No. Auto-regressive prediction produces a single path in the tree of possible sequences. At non-zero temperature (stochastic generation) the potential paths form a subtree of the full tree, or rather, a distribution over all paths in the tree. If you threshold all the paths
@ylecun
-
Retrospective Skepticism on OpenAI’s GPT-2 Safety Concerns
By
–
I trolled OpenAI when they didn't initially release gpt2 because oooooh soooo dangerous
-
Yann LeCun Clarifies Origins of LeNet and LeWorldModel Names
By
–
For the record l, Randall picked the name LeWorldModel. And Larry Jackel picked the name LeNet back in at Bell Labs in 1989.
Not me. -
Closed AI Models Benefit From Open Source Without Contributing Back
By
–
Let's be real, all closed models profit from open models WITHOUT GIVING BACK.
-
Evolution Solves Hierarchical Learning Where Classical RL Fails
By
–
Evolution figured out how to get learning to produce hierarchical models for perception and planning.
It had to find a solution to the collapse problem. What you describe is "classical" reinforcement learning. We know that's way too inefficient. -
JEPA Is an Architecture Class, Not a Single Method
By
–
JEPA is a class of architectures, not a method. There were lots of JEPAs before I proposed the name as a unifying concept. For example, Siamese nets (NIPS 1993) are a special case of JEPA (where the predictor is the identity, and the training loss is contrastive as in SimCLR).
-
JEPA Learns Abstractions to Enable Prediction Like Science
By
–
JEPA finds abstractions that enable prediction. This is how intelligence and science work: find abstract representations of reality that allow to make predictions.
-
Research Papers Predict Future Products Years in Advance
By
–
Yet another dude who doesn't realize that before you get a product in your hands, there may be 5 years of technology development preceded by 10 years of fundamental research. Want to know what products will become available in a few years? Read research papers.
-
AI Agents Need World Models and Safety Guardrails to Avoid Danger
By
–
Danger does not come merely from agency, but from agency with no ability to anticipate consequences and with no safety guardrails. The solution? AI agents that can predict the consequences of their actions (world models) and only take actions whose predicted outcomes satisfy
-
BADAS: Advanced Incident Prediction Model Beats State-of-the-Art
By
–
This isn’t just a game; it’s your chance to go head-to-head with a model that beat SoTA in incident prediction. Trained on 10B+ real-world data, BADAS doesn’t guess, it predicts.