The UK Leads by Example: Open Source for Safer AI 🇬🇧
We commend the AISI’s open-sourcing of Inspect to promote #AIsafety . This aligns perfectly with our call for more open source innovation *by* the public sector. Learn more: linuxfoundation.org/research/europ…

6/ I am personally also very attracted to projects that attempt to truly open up the full process of training AI models, for example GPT-NeoX, OLMo or Pythia which all have publicly available training data and OSI-licensed training and evaluation code and model weights.

🫰What did the new influx of funds change in regard to sustainability, security & mitigation planning in developer communities? (Shoutout to Cailean Osborne (review board) & Atlantic Council for initial assessments: dfrlab.org/2024/04/18/mor… & arxiv.org/abs/2404.06484) (5/x)

Researchers depend on #OSS as digital public goods but their development and maintenance are often unfunded. In this new proposal with UKDayOne, Cailean Osborne Oxford Internet Institute makes the case for the UK Gov to set up OSS Fund to support maintenance and spur research & innovation👇🏼

📯 Upcoming #GeneralElections2024 present an opportunity for UK Gov to support development & maintenance of OSS used widely in research & innovation, following examples such as Sovereign Tech Fund in 🇩🇪

👇🏼We propose UK Gov establishes an OSS Fund (£12.5/y). Let us know what you think!

📣 NEWS: our very own Cailean Osborne has contributed to a new call for the UK Government to establish an OSS Fund, similar to Germany’s Sovereign Tech Fund. This fund aims to support digital public goods used in research and industry, fostering innovation through open source software.

Google, who massively benefits from Python, stops investing in its maintenance and improvement!? 😱

Is that true, Jeff Dean (@🏡)? How is that justified?
Digital commons, as OSS, work only if all pitch in, in particular the privileged.

🔥 Yet another mic drop from Hannah Rose Kirk: PRISM, an alignment dataset feat. sociodemographics & stated preferences from 1.5k ppl from 75 countries who gave 68k responses in 8k convos. Available on GitHub & HF. Kudos to Hannah & team for their excellent work 👏🏼👏🏼

Published in Nature Machine Intelligence today, our new article explores the trade-offs of personalised alignment in large language models ⚖️ Personalisation has potential to democratise decisions over how LLMs behave, but brings its own set of risks...
nature.com/articles/s4225…

A very sad day for technology as yet another an amazingly impactful organization folds 🥺

.NTIA recently closed a policy consultation on openly available AI models. Now that I’ve recovered from the writing sprint, here are a few takeaways:

In support of developers building AI in the open, we submitted a filing responding to the NTIA request for comment on the risks, benefits, and policy implications of widely available model weights. Here’s what we said👇github.blog/2024-04-10-hel…