4 ms·
I'm familiar. It's engineered to attend , attend , attend and fill in blanks. Nowhere is there any consideration for semantics
by cblconfederate 6y ago
I'm familiar. It's engineered to attend , attend , attend and fill in blanks. Nowhere is there any consideration for semantics
- tiborsaas 6y agoWeird, when I spent lots of time with Philosopher AI all I've seen is semantics in action. How else do you expand on a topic based on a simple question and keep the context? How can it then compress text to summarize it well? How can it obtain high level answers for complex questions?
- cblconfederate 6y agoTo answer these questions one would need to begin with rigorous definitions of 'semantics', 'context', 'memory', 'complex' etc. What we know is that attention models are heavily trained associative memories. Anecdotal perceptions are deceiving, as humans we are often prone to see human/animal-like behavior where it doesnt exist (e.g. cgi/games).