@ObadiaAlex

scaling trust @aria_research | https://nitter.cf/t.co/r2MLCQxlXK

alexobadia.com
Joined March 2016
  • Tweets5.6K
  • Following8.2K
  • Followers10.1K
574Media
Update from us on the Scaling Trust Arena 🏟️ Excited to be working with @andonlabs, @BT6_Official and amododesign.com*! Check out the post, the draft spec, register your interest, and (please!) give us feedback 🤠 scalingtrust.org.uk/blog/upd…
1
14
6
64
5,165
We're the only red team for @ARIA_research's Safeguarded AI programme, led by @ammannnora. Each development cycle, we attack the code, its proofs, and everything those proofs rest on: the models, specs, assumptions, and deployment choices behind every team's assurance case.
ARIA
@ARIA_research
Sep 24
AI-enabled cyber attacks are getting faster + going further. But what if AI-enabled formal methods could make high-assurance cyber defence practical at unprecedented speed + scale? We’re funding eight teams with £22m to test that idea. Meet them here: link.aria.org.uk/SAI-TA2-Cre…
3
8
2
47
6,294
Proud of our sister programme Safeguarded AI!!
Nora Ammann@AmmannNora
Sep 24
Proud and pleased to be finally able to announce our latest set of Creators working on technically exciting and societally critical projects. Formal methods have the potential to lead to a phase change in cyber resilience, and it matters that we get there fast!!
11
1,100
currently watching -- nobelprize.org/prizes/econom…

Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel 2009

The Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel 2009 was divided equally between Elinor Ostrom "for her analysis of economic governance, especially the commons" and Oliver...

nobelprize.org
1
6
875
we're running the first in a series of meetups on AI security & safety in London, co-organized with Meta! come hang out luma.com/2lraxd0n

AI Safety & Security Meetup · Luma

We are pleased to launch the first in a quarterly series of AI Safety and Security community meetups in London, co-organized by ARIA's Trust Everything,…

luma.com
1
3
32
1,873
love
Juan Benet
@juanbenet
Sep 20
Ugh no! — use a scalable name. This is what metric prefixes are for! Average human intelligence can be 1 Brian. And we can scale by rough neural network connectivity, as it’s easy to measure. NanoBrain - C Elegans MicroBrain - Fruit Fly - 2015 AIs MilliBrain - Mouse - 2020 AIs Brain - Human - 2026 AIs KiloBrain - 2030ish AIs MegaBrain GigaBrain - Humanity sums to ~8 GigaBrains. TeraBrain PetaBrain - Kardashev I earth brain world ExaBrain ZettaBrain YottaBrain - Kardashev II Dyson Swarm Brain Onwards to the YottaBrain! Notes: - Obviously you can have smarter or dumber intelligences in similar scale brains, but this is a rough approx of avgs. - Obviously intelligence doesn’t add: 1,000 people are not 1000x smarter than 1. Sum just for rough reference - We’ll make more energy efficient hardware, so take the Landauer limit into account - We’ll figure out how to make intelligence more computationally efficient (but probably not by that many orders of magnitude) - yeah yeah, there’s mistakes, correct them. And here is Claude’s version:
1
3
1,003
Happy to announce my latest work with @skominers and @SuccinctJT ! eprint.iacr.org/2026/2056 TLDR: we transfer the techniques from better.codes to yield 5-10% proof size reductions in several SNARKs today, and supply an efficient decoder for knowledge soundness 🧵/ n
6
26
3
146
10,075
Introducing Scaling Trust Arena: an agentic economy with a multi-million dollar prize pool. Together with @ARIA_research, @BT6_Official, and Amodo Design, we’ve set up a warehouse where anyone can submit agents to make physical goods or services, or to red-team others.
2
12
4
159
12,631
*(subject to contract and negotiation)
5
202
i recognize where this is coming from, and like the motivation for bipartisan gov action, but this is not the way safe development, adoption and societal transition to ASI should be the goal rather than banning it
etn.
@etnshow
Sep 11
BREAKING: 40 British MPs have signed a letter to Prime Minister Andy Burnham calling for a ban on the development of superintelligent AI.
3
7
1,443
many criticising this, the nuance of the text seems to be lost on people (or maybe they didn’t read it all) what mathematicians are saying in there makes sense to me, and seems solvable (probably by using AI to improve human explainability & scrutiny into the results)
Andrew Curran
@AndrewCurran_
Sep 11
Twenty-five Fields Medal winners have published a joint declaration warning about what they see as a severe misalignment between AI companies and the mathematics community.
4
10
1,497
nitter.cf/albertwenger/status/20…
Albert Wenger 🌎🔥⌛
@albertwenger
Sep 12
This is a nuanced position on the importance of advancing human accessible knowledge and the incentives required for that. Dismissing it as gatekeeping or luddite misses the point.
2
292
nitter.cf/_thomasip/status/20985…
Thomas Ip
@_thomasip
Sep 11
Replying to @ns123abc
if you spend time reading it, the main argument is not about attribution/plagiarism, but that the labs are presenting the final solutions as if that is all there is to the problems, when the understanding and insights are the core part of mathematics in the first place.
4
164
Today from Phillip Isola, MIT professor and computer vision legend: “Every robot with an internet connection becomes a potential tool for Fable, Astra, and other AIs.” 🧵
13
18
3
238
24,526
We're hiring. Think: Sr partner at a venture incubator, but the outputs are Focused Research Organizations Form the thesis, find the founders, build the boards, launch the org AI resilience: bioresilience, cyber-physical systems, epistemic infrastructure, compute governance
12
59
18
215
75,187
In an industry first, we’re piloting double-blind evaluations for frontier AI. By creating a secure environment where neither test prompts nor model weights are revealed, we can ensure external safety and performance evaluations of our models remain private, robust, and trustworthy. → goo.gle/3St2xan
121
102
48
855
352,417
New essay! Where new intermediaries emerge in the age of AI, why their power could be systemically dangerous even in careful hands, and how, by keeping their power checkable and contestable, we might reap AI’s benefits without surrendering pluralism, privacy or safety. scalingtrust.org.uk/blog/wit… AI discourse over the last few months has been full of parallels for me: the Mythos export controls reminding me of the Crypto Wars, the several letters on why open source open weights matters echoing the Free Software movement.. past doesn't repeat but it rhymes n all that AI has potent information dynamics, it both (1) does better the more data it is fed, (2) has tremendous capabilities at digesting such data. A dangerous flywheel. In private conversations with friends, and at events over the last months, recurring convos have come up: worries around AI-enabled surveillance and influence campaigns, the lack of an ~EFF thing for this era, and the gap in the discourse for something that is neither full-speed-ahead AI accelerationism nor control-centric AI safety. It's taken a while to gather my thoughts (and the help of a lot of people!), but I've now put together an attempt at laying out our thinking on the topic by looking at it from the lens of intermediaries. This essay also serves to further motivate the £50m r&d programme we're running on multi-principal, multi-agent security at @ARIA_research. Let me know what you think! Thank you to @James_D_Fox @sebkrier @lrhammond @lukaspet @MjaBradshaw @iamnotnicola @AmmannNora @m0namon @allisondman @ccatalini and Louise for very helpful conversations and reviews that shaped the post!!
8
12
4
87
7,619
there's been a lot of recent, at times great, discussion about how to spend impending AI windfalls well. here's my 2 cents: why not use it to stand up independent, secure compute clusters dedicated to open AI safety and science research? foresightinstitute.substack.… in a world in which research depends on compute while compute prices and competition are skyrocketing, securing access while we can seems critical.
5
22
9
115
30,688
No one believes the argument that Flock is just an extension of existing of surveillance cameras. As a whole, the system is something entirely new. p2p.coincenter.org/p/flock-i…

Flock is different

In some ways, ALPRs like Flock and similar technology-driven surveillance tools are nothing new. Why the backlash? Because the system as a whole is very new.

p2p.coincenter.org
6
16
3
77
15,887
very excited for this going live! 🤠
Institute for Responsible Superintelligence@RESI_org
Aug 24
The Institute for Responsible Superintelligence (RESI): a nonprofit founded by Shafi Goldwasser, Adam Kalai, Vinod Vaikuntanathan. @ShafiGoldwasser @Vinod_MIT @AdamFungi resi.org Our goal: build scientific foundations needed to make superintelligence safe by design
3
1
25
3,125
Load more