Apparently Palantir has been using extra high mode unnecessarily.
EXC: British govt is searching for a UK alternative to Palantir, as the PM’s “buy British” push triggers hunt for homegrown replacement for the controversial US software group in NHS and MoD
Ministers have identified a handful of UK technology companies they hope might be candidates to take over the health service’s Federated Data Platform, as they consider applying a break clause in Palantir’s contract in March next year, people familiar with the matter told the FT
A defence minister has also told parliament that the govt is sounding out alternative suppliers for its MoD contract
w/ @Smyth_Chris & @Laura_K_Hughes
ft.com/content/9a6e17fe-55e4…
One day serious car crashes will be like airplane accidents, super rare, resulting in a big investigation and newspaper headlines. I hope that day is soon.
As if inequality wasn’t bad enough now the NHS is letting billionaires jump the queue for colonoscopies.
my allocated Tax Compliance Mandarin is already in touch and he wants to know EVERYTHING bloomberg.com/news/articles/…
Pacing the frontier of regulation.
Wow.
Eleven EU member states, including three of the largest (Germany, Italy and Poland), have told the EU to stop making new laws for at least a year, and to instead review existing laws to simplify and remove bad regulations.
politico.eu/article/11-count…
jjh retweeted
In 1956, Cybernetics was renamed "AI" by logic programming experts. Now Trump wants to rename AI again. Let's revive the old name! Modern AI is about neural nets & deep learning, closer to Cybernetics than to 1956 AI. Plus, Cybernetics sounds better than Artificial Intelligence.
What did Turing see?
Two days ago, GPT-6 Astra broke a yet unsolved German Army Enigma message from 1941. Amazingly Astra was able to autonomously:
- Search historical archives
- Compare uncertain letters
- Find contextual clues
- Build an Enigma simulator
- Write cryptanalysis code
- Run parallel experiments
- Test competing keys
- Recover the plaintext
- Cross-check the results
1/n
OpenAI: Our new model gets an almost perfect score on frontier levels maths. We are excited to see what you do with it.
Me: "How do I open a cybertruck door?"
Astra:
For similar reasons I think continual learning might be solved in AI very differently from how humans manage it.
So the way we do work now is instruct the agent and then relentlessly nitpick, make snide suggestions and insult its intelligence and genetic lineage until the task is completed.
The future clearly belongs to those of us who grew up with a lot of siblings.
Sadly I won’t be at ICML but see below if you will be
If I had known that being a founder means I can throw a party to play all of my favourite board games with fellow nerds, I would have got into this gig a lot earlier!
Food, open bar, and 60% of my board game collection including:
- Catan + expansion
- Heat
- Wingspan
- Pandemic
- Great Western Trail
- Splendor Marvel
- Set
- Carcassone & expansion
- Dungeon Mayhem
- Jaipur
- Agricola family
- Azul & variants
- Charty Party
- Pandemic
- Lost Ruins of Arnak
- 7 Wonders
- Ticket To Ride
- Chess
- Scrabble
Comment if your favourite's not on the list and I will see what I can do (both in terms of procurement and fitting it into my suitcase).
luma.com/386wuc69
When bureaucratic systems treat you inhumanely, if you have opportunity, it is a strong moral good to fight them without mercy.
First they took all the Noam’s, now they’re going after the Deans.
I can think of no better person to help shape frontier AI policy than @deanwball. He has a clear understanding of where AI is headed. I look forward to working with him at @OpenAI!
Reaching Type II on the Noamshev scale might seem minor so I've plotted it out to give a better intuition.
