Was reading with great interest the conversation between Yann LeCun and Arnaud Contival. The debate is about ChatGPT and the intelligence it really brings.
Not a new topic:
on one hand, it is surprisingly efficient on some tasks like retrieving information already existing and summarising in a language that most can understand,
on the other hand, it is really approximate, has a lot of memory but hardly mimics intelligence.
Truth is that ChatGPT is fun and may help when searching for information and writing quick something. What can be retrieved is more accurate that typical Google. Output also is more friendly :-)
All good to the extent, we discover that it is not always very accurate and sometimes a bit fuzzy. The only problem is that :
- we use it when we do not know the right answer
- it does not say when it is fuzzy and just presents things as if they were perfect.
I guess, it is wise not to expect too much. Also, improved solutions will pop up quick and replace it. By the way, some of them will be funded by advertising. So, I like to think that this is just a great step towards a better solution.
My only question is whether all that really makes sense and fits our sustainability objectives? Would dream having access to a sort of ChatGPT burning less energy than a human brain. After all, this is one of the targets. ... No ?