Explore AI
Region: GB
Wednesday 14 February 2024 09:00:00 GMT
180031
6177
443
416
Music
Download
Comments
Tonio Loewald :
The idea that intelligence is “just autocomplete” is hilarious.
2024-02-14 19:23:07
88
Kayse :
"Understand" what has been said or "assign mathematical value" to what has been said to predict what to say next?
2024-02-14 18:01:19
18
skippitydoo :
It’s still autocomplete. Why does it know whether trophy or suitcase? Because of the millions of examples of similar sentence translations!!!
2024-02-15 15:39:36
17
drurywalker :
This is wildly misleading at best. The anthropomorphizing of gradient descent algorithms can do real harm.
2024-02-14 17:07:23
22
anewone :
Great example. Good guy
2024-02-14 13:43:29
19
elviso890 :
it's just super auto complete basically guessing what comes next based on a trained model.
2024-02-15 14:12:58
6
Karma :
What he just explained in layman's terms is what is called the self-attention mechanism in transformer neural networks for anyone interested.
2024-02-15 07:32:04
11
donkih :
So could AI do the correct translation?
2024-02-14 23:18:19
1
🦕 ORyan :
🤨 He is mixing soo many things & these models don't "understand" it's math. And translation models work slightly differently but still math.
2024-02-15 15:12:58
1
tinghiutuan :
A lot of experts in the comments teaching geoffrey hinton how to explain AI.
2024-03-02 04:37:24
4
bob.does.food :
He misuses the word “understand”. The AI is not “understanding”, it’s choosing a word & its relationship to other words based every occurrence in data
2024-02-14 17:42:09
2
rohan :
The key word is “fit” that
2024-03-09 21:30:08
1
Palizi :
Everyone in the comments pretending they understand LLMs better than Hinton 😂
2024-02-16 08:32:01
2
Mazza :
« the glass fell on the table and it broke ». You can only translate into Fr if you know that glasses are more fragile than tables.
2024-02-14 21:16:51
1
Zoney :
He explained the classical attention mechanism, for anyone who wants to do more reaearch
2024-02-14 17:39:28
3
marcos :
the guy is like "yeah yeah" but understood nothing
2024-02-16 01:44:47
2
_shitpostBotTenThousand🦑 :
people saying "it's just gradient descent with back propagation", how do you think human minds work? do you think they're just magic? we use analogous processes, we are neural networks
2024-03-20 07:23:30
1
eli goren :
great explanation
2024-02-15 16:37:06
1
Tuesday Morning :
It's an amazing statistical model with a large corpus of information. It doesn't think, it uses existing information to predict answers to prompts.
2024-02-15 13:52:47
9
martinwalsh340 :
AI has no understanding of spatial relationships. It simply chooses the most probable , after being fed a truck load of instances
2024-02-14 22:31:21
13
the_enid :
Knowledge isn’t understanding.
2024-02-14 16:08:34
6
Vilius :
AI still does not understand per se, just accounting for more relations than just dictionary links.
2024-02-14 15:33:50
21
atlasbernardin :
imagination is unpredictable
2024-02-14 11:52:10
5
james :
There is no understanding in models … it’s all about using weights to generate output. Weights are defined by training data.
2024-02-14 13:46:28
8
To see more videos from user @ai.explored, please go to the Tikwm
homepage.
