool such as books and the internet will be adopted to the maximum by the academia and be employed for the pursuit of of knowledge and free rhinking, which AI cannot do
@ylecun@karpathy discrete bs continuous is the difference or linear bs nonlinear? in many cases you go from one to the other with some transformation albeit you pay in complexity