Moritz Boos

Mastodon: @[email protected]

[Nadine Spychala - https://mstdn.social/@nadinespy @NadineSpychala]

Informative & down-to-earth.

"My sense is that for all the fanfare and talk of acting for the good of humanity, OpenAI is more of a Silicon Valley, uncapped, for-profit company, than the people involved like to admit!"

https://t.co/4eLseTLhg2

[Nadine Spychala - https://mstdn.social/@nadinespy @NadineSpychala]

I’ve been writing a blog post on collaboration in science – it’s long, but there’s a summary/TL;DR (and a TL;DR of the TL;DR), and some nice photos to look at, too: https://t.co/v9dwOuMxtS

Some bits on what it’s about (mostly points from the post’s summary):
1/13

[Moritz Boos @mjoboos]

What a line up! 🤯

[Nadine Spychala - https://mstdn.social/@nadinespy @NadineSpychala]

Save the date for a one-day **online** symposium on 21st October 2022:

Rethinking Computational Approaches to the Mind - Fundamental Challenges and Future Perspectives.

Find more info - including how to register - on the event website: https://t.co/moU7U1G2sh.

1/2

[Moritz Boos @mjoboos]

One of the biggest differences between academia and industry is regular code review. This blogpost explains some of the obstacles to establishing this practice and how to overcome them.

[Moritz Boos @mjoboos]

This work was done during a visit to the Neural Acoustic Processing lab - initial results were interesting, but then I left academia. Instead of sitting on this analysis forever, I'm releasing the code as a stand-alone library so others can apply use it on their own data. 7/7

[Moritz Boos @mjoboos]

I wrote a library for easily fitting attention-based deep auditory encoding models to predict brain activity - read 🧵below to learn when this technique might be useful (TLDR: less overfitting and more interpretability) and check out the library here: https://t.co/KbYWu2oaoq 1/7

[Moritz Boos @mjoboos]

Some of the auditory features that are "uncovered" that way are all-time auditory neuroscience favorites such as word onsets/offsets and longer periods of silence. 6/7

[Moritz Boos @mjoboos]

@realdaubman Awesome!

[Subbarao Kambhampati (కంభంపాటి సుబ్బారావు) @rao2z]

There are many ways of interpreting last week's developments in large *learned* models (#LLM's) n #AI (#PALM, #DALLE2 etc), but in the long term, I think, we might remember this as the week #AI finally became an ersatz "natural science".. 🧵 1/n

[REG @ Turing (active on BlueSky and Mastodon) @turinghut23]

GREAT NEWS 📢: From today our team is hiring on a rolling basis!! Multiple amazing Research Software Engineer/Research Data Scientist permanent positions within the Research Engineering Group at the @turinginst 🤩🤩
https://t.co/LnGW7JNQX5

[Moritz Boos @mjoboos]

Good post about having a modern default data science setup - this takes most of the packing pain away. Only thing I do differently is using poetry for virtual env management as well.

[Richard Ngo @RichardMCNgo]

A good example of the agency bottleneck: there are many very talented scientists explicitly motivated by advancing scientific progress. Yet only tiny proportion work on reforming the scientific system, which would contribute far more than almost any object-level research.

[Sebastian Raschka @rasbt]

In https://t.co/bcRiIqGqzN, researchers found that pre-training CNNs the same way as transformers leads to competitive performance on NLP tasks!
Nice insight & paper I totally missed last year (just stumbled upon it thanks to @seb_ruder's excellent review https://t.co/zAtk7tHLFD)

[Parietal, INRIA @Parietal_INRIA]

We’re hiring ! https://t.co/0HHJfHks6P We're hiring a developer for t...