parthknowsai
Region: US
Tuesday 30 June 2026 12:29:01 GMT
32369
2633
120
165
Music
Download
Comments
Thor :
Self driving AI doesn’t use LLM technology does it?
2026-06-30 15:33:21
16
🪨 :
Wow so AI is also like Plato’s cave
2026-06-30 12:32:08
55
a cauema :
ok. gimme jeppa then
2026-06-30 16:36:03
5
timbgreen3 :
I suspect it's absolutely solvable I know almost nothing about computers or AI I do know lots about Neuroscience and learning of humans here's the thing they can learn or it can learn the same way humans do now I don't know the model that you talked about at the end of this video but here's what I'm going to propose is likely to work but would obviously need to be tested. we also as humans our brains one of the dominant theories currently is predictive processing that our brain uses what you may as well call statistical probabilistic models to predict what is most likely to happen based on previous experience sounds an awful lot like an llm doesn't it and if that's the case we use the step that our brain also seems to use in the same situation you may you run a predictive model you do your predictive processing you come to a conclusion then you test it against a simulation and or in a perfect situation an actual experiment in the physical world then you have three layers you have the predictive processing which is akin to human brains also you have the simulation to see how close that comes to the probabilistic prediction and then you test that with that third layer of literal real world experimentation to see what actually happens now how close the predictive processing comes to layer layer 2 and especially to layer 3 then gives you ironically a probabilistic rating of how accurate the original probabilistic processing the original predictive processing was and then use that as the reinforcement cuz that's basically what humans do
2026-07-10 05:28:38
0
Clinik :
reasoning is exactly what you described: picking the most likely path based on your understanding the world
2026-06-30 20:35:43
7
Home Bluston :
Two mistakes here: 1.Models are not just trained on written text, they are also trained on videos, including full length films. 2. There is randomization, so llm's do not always produce the most likely output. If we are aiming at human intelligence we only need to compare the amount of llm hallucinations with the amount of hallucinations in human thought (which is substantial).
2026-07-01 08:03:48
5
alejandroechever3879 :
These problems are just philosophical and not realistic evidence of AI limitations. Quite ironic considering it accuses the AI of not dealing with reality 😂. Autonomous driving has been completely solved. The problem is scale, input and compute.
2026-06-30 17:53:50
2
roarc :
all of our senses are a description of the world 🤗
2026-06-30 14:36:36
2
mba0v3r :
"Causal" not "Casual"-- huge semantic difference
2026-06-30 18:01:40
4
justin :
he might be right he may be wrong and someone else might be right. nobody in the comments are qualified to say. But it’s very interesting seeing how angry it makes some people to hear that AI is mot perfection, or not heading there.
2026-06-30 23:22:41
3
user3995620960557 :
grok understands human relationships better than I do and teaches me about them 😁
2026-06-30 13:26:19
1
Dugukfoo :
Autonomous driving is far better than the best human driver
2026-06-30 16:10:51
3
Marc Monroy :
after using AI for some time, you realize its separation from reality and it's limitations.
2026-06-30 14:55:25
6
hehehehe :
Ok, Lecun should give us a better model then
2026-06-30 17:46:49
3
Papsii :
The reasoning definition of LLMs, that's exactly how Humans think😂
2026-07-01 03:58:43
1
guleedInterMilan :
AI needs Quantum physics 😂
2026-06-30 17:34:10
0
Richaz :
JEPA models (at least currently) require a “goal frame” to reason. This limits them in any real world application. Is there a way to solve this?
2026-06-30 12:43:36
0
Pika22 :
I always feel AI has understanding of reality.
2026-07-02 05:07:57
1
user2520225071047 :
Thank you for your great videos!
2026-07-02 16:53:09
1
Wahala :
Well his damn world model is taking for ever to drop something profound. Robotics cool but what else ya got?
2026-07-01 08:38:34
1
Rez :
Remember we learn from the signals from reality not directly from reality as well 😅
2026-06-30 18:51:47
2
Auge Eidos :
No counterfactual reasoning or causal inference
2026-07-01 00:56:21
1
Le Yõ :
My wild take is that we won’t plateau. We’ll need to revise the current structural theories underpinning AI which might lead to small but intriguing discoveries in logic and mathematics. In fact, I’m currently working on it in my spare time and it’s fascinating.
2026-06-30 20:09:29
1
Janani Subramanian :
that’s a fuuny name
2026-06-30 12:42:10
2
user4517038289519 :
best content on Tiktok
2026-06-30 13:33:18
2
To see more videos from user @parthknowsai, please go to the Tikwm
homepage.
