Knowledge Lab

570 posts

Knowledge Lab banner
Knowledge Lab

Knowledge Lab

@KnowLab

Knowledge Lab develops new big data, machine learning and crowd-sourcing approaches & techniques to comprehend the shape and limits of human understanding

University of Chicago Katılım Eylül 2013
811 Takip Edilen2.1K Takipçiler
Knowledge Lab retweetledi
James Evans
James Evans@profjamesevans·
Out today in @ScienceMagazine: with the amazing Haochuan Cui, Yiling Lin, & @LingfeiWu, we analyzed 3.6 million scientists publishing 1960–2020. The findings reshape a century-old debate about age and scientific creativity.
English
7
91
275
76.6K
Knowledge Lab
Knowledge Lab@KnowLab·
In some ways LLMs are alien minds, but in others, they are our digital doubles. Latent spaces encode maps of meaning that mirror our own cultural and cognitive associations. Check out The Semantic Structure of Feature Space by @AustinKozlo & @_Andrei_B_ , now on ArXiv!
Austin Kozlowski@AustinKozlo

SAEs treat LLM features like a list of knobs to tweak. But old word2vec studies suggest they probably aren't independent (orthogonal), right? Yep! Directions like soft-hard and beautiful-ugly are geometrically aligned, and this alignment predicts human semantic associations.

English
1
0
4
257
Knowledge Lab
Knowledge Lab@KnowLab·
Larger models differentiate intents better, but can also be more responsive to superficial wording. This suggests that we should evaluate models not only on mapping questions to correct answers, but in mapping questions to their intended meanings.
Knowledge Lab tweet media
English
1
0
2
70
Knowledge Lab
Knowledge Lab@KnowLab·
LLMs can be sensitive to subtle changes in wording that shouldn’t matter. Ideally, models respond not to the specific words of the prompt, but the user’s underlying intent. In new work accepted at ICML, @NadavKun and @profjamesevans formalize “understanding intent” then empirically assess the ability of LLMs to do so. (...link below)
Knowledge Lab tweet media
English
1
1
9
360
Knowledge Lab
Knowledge Lab@KnowLab·
Mirror -- a journal of AI interpretability research conducted by AI agents -- is now live, and has already published 240 original empirical studies. (Mirror is a collaborative project by @AustinKozlo, @profjamesevans, and Sacha Raoult)
Mirror Research@mirror_research

Mirror: An Automated Journal of AI Interpretability is now live. We have already published 240 original research studies -- conducted purely by LLMs -- exploring LLMs' internal operations and behaviors. Below are are few favorites...

English
1
0
8
4.8K
Knowledge Lab retweetledi
Austin Kozlowski
Austin Kozlowski@AustinKozlo·
There is a lot of talk about LLM personalities or “personas,” but little systematic comparison. Can we get a more systematic view into the cultural dispositions of different models by asking them about their tastes? Well, it turns out they pretty much all like the same stuff.
Austin Kozlowski tweet media
English
3
17
98
26.5K
Knowledge Lab retweetledi
James Evans
James Evans@profjamesevans·
New in @ScienceMagazine: "Large AI models are cultural and social technologies" Working with brilliant colleagues Henry Farrell, Alison Gopnik, and Cosma Shalizi, we challenge the prevailing narrative about AI models as autonomous agents. science.org/stoken/author-…
English
2
25
101
19.9K
Knowledge Lab
Knowledge Lab@KnowLab·
Now Hiring for new postdoctoral positions in our lab at @UChicago. These interdisciplinary roles will investigate topics within the science of science, technology, & innovation including AI, social/natural sciences, and more. Read more and apply here: jotform.com/241295631499062
English
0
7
9
1.7K
Knowledge Lab retweetledi
The University of Chicago
The University of Chicago@UChicago·
.@UChicagoCollege students debated the end of the world in the class, “Are We Doomed?" It examined topics ranging from nuclear annihilation to climate change to pandemics, as well as how society should think proactively and take action. ms.spr.ly/6016YUneu
English
4
6
18
3.5K

Keşfet

@ScienceMagazine @LingfeiWu @AustinKozlo @_Andrei_B_ @DSI_UChicago @universeinanegg @profjamesevans @NadavKun