Continual learning
The ability to keep learning new skills and knowledge over time on the job, rather than being frozen after training.
As defined in dwarkesh_ilya_sutskever_were_moving_from_the_age_of_scaling
Other definitions
The ability of an AI system to learn and adapt from new experiences within a given context or over time without forgetting previous knowledge, rather than requiring complete retraining; a key research frontier where systems accumulate experience like human employees do
DeepSeek's Invisible Founder Just Talked for 4 Hours. It Leaked.Models that keep learning from new data after deployment rather than being fixed after a single training run.
AI Market Jitters
Where it comes up · 25
- Dwarkesh Patel Podcast - 8 Predictions for the Era of Continual Learning9 claims
- Dwarkesh Patel Podcast - The next big breakthrough will be AIs learning on the job7 claims
- DeepSeek's Invisible Founder Just Talked for 4 Hours. It Leaked.6 claims
- Dwarkesh Patel Podcast - The avalanche that hit math is coming for AI development5 claims
- What does the next training paradigm look like?4 claims
- Richard Sutton – Father of RL thinks LLMs are a dead end4 claims
- Ilya Sutskever – We're moving from the age of scaling to the age of research4 claims
- An audio version of my blog post, Thoughts on AI progress (Dec 2025)4 claims
and 17 more
Most important claims · 10 of 76Search all claims for “Continual learning” →
- When continual learning is solved, there will be a huge discontinuity in the value of these models: even without a software-only singularity, AIs deployed broadly across the economy learning on the job—and amalgamating their learnings across all copies so one AI learns every job—could rapidly become a superintelligence even with no further algorithmic progress, producing a broadly deployed intelligence explosion.Why I don’t think AGI is right around the corner
- The lack of continual learning is a huge bottleneck: there's no way to give a model high-level feedback, so you're stuck with the abilities you get out of the box; tweaking the system prompt does not produce anything close to the learning and improvement human employees experience.Why I don’t think AGI is right around the corner
- There are two paths to fast adaptation: continual learning (continuously updating model weights so it adapts to new information rapidly, like an employee who learns from feedback) versus in-context learning (loading a huge context window with extra information each prompt); Lambert is more bullish on simply providing very good context, since updating weights per person is too expensive.Ai Sota #2026
- Even if Sutton's Platonic ideal is not the path to the first AGI, he identifies genuine basic gaps that go unnoticed because they pervade the current paradigm—lack of continual learning, abysmal sample efficiency, and dependence on exhaustible human data—and if LLMs reach AGI first, their successor systems will almost certainly be based on Sutton's vision.Some thoughts on the Sutton interview
- Existing intelligence-explosion scenarios (software-only or software-plus-hardware singularities) neglect what will be the main driver of improvement atop AGI: continual learning, since humans become more capable mostly from experience in the relevant domain.An audio version of my blog post, Thoughts on AI progress (Dec 2025)
- Coding made fast progress because the codebase functions as an external scaffold of memory: reading the codebase into the context window gives the model everything a human would otherwise need months to learn on the job, so what we think of as 'learning the job over six months' the model accomplishes within its context—an advantage other economic activities may lack.Dwarkesh Patel Podcast - Dario Amodei — "We are near the end of the exponential"
- Under the big-world hypothesis, agents become useful on the job by encountering their particular, unanticipatable part of a world too huge to pre-load; the LLM dream of teaching the agent everything in advance fails because of all the idiosyncrasies of a particular life, so the agent must learn it along the way into its weights (not just a context window).Richard Sutton – Father of RL thinks LLMs are a dead end
- If AI progress totally stopped today, less than 25% of white collar employment would go away, because while many subtasks get automated, the inability to improve over time and learn preferences means firms still hire humans; without progress in continual learning we'd be in a substantially similar position for most white collar work.Why I don’t think AGI is right around the corner
- If AI achieves continual learning, meaning it dynamically updates its weights in real time, there would be a very fast takeoff, because current models need a million trial exposures and training-run patches while humans learn from one hot-stove experience.Gavin Baker - Watts and Wafers - [Invest Like the Best, EP.473]
- Critical capabilities are still missing from current AI systems; continual learning—the ability to learn after training and deployment—remains absent despite being central to human cognition.Demis Hassabis: Why AGI is Bigger than the Industrial Revolution & Where Are The Bottlenecks in AI