Starting with Shaanan Cohney (@shaananc.bsky.social), from the University of Melbourne, on the first topic, defining appropriate technical privacy notions for practice!

Utility in what sense? It does improve reasoning capabilities (both formally and informally).

And not just restricted to ChatGPT! There are a large number of models that are reasoning trained (or based) now.

I’m not sure I follow your key issue? Semantic meaning in context is the main training?

Some of your statements about their training are no longer true. The current generation of foundation models (o3 etc) are explicitly trained on reasoning data. Further, formal reasoning data (using special purpose languages) is increasingly integrated.

Wow this is a gorgeous CV!

!!!!!

Invite the adversaries over for tea?

I was waiting for this thread!! I knew you’d have all the deets.

Social media platforms will face financial penalties starting in the 10s of millions if they don’t implement technical measures to block under 16s from social media!

This was a terrible review.

This is all the external fulfillment I need.

Adding to the call!

Here, I illustrated it

Pick me!

I’d also love to be added even though I’m in Australia!

This was exactly what I was looking for!

Only if you look closely on the slide, the formal specification has a bug!