It shouldn't be possible to make a living playing sport professionally. It robs people with potential and drive of a future where they do something meaningful.

Ha ha. I was unaware of this!

The security on some of these humanoid robots is below IoT levels of amateur. The companies aren't responding to vulnerability disclosures so nothing is getting better.

It looks like the shuttles dissipate heat by converting it to infrared radiation using liquid coolants and radiators.

I would imagine an orbital data centre would essentially need to be a radiator with a data centre attached rather than the other way around.

In space, no one can map your heat stream

I may be being pessimistic. I am still sore from the two days I spent going through a robotics paper 16 years ago, only to realise that they had simply abused set theory to represent the concept of follow-my-leader.

Some papers need to be printed out... so you can hurl them across the room 🤣

Definitely. Shannon-Bell showed most spoken languages are around 75% redundant.

The input gets tokenized, and the token space can be mapped from the relationship between other internal tokens and the text output.

Sounds very much like a paper from the school of "proving the bleedin' obvious"

They've phrased this badly. They seem to be demonstrating that underlying internal state/output pair is unique for each prompt, which isn't as mad as suggesting every output is uniquely traceable to a specific prompt.

Reverse engineering prompt from internal state is surely a bit trivial?

Many companies forget what innovation actually is. Universities are great at invention. Most companies are great at extracting value. Innovating means coupling invention with a path to value which few are good at. The invention doesn't have to be technical and the value doesn't have to be monetary.

I'm not so sure. Their designs are atrocious (from a purely engineering standpoint.) There's no way any engineer would approve camera-only navigation, hidden emergency release mechanisms, or safety that is based on passing regulation rather than managing risk without pressure from toxic management.

Teenager. Friend brought back bangers from holiday. A larger one turned out to be a small demolition charge. We put it in a jam jar & lit it (the "why" of that decision haunts me). Watched our homemade shrapnel explosive from cover. Shard cut my face just below eye. Astonishing I retained my sight.

Yuck. Sounds horrendous

Boo! Sorry to hear that.

I find the very prospect appalling. Your incredible contributions to UK life, including successful leadership of one of our institutions and your integrity and intellectual honesty during vital debates makes your presence here something to celebrate and reward, not punish.

That's not quite true, and sadly, it's because you're being too positive! There is no minister, junior or otherwise, looking at AI safety anymore. AISI's remit has narrowed to specifically detecting and preventing malicious use of AI. Zero interest in safety, ethics, AI protection nor AI defence.

I think Paddington 2's broad range across the Paddington-Batman spectrum goes a long way to explaining its position as a cinematic masterpiece, as per The Unbearable Weight of Massive Talent.

Ironically TUEoMT is, by definition, unBEARable and therefore a Batman film.

Schindler's List is dark and gritty, but also follows a bid to avoid the unnecessary incarceration of an innocent, thus placing it firmly in the Paddington 2 mould.