AI labs should watermark text output. A less visible em dash. TBH it likely already exists and will be found in 20 years, like printer tracking dots. en.wikipedia.org/wiki/Printer_t…

Not everyone thinks only of themselves.

It doesn't appear to contradict. It completely contradicts and destroys the IDF account. Words matter.

Behind the scenes at the iconic Maida Vale studios recording with the BBC Concert Orchestra for BBC Radio 3 Unwind.

I predict a laptop with em-dash key will be released within the decade. 😅

📷 oh damn. just realised we've been nominated for a The Radio Academy award for this one

jack Nationalise AI companies then. It's trained on our work.

Assuming the treatment isn't novel, then isn't this a fun example of a needle-in-haystack eval? 😄

#sudosocial a digital relationship with an ai that feels reciprocal and personal but is fundamentally artificial. word of the year 2027

DeepMind are in a different league to the rest

In the future, we'll vote for models and weights, not leaders and parties.

Pretty sure this Spitfire Audio x Splice deal is an AI training data play. Raw samples are more valuable for AI development than completed songs, especially since the machine has ingested every song on Spotify etc.

Garry Tan AI isn't trying to be a genius - it's trying to become a genus.

Here we go! 🎹🎙️🪈 The themes I composed for BBC Radio 3 Unwind with BBC Concert Orchestra have been nominated for Best Sonic Branding at the The Radio Academy & I'll be discussing it all with Granny Eats Wolf at the The Podcast Show! youtube.com/watch?v=25sEaY…

I DON'T KNOW HOW TO DEFEAT THE X ALGORITHM & WHY IT DOESN'T SHOW MY SHIT TO ANYONE. MAYBE I NEED TO TYPE IN CAPS OR INCLUDE MORE PICTURES OF CAT 🐈‍⬛

"My track record sucks and I have genuinely no idea" but here's my take anyway 🧐😄

I weighed up the delights and irritations of visualised podcasts for The Guardian. Do they represent a brave new audiovisual frontier - or are they just “crap telly”? theguardian.com/tv-and-radio/2…

agi will require ubi.

Wow. Since launch 24 hours ago, we’ve served 85,000 streams of interactive video across the world. And it’s only accelerating! What’s wild is our model generated 250 million(!) video frames to meet the demand. Blown away by the excitement, TY. Let’s make more magic together ❤️