Grappling with the ways of neural networks, both biological and artificial 🧠 🤖
London
Joined April 2016
- Tweets16
- Following164
- Followers137
- Likes5
New blogpost! Current LLMs think in words, but what if they might one day think in opaque vectors? Google DeepMind’s recent DiffusionGemma a priori has the ability to do such “latent reasoning”. Together with @NeelNanda5, we investigated whether it is in fact harder to monitor.
Jan Bauer retweeted
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.
Jan Bauer retweeted
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
journals.plos.org/plosone/ar…
Jan Bauer retweeted
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. youtube.com/watch?v=E5hATeCZ…
Jan Bauer retweeted
Ich bin @rezomusik Fanbase. Heute gemeinsames Video gedreht. Die Antwort auf Philip A… Kommt bald.
Jan Bauer retweeted
💯 reminds me of MAML meta-learning (arxiv.org/abs/1703.03400) 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.
