I found this interesting. He explains how AI functions, and specifically machine translation here:
https://youtu.be/zl99IZvW7rE?t=756
He says that translating from Chinese is found to work the best with bitmap images as input, rather than symbols (characters). So the software reads the shape of the characters.
Geoffrey Hinton on machine translation
- tungemål
- Blue Belt
- Posts: 949
- Joined: Sat Apr 06, 2019 3:56 pm
- Location: Norway
- Languages: Norwegian (N)
English, German, Spanish, Japanese, Dutch, Polish - Language Log: https://forum.language-learners.org/vie ... 15&t=17672
- x 2192
-
- Black Belt - 3rd Dan
- Posts: 3536
- Joined: Thu Jul 30, 2015 11:04 am
- Location: Scotland
- Languages: English(N)
Advanced: French,Spanish, Scottish Gaelic
Intermediate: Italian, Catalan, Corsican
Basic: Welsh
Dabbling: Polish, Russian etc - x 8811
- Contact:
Re: Geoffrey Hinton on machine translation
I think the most interesting thing is how he talks taxing as an assumption that neural nets are comparable to brains and makes the claim that they're arguably more efficient because they have fewer neurons. What he's missing is that if a computer looks more efficient, it's not actually doing the same thing, and with language, "the same thing" is the key to everything. If you're not mimicking native speaker processing, you're not really dealing with the "same language", just something superficially similar.
The thing with A.I. is that it's supposed to be about simulating brains, and right now, we don't have computers powerful enough to do that.
I personally think the fact that the term "artificial intelligence" is so closely tied to neural nets is effectively anti-scientific. Don't get me wrong: neural nets are absolutely astounding, but the problem is that the term "AI" gets investor dollars and attempting to do things in a better way than neural nets means losing the street cred of the term AI.
The thing with A.I. is that it's supposed to be about simulating brains, and right now, we don't have computers powerful enough to do that.
I personally think the fact that the term "artificial intelligence" is so closely tied to neural nets is effectively anti-scientific. Don't get me wrong: neural nets are absolutely astounding, but the problem is that the term "AI" gets investor dollars and attempting to do things in a better way than neural nets means losing the street cred of the term AI.
2 x
Return to “General Language Discussion”
Who is online
Users browsing this forum: No registered users and 2 guests