If it can plan, it can predict the consequences of its actions and envisage multiple outcomes, then pick a sequence of actions that accomplishes the goal.
That means it has a world model that can make real-world prediction.
That means it's not an LLM.
@ylecun
-
Planning, world models, and why LLMs fall short
By
–
-
Frontier model economics and distillation push toward open models
By
–
– training frontier models is expensive
– distilling a frontier model is cheap
– this market force alone tells us that frontier foundation models will be free/open.
– there are others. -
Debate on LLM limits and path to human-level intelligence
By
–
1. LLMs have always been useful
2. LLMs are unsuitable for the capabilities I listed, which is why LLMs are not a path to human-level intelligence. -
AI progress makes systems neuroscience more exciting
By
–
I feel like the progress of AI is actually making systems neuroscience more exciting. It's not like current AI has a particularly good angle on physical intuition, common sense, world models, planning/reasoning, motivations/drives, emotions, etc. We may now have the tools to
-
Reasoning requires search, which LLMs lack
By
–
Good piece.
True reasoning must involve a search.
LLMs don't possess this capability. -
Critique of ‘general’ AI claims lacking physical-world understanding
By
–
Folks who claim their AI is "general" even though it doesn't understand the physical world as well as a house cat sure have a very strange definition of "general". Just sayin' No need to try to connect this to some sort French culture stereotype you just pulled out of your
-
Data requirements and sample efficiency for Level-5 self-driving
By
–
We have billions of hours of training data for driving: video+sensor inputs together with steering wheel and pedal actions. Despite that, we still don't have Level-5 self-driving cars. It not just about running time it's about data requirements and sample efficiency. (Before
-
Two years later, still far from human-level AI
By
–
That video is 2 years old.
Two years later, we are still far from human-level AI. Sure, AI has superhuman performance in a number of tasks (particularly in mathematics and coding and answering questions with known answers). But where is my Level-5 self-driving car?
(Before you -
LLM limits: physical intuition and domestic robots
By
–
Sooo where is my domestic robot? Where is my L5 self-driving car? I can fine-tune an LLM to answer any question about the physical world. But that doesn't mean it has enough understanding and physical intuition to predict what a 6 year-old can predict.
-
Pro-intelligence or anti-intelligence: is more AI-driven intelligence good?
By
–
Question for everyone :
Are you pro-intelligence or anti-intelligence ? Do you think more intelligence in the world is intrinsically good ? Or do you think more intelligence is intrinsically dangerous ? AI amplifies human intelligence.
It makes PEOPLE smarter and more