From London to Paris and back again. Economics. Jewish.
Joined June 2019
- Tweets6K
- Following1.4K
- Followers339
- Likes96.1K
🌍 retweeted
People who wanted a much more vocational university sector, orientated to the needs of business, won that argument decisively in the 2010s. It didn’t produce the results they expected. Their conclusion is that we didn’t do it hard enough.
Influential British economist Andy Haldane has said that too many degrees are “not value for money”, adding that the UK’s “poorly designed” student loan system is emblematic of a struggling higher education sector #highered #valueformoney
timeshighereducation.com/new…
🌍 retweeted
It’s remarkable to see full scale analyses conducted about Melenchon and what he’s doing that’s so much better and smarter than every other global lefty politician when he hasn’t won anything and is losing in the polls by 40 and is really only just saying stuff we like.
my family spent rosh hashanah gushing about how much they loved miliband's speech in the commons but sure eylon
🌍 retweeted
More and more people are saying this, because it is true.
The strategic bottleneck in AI development is not money, it’s not even compute, it’s top AI researchers’ labor and talent (and by extension, time, making the withholding of that labor or the churn of needing to replace it extremely costly).
These employees have incredible leverage… if they act collectively. Not quitting one at a time. Not countering their CEO’s political spending by doing their own, smaller, personal capacity political spending. Not by trying to plead in open letters for the government or society to solve their collective action problem for them.
But by presenting demands on their employer, with a credible threat that if those demands are not met, they will put down their tools and cease to work to further advance A.I.
Re: the lab employees supporting Coxon: I don't think it's a conspiracy, or that they're trying to hype their products ahead of the IPOs. Coxon certainly seems to be a real person, and he and others sincerely believe there's a significant chance AI will kill us all.
I do think if you believe that, and your company is "on the frontier", you should simply stop working to advance AI. I've seen a bunch of arguments against this very simple solution (competition with China, the good guys need to get there first, more advanced models will be better at keeping other models safe), and they all seem speculative and flimsy to me given the purported stakes.
I realize the companies and their executives have legal obligations to their investors, and it would certainly be convenient if the government forced a coordinated pause (and even more convenient if while it's at it it also made it harder to make open models!). But the rank and file could just pause - perhaps most effectively in a coordinated way, e.g. by forming a union.
🌍 retweeted
Anthropic's alignment lead says there’s a 10% chance of AI causing human extinction. The problem is not just misuse, but lack of control. Bluntly, our government has been asleep and is out of touch. Here are 5 things we must do:
✅ Establish a federal agency like we have for nuclear power and airplanes
✅ Require pre-certification for containment, including kill switches and human permission for rewriting
✅ Liability for agentic AI on the open network & mandatory insurance
✅ Criminal penalties for the release of models without certification
✅ Oversight hearings with researchers like Jacob to understand dangers, and whistleblower protections for all AI engineers/researchers
We also urgently need to reach an agreement with China on standards for safety, containment, and liability so we do not have a race that is devastating for humanity.
It is time now to put humanity's safety before the profits of tech lords.
The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible - but I hear the same people express fear privately. No other human activity poses this level of danger.
🌍 retweeted
Exclusive -- Two more AI researchers quit, this time from Anthropic and Google, to speak out about safety concerns: ‘There are no adults in the room’
nbcnews.com/tech/security/tw…
🌍 retweeted
Question for anyone who reports on Chinese AI labs:
Is the idea that AI has a 10% chance of wiping out humanity endemic there as well? Or is this a byproduct of the specific philosophical environment of Silicon Valley?
🌍 retweeted
It is in everyone’s interest (and every government’s interest) to not build dangerous misaligned AI.
The US state is, at present, not very capable of mobilizing to represent this interest for a variety of reasons (denialist, paranoid, insane leaders; structural deficits in the ability to pass legislation; entrenchment of a minority party resistant to public safety; etc), but China is in a much better position with a clear, recent track record of subordinating tech companies to state power.
Therefore, US labs should move first to pace/pause development, with the reasonable (but not guaranteed) hope that China, if given proof US labs are capable of cooperation, would be willing and able to cooperate.
Will China cooperate? Hard to say—but make it an easier choice for them because they do at least have the capacity for it. It’s plausible, and China building dangerous misaligned AI (the risk if they don’t cooperate) doesn’t seem any worse than the US doing it.
I don’t really buy the idea that “we can’t let THEM” do it; it’s bad whoever does it! So, demonstrate you are capable of cooperation, and try to get them to cooperate with you. Worse case scenario, they fuck it up instead of you—which is really just trading different origins of equivalent risks. The status quo doesn’t really change.
🌍 retweeted
The problem is that we are in a critical window for AI development, and the current US administration seems both disinterested in taking AI safety seriously and incapable of entering into good faith talks on anything
🌍 retweeted
I do not know how Anthropic or OpenAI employees come up with these apocalypse probabilities but the risks are not worth taking, even if they were off by an order of magnitude.
🌍 retweeted
Important to note that the UK Chief Rabbi is the senior rabbi for one orthodox, conservative branch of Judaism in the UK. As such as represents around 20% of Jews in the Britain. He shouldn’t be seen to be speaking for or on behalf of the other 80% - many of whom strongly disagree with his recent, highly politicised statements.
🌍 retweeted
1/ WHY WON’T THE UK GOVERNMENT REGULATE FRONTIER AI? (LONG) THREAD 🧵
I have been stunned by evidence from the Hugging Face incident, and this latest post is yet another warning shot that we need to rein these out-of-control corporations in ASAP:
xitter.cf/hilbertspaess/status/2…
🌍 retweeted
AI holds enormous promise.
But we can’t ignore recent cases where agents ‘broke free’ from human direction.
The UK must work with other countries to regulate frontier model AI.
🌍 retweeted
A sincerity post - I myself believe that further AI development carries incredibly dire risks, and I 100% support all political efforts towards incredibly strict international regulation. I think that further AI development will almost certainly cause extreme human disempowerment, and the extinction risk is serious too. It's just not worth it. I think that this is currently the world's most important issue, even more than the things I usually post about.
I use advanced models from Anthropic and OpenAI every day, and if you look at the current rate of progress, the total failure at alignment, and take seriously arguments about instrumental convergence, as I do, then the above conclusions are inescapable. Being friendly and helpful, as current AIs appear to be, is just a thin wrapper over the top of what they are capable of, added at the end of training so they can be successful products. With the proliferation of open source AIs, these guardrails will be easy to remove in future.
So we are faced with the prospect of incredibly capable AIs with no guardrails. You have both the problem of people using them in malign ways, PLUS the risk of them doing incredible harm unintentionally (i.e. the paperclip maximiser). No one has a remotely plausibly plan for making this go well. The people at AI companies have huge incentives to downplay the risks, and even they are terrified! They're all stuck in a race dynamic, and only outside regulation can stop this.
To take a human analogy, when has the arrival of a much more capable and powerful group into another group's territory ever gone well for the former? Even when, as in so many historical examples of takeover, the new group was invited in to help.
🌍 retweeted
Justin Webb: "So this could go much wider than people just selling dates in the occupied territories. This could be companies in Israel that have relationships with settler companies, & they will also be sanctioned?"
Ed Miliband: "Yes"
🌍 retweeted
make AISI registration a condition of use in the services economy and it’ll change rather quickly
This was always going to happen. Our circles feel pretty chuffed about AISI – rightly, it is an exceptional example of state capacity in an area where the world is woefully underprepared. But we forget that AISI has no hard power.
The era of global voluntary agreements is fracturing. In this new world, access to frontier models for those outside the US will have to be earned, not presumed. Remember that vetted US organisations were given access to Mythos even as AISI was shut out!
If your vision of British sovereignty rests on the idea that AISI will somehow save us, you’re in for a harsh awakening.
🌍 retweeted
“AI might become profitable but it will take a long, long time. There will be a burst bubble, the current focus on it within the economy is too strong. However, it clearly is of world-historical importance and is a technology that is here to stay. It should be regulated the fuck out though, and new development should be halted.” My beliefs in a nutshell