Yesterday, the
@FT
ran a piece about government assurances around compute. We’re confused by the government creating an AI Opportunities taskforce days before canning investment, and now backtracking.
We’ve spoken to multiple insiders to figure out what’s going on. It's bad. 🧵
1) there is no early stage funding gap in the UK
2) this is a bit rich from the TTO that has a track record of offering UK spinouts some of the worst terms out there
OUI has probably incinerated more potential value than any other technology transfer office in the UK
This from the FT is a peak example of the European war on dynamism. There’s no social or economic problem that isn’t solved by more tax and more regulation. 🧵
@Samfr
as someone with a depressingly detailed knowledge of MoD contracting, this will involve, among other things, ministers signing off paint supplies, uniform badges and buttons, conference booking services, and plumbing works on individual bases. A good use of time for all involved.
government seems to have a pathological need to make British science/tech ambitions look ridiculous with this kind of over-claiming. it just reeks of unseriousness and is *completely unnecessary*
Criticizing a new government at the height of its powers might seem mad, but when institutions make decisions that are bad for the ecosystem, we’ll always call them out. And if you have any experiences or tips you want to share in confidence, our DMs are always open.
Rebranding it as an AI-first or AI-only project is a recipe for a procurement disaster. NVIDIA’s sales team are rubbing their hands together with glee, knowing they’ll be able to name their price.
My goodness me this is *bang on* from my brilliant colleague
@Patrick_S_King
Politicians are elected to make decisions, not duck them by setting up reviews & quangos
Decisions, accountable to the people - that’s the job.
Our sense is that the Treasury wielded its axe in haste, the government is repenting at leisure, and is desperately trying to spin its way out of this. In the process, we’ve seen them make a few claims that are either flatly untrue or misleading.
This is not how to make policy.
It’s not a coincidence that this occurred within days of the government firing a large number of its independent advisors. The handful of successes around science and tech under the last government came from outsiders’ work. We’re witnessing the return to business as usual.
the most comprehensive take on the UK's natsec/defence challenges so far - realistic proposals to correct our lack of seriousness across strategy, technology, procurement. next govt will hopefully implement in full
(and pleased to have made a very small contribution)
The world's in a dangerous state. The UK can show leadership, but to secure peace we need a new plan.
New technological capabilities are critical and we need to prepare for the possibility of war as we strive to ensure it doesn’t happen.
Latest from
@InstituteGC
you know the
@stateofaireport
has become canonical when the
@turinginst
feel able to lift your graphics without citing them. They could've at least varied the examples and the labels...
This is unforgivable, given how poor the UK’s compute capacity already is by international standards. I can only wonder how Edinburgh is feeling right now.
@_charliedickens
they're received hundreds of millions of pounds in government funding + secured an even larger funding deal for the years ahead while making a limited contribution to the country's AI ecosystem and utterly failing to prepare government for recent capability advances
great to speak to
@katyanna_q
about the need for UK universities to stop treating founders as a source of rent and take a longer-term view when it comes to spinouts
Even referring to it as the “Edinburgh project” is slightly tricksy - we should be clear that the government is defunding a national supercomputing facility. It would be equivalent of abolishing DSIT and calling it ‘a reallocation of spending power between offices in the
We’ve confirmed what
@griffitha
says on the spend being ‘unfunded’. It’s misleading and ministers shouldn’t resort to these kinds of tactics in the hope no one will understand how govt spending works.
The previous compute review was conducted over the best part of the year by a senior Google research director.
I have immense respect for Matt, but it’s not realistic to believe that this can be reworked entirely (along with a set of other recommendations) in the space of a few
Far from the pacifist and anti-imperialist aims of Corbyn, the Labour Party is now the party of war and militarism.
Labour is beating Nato's drums of war, writes
@battleforeurope
:👇
This is a misrepresentation of the exascale project. The design of Europe’s Jupiter project shows that exascale supports all types of research. We’ve heard from multiple people close to the previous compute strategy that this was the case in the UK - govt is hiding behind a lack
I'm not 100% sure how anyone can look at the UK's hopelessly unambitious defence spending plans and conclude they're a declaration of war against anything other than our own national sense of self-respect
in the UK, a start-up would win three 50k grants to do this, before the MOD goes to a prime. despite it requiring 3 bits of metal, it would somehow run 40% over budget and be delivered years late. a select committee in 2035 will describe it as a "baffling use of taxpayers' money"
It ain't pretty, but it works. Russian mil-industrial complex is moving fast and iterating.
Do you think they had to wait 3-5 years for a procurement to sell glide kits for dumb bombs like one would in EU/US?
@andrewjb_
Just minor point - lifetime costs were included in the budget. Upfront costs would be lower than headline £. Budget also included skills/software progs etc to exploit.
good stuff. real wake-up call for anyone who think govts. are taking this remotely seriously - the picture is worse than you think and decel Treasury Brain isn't going to fix it
Access to compute is becoming the next major global digital divide.
The latest
@institutegc
report, on ‘The State of Compute Access’, has the data to show it. Policy and international institutions must adapt, fast. 🧵
when
@nathanbenaich
and I wrote about the perils of large-scale horizontal AI regulation - even we didn’t quite appreciate the EU’s ability to so thoroughly clown itself
Boom 💥.
French lead EU Commissioner for the AI Act,
@ThierryBreton
strikes back with an interview 'au risque de choquer' 😲:
'Big Tech and AI startup Mistral do NOT represent the public interest'.
The new government has an opportunity to reset science and tech policy - ending the failed blend of short-termism and subsidy.
@nathanbenaich
and I outline how they can do it.
New on
@airstreetpress
: The UK Secretary of State for Science, Innovation and Technology commissioned
@matthewclifford
to produce a roadmap on harnessing the benefits of AI for growth and productivity.
As believers in openness, we’re sharing our submission in full.
There’s of course a case for assessing whether we need to recalibrate big projects, but the right time to do that was the upcoming spending review. Instead, they've thrown something that’s been in the works for two years into uncertainty.
tldr; stop handing out grants that won't go anywhere or pretending that the EU can recreate big tech companies. do start funding defence properly and building out public cloud capacity. thank you for coming to our ted talk
New essay: The growing interest in AI from governments is a welcome development, but we believe that excitement should be tempered with discipline.
With
@chalmermagne
, we look at how government technology investment often fails to accomplish its goals.
This usually stems from
My inaugural contribution to
@airstreet
blog. Inspired by
@duncanrobinson
’s Reform Fairy and the recent Future of Britain conference, this dives into how the UK is keen to "embrace the opportunities of AI", but incurious about what this means practically. 🧵
@shashj
looking at my old contract spreadsheet, it would also involve ministerial sign-off on horse feed + bedding, paint supplies, and door and window replacement in the Lichfield area
last time we wrote an essay on China, you’d have thought we single-handedly started the new Cold War based on some of the reactions. So naturally, we’ve done it again.
New on
@airstreetpress
- the state of Chinese AI.
We dive into the SOTA from Chinese AI labs, the chip wars, sanctions, and loopholes.
Some highlights below 🧵
i) correlation =/= causation - there are some important differences between Imperial and Southampton that aren't connected to UPFs!
ii) can we not use Oxford's diabolical setup as an example of good practice?
@sebkrier
have you considered leveraging the transformative capabilities of generative AI in a responsible and sustainable way? my invoice is in the post
@v_j_freeman
disturbing to see how a vocal minority’s critical thinking abilities can absolutely melt - saw this a lot in my labour students days. Sending my best.
If we’re to bring dynamism to European defense, investors need to see a path to exit. With lower margins than enterprise saas, few making it to IPO, and a limited pool of acquirers.
What’s the answer?
We dive into the surprising lessons defense can learn from pharma. 🧵
New from me and
@nathanbenaich
for the
@AirStreet
website - a defence of the UK’s “pro-innovation” approach to AI regulation, ahead of the upcoming AI safety summit. 🧵
the only AI report in the universe has landed!
we have LLMs, things that aren't LLMs, chips, drones, regulation, safety, geopolitics, elections, VC, x-risk shouting matches. joy to work on my first
@stateofaireport
with
@nathanbenaich
,
@corina_gurau
, and
@osebbouh
🪩The
@stateofaireport
2023 is now here.
Our 6th installment is one of the most exciting years I can remember. The
#stateofai
report covers everything you *need* to know, covering research, industry, safety and politics.
There’s lots in there, so here’s my director’s cut 🧵
I've managed to draft a whole seven sentences on the UK election for this week's Guide to AI, which is probably twice as much as the main party's manifesto commitments on AI combined
@adb0wen
and they are all significantly better at it than their UK equivalents. guess it's a product of a political culture where politicians debating their opponents on the reg directly is a bit more of a norm, rather than the campaigning from safe house to safe house we get here
Brexit has had a profound impact on young people who didn’t get a say in determining our country’s future.
If I'm re-elected as Mayor, we'll bring in a new London version of Erasmus, helping students in our city to travel and study abroad once again.
this dross from the new statesman is typical of the scare-mongering around actually useful innovation. they can't point to any specific in the contract that's bad, clearly don't understand the controller/processor relationship, so have opted just to make spooky nosies
uni lobbying becomes less coherent and more blatant by the day, as different policy solutions and rationales get mixed up: "we need government subsidy for companies at their earliest stage because there's a growth stage funding gap"
Dross, even by the UN's standards:
- no proposals beyond a committee of committees
- tortured efforts to pretend the SDGs mean something
- random references to the Global South that don't go anywhere
- responsibility principles a public affairs intern could write using Claude
The Secretary-General's Advisory Body on AI published its interim report
The report calls for anchoring AI in international law, human rights & SDGs.
It also identifies AI governance critical functions & principles
You can read it here
@arisroussinos
not just true of foreign policy - we're enjoying a total holiday holiday from reality. Not helped by a Westminster/media bubble whose sole interest is debating which cut-out landed their meaningless talking points slightly better in front of an ever diminishing audience
@nimivashi15
because that would involve them being less greedy with equity and not treating start-ups as a form of rental income. much easier to cry about 'funding gaps' and ask for subsidy
data, oil, chatgpt for x: imprecise analogies for technology lead to bad policy, bad business decisions, and frankly, just irritate me on a personal level. I can't promise the last of these isn't the reason we wrote this.
New from
@airstreetpress
: a definitive answer to the question of whether data is the new oil (spoiler alert in image), before we dive into the dangers of easy analogies from policy through to entrepreneurship.
A few highlights below 🧵
returning to my favourite subject, six months on: the horrors of European defence procurement. our take on what's changed, what's stayed the same and where next
Last year on
@airstreetpress
,
@chalmermagne
and I dived into Europe’s lack of defense readiness and the failure of many governments to engage productively with AI-first challengers.
Six months on - what’s changed? 🧵
it's legit insane how much of start-ups' time government wastes consulting (and then ignoring them) on policy in exchange for intangible 'influence', because it lacks political courage and is too stingy to build out capacity itself. not superpower vibes tbh.
I know it’s not breaking news, but the bizarre political journey of some self-proclaimed centrists worried about supposed government overreach in the Brexit years is one of my favourite subplots
Pass a short Bill closing Twitter down in the U.K.
@Keir_Starmer
? There is more than enough reason to do so. One of the richest men in the world is using his platform to cause serious harm - putting lives & communities at risk.
@YvetteCooperMP
@lisanandy
Attention, United Kingdom!
"We don't need to touch our strategic weapons. We've got enough of others to make sure that there is nothing left of Britain, our sworn enemy. That's where everything stems from, all the rest are underlings." - Russian propagandist Gurulev.
Maybe we'll have to rethink some of these questions in future, but these attempts to disincentivise innovation before we've even see its effects are exactly why
@nathanbenaich
and I keep banging on about European Dynamism
This is a peak e.g. of the phenomenon in our recent essay on UK AI policy - instead of resolving the core barriers to wider AI adoption (e.g. rubbish data and procurement), announcing tiny pots of money that won't move the needle
Second, we're putting AI to further use to improve the NHS.
To support services like robotic surgery, we're giving an extra £13 million to develop cutting-edge AI technology.
I guess I shouldn't be surprised at this stage, but still taken aback by the extent to which a section of the UK political class is prepared to continually degrade itself for a quick payday
April 1: Israel bombed the Iranian consulate, an act of war.
April 13: Iran attacked Israel, an act of war.
Biden says: you're even-steven -- and he's right.
Israel says: no way.
This is going to escalate. Of course it will.