Active learning
Learning in which the learner actively chooses actions or experiments to gather the most informative data, rather than passively observing.
As defined in 308 | Alison Gopnik on Children, AI, and Modes of Thinking
Other definitions
A machine learning strategy where the system chooses which new data to collect next to maximize what it learns, accelerating model improvement.
Alex Wiltschko - Giving Computers A Sense Of Smell - [Invest Like the Best, EP.415]
Where it comes up · 4
- Alex Wiltschko - Giving Computers A Sense Of Smell - [Invest Like the Best, EP.415]3 claims
- Ilya Sutskever: Deep Learning | Lex Fridman Podcast #942 claims
- John Schulman (OpenAI Cofounder) — Reasoning, RLHF, & plan for 2027 AGI1 claim
- 308 | Alison Gopnik on Children, AI, and Modes of Thinking1 claim
Most important claims · 7 of 7Search all claims for “Active learning” →
- Being able to both read a smell into data and write it back out creates a virtuous cycle in which every run produces new data, and if the next day's experiments are chosen to teach the system more, this is active learning that makes AI systems improve quickly.Alex Wiltschko - Giving Computers A Sense Of Smell - [Invest Like the Best, EP.415]
- Even very young children are active learners who do not merely observe statistics passively but actively experiment on the world; what looks like a toddler getting into everything is functionally equivalent to being a good experimental physicist, and this active intervention may be the key to why children learn so much, so fast, on so few calories compared to AI.308 | Alison Gopnik on Children, AI, and Modes of Thinking
- Models have some calibration about what they know—which is why they don't hallucinate that badly—and that same ability to know what they know could automatically give rise to active learning, the capacity to identify and seek out knowledge gaps.John Schulman (OpenAI Cofounder) — Reasoning, RLHF, & plan for 2027 AGI
- The binding constraint for AI progress in a new sense like smell is data rather than model architecture, so the real moat is the infrastructure that generates labeled physical-digital data.Alex Wiltschko - Giving Computers A Sense Of Smell - [Invest Like the Best, EP.415]
- Licensed scent datasets were not AI-compatible, so Osmo had to build its own data pipeline from the ground up, including a staffed sensory labeling operation and sensor labs running around the clock.Alex Wiltschko - Giving Computers A Sense Of Smell - [Invest Like the Best, EP.415]
- Language models will likely continue to improve with scale and will probably show larger results in the future, and further directions include exploring how models can use intelligence to decide what data to study rather than indiscriminately memorizing all data, similar to how humans selectively learn.Ilya Sutskever: Deep Learning | Lex Fridman Podcast #94
- Active learning (selecting which data to train on) will be a major breakthrough area, though it requires problems that actually need it to succeed; artificial tasks on MNIST won't convince anyone, which is why active learning research might be more fruitful when driven by real problems requiring it.Ilya Sutskever: Deep Learning | Lex Fridman Podcast #94