Tiring. I've come to the conclusion that those ppl who write like that about AGI/AI, including some critical minds, have never taken their time to 'understand' what NN/algos/models actually are. They just work w/ ideas from those who are delusional about it.
I guess this is a joke, but it bothers me. GPT doesn't have any "thoughtful answers" or thoughts at all about consciousness, it assembles texts that have been written by humans on the topic into answers. That's different. (to be continued)https://twitter.com/woj_zaremba/status/1193958784827154432 …
-
-
-
Totally agree. GPT is just a language model. In theory, the worlds biggest language model could convince folks it's smart, but it's simply a "Highly Compressed Humongous Table" twist on McDermott. It's worth applying McDermott's argument to GPT. http://www.cs.yale.edu/homes/dvm/papers/humongous.pdf …
- 1 more reply
New conversation -
-
-
GPT has as much understanding of a topic as Eliza had of human Rogerian psychotherapy.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
You never know whats is a joke and what not with these openai guys...
-
Also, there are multiple books written about this question (e.g. Dennett). It didn't require a rainforest of carbon to figure out that mindless algorithms can appear to be conscious
- 3 more replies
New conversation -
-
-
GPT,BERT or any generalization language models never really understand. They can't create, only mimic. Given gigantic capacity of the models and training data, they simply memorize most sequences possible during training and then spit out sequences which it has seen in the past.
-
This is very close to the truth. It is not exactly repeating the training sequences, but almost exactly repeating. There is some interpolation happening. But still, would anyone believe that interpolation is "understanding"?
End of conversation
New conversation -
-
-
There is any “consciousness” in current
#AI architectures. We still do SGD after all, and we have added encoder/attention/decoder as new Language Model’s approach. This does not imply a conscious reasoning.Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Hawking that "pre-AGI" line of theirs. Must be looking for customers now that MS has invested in them, maybe they feel pressured to deliver? I can't trust anything they say about AI anymore...
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
But Weizenbaum’s assistant would shut the door to her office for private time w Eliza. Apparently some want thought-free & judgement free conversations. But then there is thishttps://m.youtube.com/watch?v=a-SnsqKFHLY&client=mv-google …
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.