Jan Bauer
@japhba
Followers
71
Following
5
Media
0
Statuses
14
PhD student at @GatsbyUCL and @ELSCbrain. Studying the mechanisms behind of intelligence 🧠🤖
London
Joined April 2016
Had a lot of fun thinking about the neural basis of cognitive flexibility with Alex, Kai, and Ali! Excited to see what other work bridging ML and Cognition will come out of the Analytical Connections summer school 🤖🧠
Thrilled to share our NeurIPS Spotlight paper with @japhba* @akaijsa* @SaxeLab @summerfieldlab @Ahummos*! We study how task abstractions emerge in gated linear networks and show that these abstractions support cognitive flexibility. https://t.co/OHlg5hnQVM
0
2
8
pip ⊂ conda ⊂ docker Choose the minimum feasible isolation, but no less (Lightweight but Brittle vs Involved but Stable).
0
0
0
Couldn't agree more: we still don't have good solutions for scalable abstraction/concept learning from "raw" data (be it language, vision, or other sensory data). Solving this would likely unlock important new capabilities.
For anyone interested in future LLM development One of the bigger unsolved deep learning problems: learning of hierarchical structure Example: we still use tokenizers to train SOTA LLMs. We should be able to feed in bits/chars/bytes and get SOTA Related: larger context window
4
15
119
The longer people are in academia, the more they realize that when reading papers it's best to ignore Intro, Discussion etc. and just look at Methods and Results https://t.co/EzQL7WI71Q
196
1K
5K
While remote, both science and hiking were very much on the edge! ⛰️
Had the most amazing time at the Mathematical Methods in Computational Neuroscience Summer School in Norway 🧠 Incredible speakers, organizers, peers, science conversations, food, views, hikes. So beyond grateful to attend 💛
0
0
6
We often discuss if the brain does something like gradient descent. Here is me discussing the issue for @OpenNeuroMorph. I focus on exposing the weaknesses of the hypothesis as well. Should be particularly useful for newly interested people.
3
28
87
Ich bin @rezomusik Fanbase. Heute gemeinsames Video gedreht. Die Antwort auf Philip A… Kommt bald.
687
230
9K
@gdb 💯 reminds me of MAML meta-learning ( https://t.co/H9CIfVdxHd) where the objective is to find weights of a network such that any new task finetunes fast. In Software 1.0 land, equivalent is writing code such that any new desired functionality is simple and doesn't need a refactor.
7
12
145
Which systems come to mind when reading this? A lot seems to be vaguely familiar in terms of building complex software and testing…
Read these 3 pages. I post them every so often because I think it is some of the tightest, wisdom-packed writing on managing complex systems ever. And almost every system is a complex system today, which is why cascading failures are swirling around us. https://t.co/pnNlFLvVdz
0
0
0