Connor Axiotes (Paris & San Fran in Feb)
@connoraxiotes
Followers
3K
Following
30K
Media
712
Statuses
20K
Advanced AI poses risks we aren’t ready for | @apartresearch | Making a Documentary on AI | Boxing, MMA, @LFC, @NBA fan | Views my own
London, England
Joined November 2012
@Carra23 Omg I’m falling in love with my Dutch manager - I said I wouldn’t fall in love again after Jurgen.
5
2
438
Just became @UKLabour’s Junior Minister for AI Alignment - super excited to get started! 🚀 🇬🇧
7
1
227
@Con_Tomlinson @calvinrobinson Spiteful takes here from a spiteful duo - sorry to the women in politics they’re referring to.
0
2
128
From @jackclarkSF . OpenAI's new O3 model shows that there are huge returns to scaling up a new approach (getting LLMs to 'think out loud' at inference time, otherwise known as test-time compute) on top of already existing powerful base models. I expect the next logical thing to.
6
15
142
@CharlotteCGill I will not sacrifice my pro-immigration views for my YIMBYism, though. My moral responsibility matters more than my hope for more homes. But this is a false dichotomy - we do not have to cut immigration to fix the housing crisis. We just need to build more houses. A win-win.
262
12
114
Very happy to say I have been promoted/moved on to a full-time position in the @HouseofCommons as a Member of Parliament’s Parliamentary Researcher! Excited to get started 🏛.
10
3
118
@trussliz Should we have not involved the Soviets when it came to Nuclear weapon non-proliferation?.
2
2
105
@OliviaUtley Do we forget what the lords is? And has been for centuries? A bunch of unelected and usually low quality political appointees? . Let’s reform it then if we don’t like it - but to say that Charlotte has no right as a result of her age is silly. Do you think we shouldn’t let 30.
46
1
88
Prof. Russell on artificial general intelligence (AGI) timelines:. “Everyone has gone from 30-50 years, to 3-5 years.”. #AISummitTalks.
6
23
83
Genocide, rape and the harvesting of organs. That is what the Chinese Communist Party is doing in Xinjiang to the Uighur people. We cannot allow it. (Fantastic to see @MaajidNawaz really kick up a fuss about these atrocities.).
2
31
83
Spiteful takes by a spiteful duo here - I’m sorry women in politics have to put up with this.
Seeking attention on social media is a form of digital infidelity. It makes your man feel unwanted, and profanes the sanctity of your relationship. Your husband should be your sole source of validation. Inspire him to make sacrifices. From yesterday's podcast with @calvinrobinson
2
3
73
🚨I'm excited to announce that I'll be joining the premier AI Safety and Alignment Lab @ConjectureAI team next month - who are on the frontline of making the world a safer place. Thanks to @NPCollapse and @_andreamiotti for letting me part of the project.
11
2
74
Tomorrow we have @ConjectureAI's packed out event in the @UKParliament's @UKHouseofLords. Our CEO (@NPCollapse) will be making the case to legislators, journalists, and other interested parties as to why the regulatory focus should be on artificial *general* intelligence.
4
11
69
Thanks @RishiSunak - was a pleasure being on the road with you one last time. No world leader has done more to put AI safety at the forefront of government than you with the @AISafetyInst. Now I hope @UKLabour and @peterkyle keep this legacy intact and extend it.
1
5
61
@ESYudkowsky @elonmusk @liron @BanklessHQ This is why you need to keep doing as much media as possible @ESYudkowsky - become the most popular communicator of AI alignment worries. Keep fighting!.
2
1
54
🌎My pro-immigration case for @DailyExpress🚨. "It is our moral responsibility to welcome immigrants into our country. To give them a chance to never starve, be persecuted by their state, or lose a kid to ill-health again.". (I didn't write 'lazy brits').
200
12
52
🚨AI SUMMIT TALKS🚨. 📅31/10/23.⏲️2pm.📌(Wilton Hall) Bletchley Park. - Professor Stuart Russell.- @tegmark .- Jaan Tallinn.- @NPCollapse .- @_andreamiotti .- Annika Brack.- @halhod .- @Ron_Roozendaal .- @MarkBrakel .- @AlexandraMousav .- @dw2 . RSVP now:
4
21
49
For the @WSJ: “We’re seeing a ridiculous death race to godlike AIs from all the major players,” said @NPCollapse, CEO of @ConjectureAI, an AI lab working on solutions to the alignment problem.
2
9
46
I will not sacrifice my pro-immigration views for my YIMBYism, though. My moral responsibility to migrants matters more than my hope for more homes. But this is a false dichotomy - we do not have to cut immigration to fix housing. We just need to build more houses. A win-win.
Immigration is too high. If more young adults could engage with this fact, we would be able to apply more political pressure. There's no way the UK can build at the rate people are moving here. Rent will go up until we accept this.
61
2
43
@robertskmiles Sub-optimal times means we sometimes have to make sub-optimal choices. But you’re one of our best communicators and I think it’s probably net positive to have you on the air waves, Rob. You’re killing it and we appreciate you.
2
1
40
@ESYudkowsky @ylecun @erikbryn What’s the concept for this kind of logical misunderstanding again? The fallacy that just because something isn’t here *now* means it won’t be here *soon* or at a slightly later date? The immediacy fallacy?.
14
1
39
@CosmicSkeptic @jordanbpeterson - @CosmicSkeptic is a brilliant young YouTuber, extremely intelligent, great interviewer and broadly read. You should without doubt get on his podcast! He’s been on @triggerpod which I’d seen you have been on recent.
0
0
37
Libertarian take - it’s so damn annoying that this can’t happen with just any country we want to go to. We have to wait for our governments to give us permission? Weird.
Wonder how many people are planning to take this up "From 2024, any Brit up to the age of 36 will be able to live and work – at any skill level – in NZ for up to three years"
3
5
35
@melissasigodo For those of you making fun of him - he makes a very salient point: recidivism rates are very high and when prisoners leave prison they are set up to fail. He’s being very honest with himself and telling the truth.
0
0
32
A fantastic but sobering talk from @ConjectureAI’s @NPCollapse at @CogX_Festival. Tl;dr - the trajectory of this risky death race to artificial general intelligence is the extinction of humanity.
1
3
30
@tomhfh But Tom, why risk those 4 lives in a million when we have alternative vaccines that may not cause ANY of these 4 deaths?.
20
0
26
🤖 Me for the @Telegraph today:. ‘Connor Axiotes, of the @ASI, says the ability of advanced AI to make its own decisions *could* mean one day that humans will be displaced from the jobs market altogether.’.
5
10
29
UK AI Minister in @SciTechgovuk says: ‘Britain should have a physical center “looking at [AI] frontier risks, constantly scanning the horizon and understanding how close or how far that’s getting.”’. Maybe CERN and the IAEA can inspire us here when creating our monitoring system.
Meet Rishi Sunak’s minister for artificial intelligence — the hereditary peer sci-fi enthusiast who only joined the House of Lords last year. I sat down with Jonathan Berry — aka the 5th Viscount Camrose.
3
6
30
@ConjectureAI CEO Connor Leahy (@NPCollapse) on @CNN talking AI regulations and the potential for training run moratoriums.
0
6
27
Proud my MP and fellow Brexit campaigner has the top job! @DavidDavisMP deserves it thoroughly #BrexitNegotiator.
0
1
27
📺@ConjectureAI CEO @NPCollapse:. AGI companies are "rushing forward into newer, larger, more dangerous regimes before we can even control or understand the systems that already exist.”
1
6
27
Very very happy to say I have achieved a Distinction in my overall MSc, and in my Dissertation at @durham_uni within @Durham_SGIA. My thesis utilised a simulated coded-game to ascertain the effects of Quadratic Voting on BES cross-sectional data.
3
0
29
I mean I’m glad the Government is doing something about it but very annoying it took a viral video for them to take notice… this should’ve been done a long time ago. Well done to @pursuitofprog for applying consistent policy pressure throughout.
This is appalling. The American XL Bully is a clear and lethal danger to our communities, particularly to children. We can’t go on like this. I have commissioned urgent advice on banning them.
1
2
26
Seems like @AnthropicAI’s commitment to do serious AI safety work on their state-of-the-art models is a promise they’ve kept.
New Anthropic research paper: Scaling Monosemanticity. The first ever detailed look inside a leading large language model. Read the blog post here:
2
0
29
🏛️ Alongside speakers from @CSERCambridge, @ConjectureAI, @GoogleDeepMind and others - I’ll be speaking about… . “What would a ‘CERN for AI’ look like?”. We’ve just added another few spots as we were a *sellout*… sign up here:.
🚨 NEW EVENT 🚨. 🤖 AI Governance Talks: *Institutions*. August 2nd @ 7pm (Shoreditch) - come and learn about what kind of institutions might be best for dealing with risk from frontier AI models!.
3
2
27
I pray for the day @s8mb doesn't have to write about building more houses anymore - because the UK just built them! So what stopped a 'net gain of 1252 homes, including an extra 116 subsidised properties' this time? Find out👇.
2
5
27
Our new Prime Minister in-waiting has the character necessary to help remedy our situation. As people struggle to pay their bills we need to do what @RishiSunak has always done - look after the most vulnerable in society.
0
2
26
@alexkaplan0 I hope @TomChivers and @StuartJRitchie have an episode on this for their new podcast! (here: .
2
1
25
Want to join 250+ guests alongside 8 incredible speakers? Including from @ConjectureAI, @FLIxrisk, and former Cabinet Minister @RobertBuckland!. Well you still can - sign up here:
AI Summit Talks - Together with @ConjectureAI, we're proud to announce new speakers: investor Jaan Tallinn, co-founder of @cser, policy analyst Eva Behrens from ICFG, journalist @tomough, along with a special message by @RobertBuckland. More info:
1
2
25
🤖 @ASI-@BloombergUK AI event ✅. @nadhimzahawi on existential risk & how Britain can become an innovative AI superpower *safely* ☄️ . @OmarSAlolama on how he created a safe, responsible approach to AI governance 🇦🇪 . @BrandonLewis on the @ASI at the forefront of AI policy 🧠
0
4
25
I have parted ways with @ConjectureAI after a great run. I am now looking for opportunities in AI policy, governance, or communications. DM me if you want a coffee or a call, even if just to catch up!.
0
2
22
@pmddomingos What AI advancement would have to happen for you to be shocked? What year do you estimate this will happen, if ever?.
2
0
23
A crazy six months!. I had to reject my LSE offer as I had got myself a job in D.C. with @mercatus - however, immigration lawyers told me that Trump's Visa Ban would mean I wouldn't be able to take it. So now I am now starting an MSc at @durham_uni in a few weeks!.
3
0
23
Was slightly strange to graduate from my MSc at @durham_uni before my BA at @LancasterUni… but fun nonetheless. Lovely also to spend a full day with nothing significant happening politically. Hopefully things liven up again next week… @Durham_SGIA #graduation2022
1
0
22
@david_okla @CharlotteCGill Yeah full uni low-income background grants, from the middle of Hull and from a state school with a heavily disabled mother - made me so bitter that now I want the UK to have housing and immigrants to be able to improve their lives x.
19
0
20
@CosmicSkeptic Your net-good in the world far outweighs this. You’ve inspired me and many others to do more for non-human beings. Your inspiration and philosophy goes beyond your diet. Just make sure you fight against factory farming that little bit harder Alex….
2
0
17
Some slides from my ‘Decreasing risk though policy’ talk at @durham_uni! . - Is Growth a good?.- What’s the difference between an AI and an AGI?.- How prepared are we for the next pandemic?.- Which policies might be help to reduce X-risk?
2
4
22
When we were at Darlington his reception was huge! @RishiSunak.
Sorry but the idea Rishi isn't, and wouldn't, be popular in the Red Wall is such a chronically online take. Please just visit Teesside.
0
4
20
@CJCHowarth @CharlotteCGill Show me your simple maths and I might be more inclined to agree!.
18
0
19
Apple Engineer 1: "Phil?". Apple Engineer 2: "Yeah?". Apple Engineer 1: "I reckon we should reinforce the institutional patriarchy by making sure women can't hold our phones.". Apple Engineer 2: "Yeah mate good banter, go for it.".
0
1
20
Statecraft is consistently one of the best blogs I read - @rSanti97 is great.
Today's Statecraft is about the ways we spent and misspent development money in Afghanistan. During my interview prep, I read a bunch of the GAO and SIGAR reports on waste and fraud. Here are the highlights. 🧵
1
4
18
🦠@juan_cambeiro for the brilliant @IFP on the Mink-related H5N1:. '[T]hat H5N1 causes a pandemic as bad as or worse than COVID beginning in the next year is ~4%, [. ] the expected cost in terms of potential harms to the U.S. is at least $640 billion.'.
2
6
19
This is so so dumb. So no new invention can be dangerous people, Yann has decreed.
- Engineer: I invented this new thing. I call it a ballpen 🖊️.- TwitterSphere: OMG, people could write horrible things with it, like misinformation, propaganda, hate speech. Ban it now!.- Writing Doomers: imagine if everyone can get a ballpen. This could destroy society. There.
1
1
18
At @ConjectureAI we are working on 'Cognitive Emulation': building predictably boundable systems. We are not working on developing directly aligned AGIs. We believe the former to be a far simpler, more controllable, and useful step towards a full AI alignment solution.
2
3
18
“Shutting China out from AI Safety talks would be like disinviting the Soviets from the nuclear arms talks.” . Great piece from @AlexTPet.
I touched on lots of the same themes and problems outlined in Rishi's speech - check it out below if you're interested. This is a stronger start for AI Safety than anyone could have hoped for. Now we have to make it lasting 🫡.
3
2
17