Slowing down AI + Twitter Open-Sources Its Algorithm
Quite a bit happening this week.
Twitter open-sourced their recommender algorithm - you see it here and listen to the Twitter Spaces recording here.
This letter (signed by Yoshua Bengio, Stuart Russell, Elon Musk, Steve Wozniak, Yuval Noah Harari, Andrew Yang, amongst others) was released, asking for a 6-month pause on training AI models more powerful than GPT-4. This spurred quite a bit of discussion:
This article by Eliezer Yudkowsy argued 6 months isn’t enough
Timnit Gebru and others argued that this is all hype, and we should focus on preventing increased social inequities from AI
And a few others shared their perspectives too.
I don’t feel I’ve thought or researched this enough to share a nuanced opinion, but I have a general feeling we’re a bit out of our depth 😅
Even though I work in AI, I feel one step removed from a lot of this discussion. I’m still focussed on how (1) more people can build ML and data science skills and (2) how we can use existing tools to improve our work.
So this week I wrote an article on learning data as a busy person (such as alongside a full-time job, having kids, etc): The Busy Person's Guide to Building Technical Data Science Skills (in Six Steps).
I also wrote a topic summary on Methods for interpreting and explaining AI models, while revisiting the topic this week.
Other updates
This week I ran a “crash course data science” session for the BiteLabs Fellowship. You can check out slides from the talk here.
I also came across this awesome substack today, which is full of useful tips for data scientists looking to level up the quality of their code.
About Me
👋 Hi, I'm Chris Lovejoy.
I'm a medical doctor 🩺 -> machine learning engineer 👨💻 -> start-up founder 💡
I'm on a mission to improve how we manage our health - and share my learnings and experiences here, on my personal website and on YouTube.
I also throw in my favourite things from the internet, and the occasional joke (humour is work-in-progress).