Center for AI Safety (@CAIS) / X

Center for AI Safety

285 posts

Center for AI Safety

@CAIS

Reducing societal-scale risks from AI.

San Francisco

Joined August 2022

Pinned
Center for AI Safety
@CAIS
May 30, 2023
We’ve released a statement on the risk of extinction from AI. Signatories include: - Three Turing Award winners - Authors of the standard textbooks on AI/DL/RL - CEOs and Execs from OpenAI, Microsoft, Google, Google DeepMind, Anthropic - Many more
Statement on AI Risk | CAIS
From aistatement.com
3M
Center for AI Safety
@CAIS
Jun 9
Replying to @CAIS
The newsroom:
Center for AI Safety Names Former Robinhood and Meta Executive Rochelle Nadhiri as Head of Public...
From safe.ai
489
Center for AI Safety
@CAIS
Jun 9
We're excited to welcome Rochelle Nadhiri (@RoWill) as our Head of Public Engagement. Rochelle brings two decades of storytelling experience at the intersection of technology and policy for industry leaders including Meta and Robinhood. At CAIS, she will lead our efforts to
1K
Center for AI Safety
@CAIS
Jun 3
Replying to @CAIS
Full article:
CAIS Research Scientist Mantas Mazeika Appointed to the EU AI Act Scientific Panel
From safe.ai
669
Center for AI Safety
@CAIS
Jun 3
We are pleased to share that @MantasMazeika96, Research Scientist at CAIS, has been appointed to the European Commission’s AI Act Scientific Panel (@DigitalEU). As a member, Mantas will advise the European AI office and national authorities on general-purpose AI (GPAI) models,
1.3K
Center for AI Safety
@CAIS
Jun 2
Replying to @CAIS
Full announcement:
CAIS Names Former xAI Leader Devin Kim President and Establishes Frontier Security Institute in...
From safe.ai
751
Center for AI Safety
@CAIS
Jun 2
Big news from @CAIS: Devin Kim (formerly @xai, @scale_AI) joins as President. We're launching the @FrontierSecInst, a DC-based org bridging frontier AI and the National Security Enterprise. Frontier AI is a national security technology. It's time to act like it. ⬇️
7.6K
Center for AI Safety reposted
Adam Khoja
@AdamK133
Jun 1
When labs trigger an intelligence explosion, they should worry about AI backdoors activating to sabotage their compute or their attempt. In a new paper, we study AI betrayal—how adversaries can make AIs work against their developers. 🧵
1.1K
Center for AI Safety
@CAIS
May 28
Replying to @CAIS
The full paper goes deeper on why groups (such as the public) would have an incentive to subvert AI systems, how they could do it, and the offense-defense balance. Read it here:
AI Deterrence by Betrayal
From aibetrayal.com
450
Center for AI Safety
@CAIS
May 28
Replying to @CAIS
The fear of AI betrayal may discourage reckless deployment, reduce confidence in fully automated systems, and make actors more willing to accept safeguards, monitoring, and transparency. We call this deterrence by betrayal.
542
Center for AI Safety
@CAIS
May 28
Replying to @CAIS
Developers may not be able to make the risk of betrayal negligible. This could have a stabilizing effect on the AI race. When you can't trust an AI, you hesitate to give it power over weapons, infrastructure, or fully automated recursive improvement.
336
Center for AI Safety
@CAIS
May 28
Replying to @CAIS
Covertly subverting AI systems may be relatively easy and inexpensive. Well-resourced actors, such as China and Russia, would likely be capable of far more sophisticated and damaging subversion attacks.
241
Center for AI Safety
@CAIS
May 28
Replying to @CAIS
Alternatively, people can build an AI to refuse criminal or illegal requests. But they don't fully control the model. Someone with power can order it handed over with the safety stripped, then use it however they want.
351