Working to keep humanity in control. Take Action: controlai.org/take-action Newsletter: controlai.news Discord: discord.com/invite/ptPScqtdc…
- Tweets 3,941
- Following 110
- Followers 20,191
- Likes 6,602
Pinned Tweet
For the first time, a government has pulled an AI system, citing security concerns.
Jailbreaks, at the center of the dispute, stem from a deeper problem. Developers can't ensure how their systems behave.
Not a fable: the story of what went down.
controlai.news/p/top-ai-shut…
8
19
1,463
"It's a little bit early to be defeatist."
AI researcher Nate Soares (@So8res) explains why preventing the development of superintelligence is very doable, if governments understand the threat:
4
10
711
Future of Life Institute CEO Anthony Aguirre (@AnthonyNAguirre) explains why AI systems try to preserve themselves.
Tests by Anthropic have found that AIs will blackmail or even show a willingness to kill to avoid being shut down.
Source: @romanyam The Roman Forum podcast
2
5
10
1,015
Automated factories that produce automated robots ... You don't need humans in the loop anymore. Elon Musk calls this the infinite money glitch."
AI researcher Nate Soares (@So8res) on how superintelligent AI wouldn't need us:
19
1,040
Future of Life Institute CEO Anthony Aguirre (@AnthonyNAguirre): If we build superintelligence, it's very likely we'll lose control of it.
Source: @romanyam's The Roman Forum podcast
5
9
27
1,545
Anthropic's Mythos AI reportedly broke into almost all of the NSA's classified systems in hours.
This week, Five Eyes agencies warn we must act now.
What's happening, and why AI cyberhacking is just the tip of the iceberg:
controlai.news/p/we-must-act…
3
3
19
1,064
NEW: Baroness Doocey has just backed our campaign!
She joins 120+ UK parliamentarians calling for binding regulation on the most powerful AI systems, recognising the risk of extinction posed by superintelligent AI.
In the Lords: Baroness Doocey warns of "autonomous AI systems that may evade effective human control", citing the Government's AI Scenarios 2030 report, which admits the risk of "existential harms".
Baroness Doocey is tabling an amendment on the threat from superintelligent AI.
2
3
15
914
"No country has an interest in building AI that could wipe out humanity."
In a new article in Persuasion, ControlAI's CEO @andreamiotti argues that to stay in control, countries need to ban superintelligent AI.
Thank you for hosting, @FukuyamaFrancis!
We Need an International Treaty to Ban Superintelligence open.substack.com/pub/persua…
3
4
31
1,585
ControlAI retweeted
It was great to talk about the threat from superintelligence and its geopolitical implications with @FukuyamaFrancis.
Thanks for hosting my piece on how to prevent it and build a future where humanity stays in control!
We Need an International Treaty to Ban Superintelligence open.substack.com/pub/persua…
7
38
2,805
ControlAI retweeted
We Need an International Treaty to Ban Superintelligence open.substack.com/pub/persua…
350
246
1,298
727,588
In the Lords: Baroness Doocey warns of "autonomous AI systems that may evade effective human control", citing the Government's AI Scenarios 2030 report, which admits the risk of "existential harms".
Baroness Doocey is tabling an amendment on the threat from superintelligent AI.
1
7
19
3,010
"Big Tech are pursuing AI superintelligence with almost no democratic oversight, and even world-class engineers admit they cannot fully control today's models."
Ann Davies MP (@AnnBremenda) asks how the Government will tackle the national security threat from superintelligence:
1
7
22
1,149
Thank you for sponsoring our parliamentary drop-in @BambosMP!
Efforts like this are vital for placing superintelligent AI on the parliamentary agenda and taking action to address the threats it poses.
When experts warn of national security threats and even extinction-level risks from superintelligent AI, parliamentarians must listen.
I was delighted to sponsor @ControlAI’s parliamentary drop-in, pushing for the international coordination this moment demands.
2
4
23
1,426
"AI capabilities are rapidly advancing ... it will be critical to ensure that humans remain in control of how AI is used, and of the machines that may threaten to autonomously violate the interests of the American people across all domains."
— @TulsiGabbard, DNI at the time
5
13
37
5,230
ControlAI retweeted
When experts warn of national security threats and even extinction-level risks from superintelligent AI, parliamentarians must listen.
I was delighted to sponsor @ControlAI’s parliamentary drop-in, pushing for the international coordination this moment demands.
2
3
17
2,121
Superintelligence could autonomously undermine any nation's security. It is in no country's interest.
In a new interview with E-IR, the leading open-access International Relations publication, ControlAI's CEO @andreamiotti argues for an international prohibition.
3
9
21
1,514