|
|
|
|
|
|
by WheelsAtLarge
945 days ago
|
|
|
In theory, that's what they are, in effect, they are more than that. Yes they are searching for the most likely next word but there is some randomness to the next word that's picked so that can lead to some very unlikely results. Additionally, they have a huge corpus of words to pick from. The part that people have not emphasized enough is its ability to turn your human request into something that can be used to search the corpus of words. That's the real superpower of these models. People are now able to use their own language without any special training to get very good results. No more need to have a computer programmer translate your thoughts into computer language. It's so good that people are beginning to think that it's sentient. It's not. Also, it's not a path to AGI but it is very useful anyway. |
|
|