Two new preprints from the lab, First up, causal exploration in humans and LLM's. arxiv.org Human Adults and LLMs as Scientists: Who Benefits from Active Exploration?
arXiv.org e-Print archive arxiv.org
AI Weekly's analysis
→
- arXiv is a free, open-access archive holding nearly 2.4 million scholarly articles across nine major fields.
- None of the materials on arXiv are peer-reviewed by the archive itself, a structural fact affecting how preprints should be read.
- The archive runs as a nonprofit through Cornell University, backed by the Simons Foundation, member institutions, and contributors.
Read full analysis →
View on Bluesky ·
♥ 14
↻ 2
↩ 1
·
7 from the directory shared this ·
101d ago
In the Times this morning- Henry Farrell James Evans and I on why we desperately need the social sciences to understand AI and its impact , just at the moment when the admin is trying to destroy them. www.nytimes.com/2026/09/06/o...
nytimes.com
View on Bluesky ·
♥ 37
↻ 6
↩ 1
·
8 from the directory shared this ·
20d ago
↻
Alison Gopnik reposted
@birdsnfrogs.bsky.social
Do yourself a favor and please read @himself.bsky.social @alisongopnik.bsky.social and collaborators’ works on LLMs as cultural and social technologies www.science.org/doi/10.1126/...
science.org
View on Bluesky →
A very interesting and clear (and reassuring?) piece relating the Farrell et al (www.science.org/doi/full/10....) idea of LLM's as cultural technology to the recent developments in AI Mathematics. terrytao.wordpress.com/2026/09/13/h...
science.org
View on Bluesky ·
♥ 22
↻ 7
↩ 1
·
3 from the directory shared this ·
12d ago
Good piece on why kids beat AI at learning, with Mike Frank and me. Behind a paywall but available elsewhere. www.technologyreview.com/2026/08/24/1...
Kids outlearn AI—and we still don’t know why technologyreview.com
AI Weekly's analysis
→
- A pre-teen has heard about 100 million words; Meta's Llama 3.1 trained on 15 trillion tokens, roughly 150,000 times more.
- The 2024 BabyLM champion, GPT-BERT, beat Meta's Llama 2 70B on some benchmarks despite training on 15,000 times less data.
- Princeton's Brenden Lake trained a model on 61 hours of SAYCam egocentric video and got object recognition without innate biases.
Read full analysis →
View on Bluesky ·
♥ 35
↻ 9
↩ 3
·
3 from the directory shared this ·
33d ago
↻
Alison Gopnik reposted
So pleased that our paper on empowerment and causal models is out and freely available as part of this impressive special issue on world models and AI, with Melanie Mitchell, Josh Tenenbaum, Tom Griffiths and many other stars. royalsocietypublishing.org/rsta/article...
royalsocietypublishing.org
View on Bluesky ·
♥ 105
↻ 34
↩ 0
·
5 from the directory shared this ·
133d ago
↻
Alison Gopnik reposted
@berkeleysocsci.bsky.social
As AI systems continue to advance, Psychology Prof. @alisongopnik.bsky.social points to the original learning model: the child. In a recent Berkeley Talks episode, she explains why children’s curiosity, exploration + learning capacity set them apart from today’s AI systems. Li…
Berkeley Talks: Alison Gopnik on why AI is no match for a 4-year-old - Berkeley News news.berkeley.edu
View on Bluesky →
Really great interview with Michael Pollan on consciousness ( and whether it could be found in current AI) as part of CHris Hayes excellent AI podcast series. podcasts.apple.com/us/podcast/t...
The AI End Game: Is AI Alive? with Michael Pollan podcasts.apple.com
A very interesting and clear (and reassuring?) piece relating the Farrell et al (www.science.org/doi/full/10....) idea of LLM's as cultural technology to the recent developments in AI Mathematics. terrytao.wordpress.com/2026/09/13/h...
Happy, those able to know the causes of things terrytao.wordpress.com
View on Bluesky ·
♥ 22
↻ 7
↩ 1
·
4 from the directory shared this ·
12d ago
An especially clear and insightful piece on AI from Melanie Mitchell Misleading Metaphors and Real Risks aiguide.substack.com/p/misleading...
Misleading Metaphors, Real Risks aiguide.substack.com
View on Bluesky ·
♥ 33
↻ 14
↩ 0
·
2 from the directory shared this ·
16d ago
