Marc Warner

[Marc Warner @MarcWarner10]

At faculty, we think that we should build human-first ai systems, both for principled reasons (it’s safer and more ethical to ensure humans control the systems) and pragmatic reasons (we don’t have agi, so the best decisions will be made by people and machines working together)

[Marc Warner @MarcWarner10]

@sebkrier I’d like to read that piece!

[Marc Warner @MarcWarner10]

The flip to safety continues.

[The Free Press @TheFP]

AI safety has become too important to be left up to Washington’s whims. A partnership among the labs could reduce risk without stifling progress. https://t.co/9m72TZylJq

[Marc Warner @MarcWarner10]

I think this is a big deal!

A16Z were central to the resistance to ai safety.

If they are (correctly and admirably publicly) walking back from that, then I think it will be hard for others to sustain.

That’s a good thing, because it means we can stop having the debate “is safety important?” and start talking about the more important and interesting questions:

“How do we know what the frontier capabilities actually are?”

“What safety might actually work?”

“How do we trade off safety and concentration of power?”

“How can we govern these models for broad benefit?”

[a16z @a16z]

Martin Casado spent years dismissing the AI doom scenario and says what he got wrong was how long the scaling laws would keep holding:

"I was responding to this Bostrom notion of recursive self-improvement, fast takeoff. You create one of these things, you step back, and it takes over the world... That's clearly not what's happening."

"But here's what I got wrong. What I got wrong is I did not know that we could effectively just continue to pour money in this. The scaling laws are holding."

"I don't know what it means to do a $100 billion training run, to have this thing that you're putting $100 billion in, and then that money comes from this meta-economic machinery that may want to solve whatever."

"They may want to solve cancer, but they may also want to create a weapon. Like, who knows?"

"This concentration of this many resources in a useful way, I think, is very new. I don't think we understand the implications. I think you could reasonably argue that that's very dangerous if you apply that $100 billion in the wrong way... What does it mean to be able to concentrate resources?"

@martin_casado

[Kanishka Narayan MP @KanishkaNarayan]

Britain is becoming the first country outside Ukraine to access the Avengers AI Labs - a landmark step for our shared security.

Today’s visit by @10DowningStreet further reaffirms that Britain and Ukraine are the closest of partners.

It’s been a privilege working with @wesstreeting and @DefenceU to merge our technological capacity with Ukraine’s frontline ingenuity.

https://t.co/UBoM7TfoVr

[Tom Westgarth @Tom_Westgarth15]

This is a pretty big win.

Through its drone R&D, Ukraine has built some of the world’s most advanced data-labelling and AI testing capabilities. The Avengers platform contains an enormous volume of labelled real-world data. We are talking multiple times the quantity of data labelled that trained GPT4

The long tail of this has exciting public service implications too. Hospitals, grids, etc could allow AI systems to be tested against thousands of real scenarios, rather than just the demo data.

More govts should be collaborating on major data sharing projects like this!

[Jakob Foerster @j_foerst]

TL;DR: This is one of the most important and exciting opportunities in AI on the planet - please read on.

The British Open-ended Learning & Discovery Lab is creating the perfect place for paradigm breaking AI research in the name of open-source and open-science. We have agency, we funding, we have unprecedented amounts of compute*, but WE NEED YOU!

..and we have created the dream job for you: The BOLD Fellow. This job combines a fast-moving, high agency, collaborative environment with full academic freedom and a salary that pays the bills.

Apply by noon UK time on the 15th of September for this once in a lifetime opportunity to shape the history of our field and of our planet:
https://t.co/wN009tQEkt

*by academic standards

[Michael Gove @michaelgove]

There is no one better equipped for this vital role - @HZoete is an amazing talent

[Kanishka Narayan MP @KanishkaNarayan]

Great to see @AISecurityInst appoint Henry and Nate to these important roles.

Both bring significant experience and will strengthen the Institute’s leadership.

My thanks, too, to Adam for his leadership of AISI and everything he has done to build the organisation.

[Ian Hogarth @soundboy]

Henry is many things - a startup founder, an experienced operator inside goverment and someone who has done great work at the intersection of AI and politics.

I'm thrilled to see him take on this role at @AISecurityInst

[Henry de Zoete @HZoete]

The UK’s AI Security Institute is an organisation close to my heart. Back in 2023 I was involved in its conception and creation. It’s gone on to do incredible things. A genuinely world leading org that shatters the myth that government can’t do things.

So I couldn’t be more delighted to say that I’ve been appointed as AISI’s next Director, working alongside the brilliant @NateBurnikell as its new Chief Strategy Officer.

AISI is the world’s most respected organisation for understanding the risks from powerful AI. That is thanks to their superb team of dedicated civil servants and technical researchers.

Special thanks to Adam Beaumont who has done a great job as AISI’s interim Director for the last 9 months. He returns to GCHQ having seen AISI go from strength to strength under his watch. I’m hugely grateful for that and excited to work with and learn from him.

Last year I wrote about AISI and why its form and function is so important. You can read that piece on the link below - it gives a good flavour of why I’m excited to be joining.

I’m looking forward to getting started in September.

https://t.co/dAhChKjwUw

[Marc Warner @MarcWarner10]

Finally psychohistory!!

[James Zou @james_y_zou]

New work I'm very excited about: Physics of Agents!🤖

As AI agents become more prevalent, they'll interact and influence one another.

Can we predict the collective behavior that emerges?

Surprisingly, their dynamics follow compact, predictive laws of statistical physics🧵

[Marc Warner @MarcWarner10]

Strange that Hanson thought we’d get Em’s via brain modelling, but we’ve actually got Em’s via language modelling.

[Richard Ngo @RichardMCNgo]

@nabeelqu First part of my 13,000 word retrospective on this phenomenon is out today: https://t.co/jhD4KWd3vS

[Mike Butcher (BlueSky/Threads: @mikebutcher) @mikebutcher]

What some mainstream, non-technical journalists don’t seem to have quite grasped yet is that the era of AI naturally leads to “Sovereign AI” which therefore becomes an issue of national security, fundamental infrastructure and a source of economic growth. (And I’m not just talking about LLMs). It’s doubtful if Burnham’s “Cabinet Minister for AI” appointment will have the heft to address these huge issues on their own without a full department, and DSIT appears to be for the chop. This is the major concern of most in the tech industry, and it’s not a self-serving position - it’s an issue of national importance.

[Marc Warner @MarcWarner10]

Strongly agree with this!

The Uk needs growth. Our institutions of science need fixing. But if we attempt to rebuild the institutions of science directly for growth, we’ll get neither great science nor growth!!

[Brian Cox @ProfBrianCox]

My view is that there is a fundamental confusion within government and UKRI about the relationship between scientific research, Universities and growth. A strong science base and strong Universities are necessary for growth, but they are not levers you can pull to generate it, at least over the typical lifetime of a parliament. Science is the curiosity led exploration of Nature. It delivered the modern world, but it was not invented to do that. If you try to conflate all these things together you’ll get neither economic growth nor transformative scientific progress, and you run a serious risk of destroying the Universities and the science base upon which future growth depends.

[Garrison Lovely @GarrisonLovely]

As a journalist covering AI—and certainly not a part of the industry— the UK AI Security Institute (housed in a dept that might get scrapped by Burnham) has been an invaluable resource for understanding AI risks and capabilities. Without it, we’ll be more dependent on conflicted assessments from industry.

[Nathan Benaich @nathanbenaich]

the new uk gov shouldn’t dismantle its single dept that understands technology

[Tom Westgarth @Tom_Westgarth15]

Speaking from the perspective of AI:

1.) Some parts of DBT actually are on course to be really helpful on AI (e.g. Global Talent Taskforce, UK Export Finance). But these are areas where incentives are already broadly aligned and 'join-up' might actually do more damage than good.

2.) You don't get critical institutional innovation like AISI emerging through a 'beefed up' (ie unfocused) department - AISI only could have emerged through DSIT. Many officials across civil service recognise that DSIT is simply a more dynamic department than much of govt. The moment we are in demands urgency, creativity, and effectiveness. I don't see how a 'super department' delivers this.

3.) This is a crude heuristic, but broadly on AI, DSIT is performing well in areas where it owns the levers, such as compute procurement, funding of novel research, SovAI, AISI, etc.

4.) Where DSIT doesn't own those levers (e.g. things like planning/grid connection) there is a lot of room for improvement. But that gets fixed from an effective No10 that can exercise authority over differing departments, not a 'super department' that would end up running into similar coordination problems.

While the US and China race ahead to automate AI R&D and bolster their own sovereign capabilities, the last thing our government AI teams should be focused on is designing new org charts and creating new departmental strategies.

[Barney Hussey-Yeo @Barney_H_Y]

Sad to hear that Burnham's team is considering closing DSIT.

The UK has a major competitive advantage in its scientific capacity. Turning that strength into economic power — DSIT's job — would be my idée fixe as Prime Minister.

DeepMind is our biggest success story as a nation since the GFC. Imagine if we could create more companies like that and actually scale and list them in the UK.

A $1 trillion company would dramatically turn around our GDP growth rate and allow us to fund everything the state wants to do today...

[Ian Hogarth @soundboy]

1/ Some thoughts on the rumour of yet another government department reorg...