ControlAI @ai_ctrl Twitter profile | Pikagi

Pikagi

ControlAI

@ai_ctrl

6,172

Followers

46

Following

367

Media

1,072

Statuses

We work to keep humanity in control. Our site: Our Discord: Our blog:

https://t.co/wOmlh5aAQH

Joined October 2023

Don't wanna be here? Send us removal request.

Pinned Tweet

@ai_ctrl

ControlAI

9 days

We need to do far more to stay in control of upcoming superintelligent AI. The new Head of AI Safety for the US government's AI Safety Institute – as well as multiple presidents of the most powerful AI companies – believe that superintelligent AI has a strong chance of killing

20

36

168

Last Seen Profiles

@naytayshaydraws

@perthTCH

@rey_cruz_2

@Mendoza40550220

@YiITyjRF1yoJgP9

@VidaBarrJones

@knknymdsk

@HalimaBidaoui

@itsu_rap

@adkeir_

@amoebamusic

@NiaArtist1

@eneadeiaa

@CoronadoEd96323

@qpnHH6TeTJrC5W7

@HrysSt

@CHZ5FmexkK46973

@leydicyrpto

@Lialarsen

@NickSnyder51

@rangaba

@sineadtaylor

@OceanTerra

@SheunopaTim

@abdulahi_carays

@Shoki4495126931

@Nommiexo

@shaqmwa

@TheLifeZoomer

@eno_muchinjo

@SK8141499408

@rikurus

@OYADO_TAKAMATSU

@shy459984577762

@tractor_owner

@gummywyvern

@ai_ctrl

ControlAI

13 minutes

In short: "In 2-3 years, AI will be so intelligent and powerful that it'll make the vast majority of human work redundant." "I see, and do you at all have a plan for this?" "Yeah, I don't have a good answer for that. Uh, maybe we could collaborate a bit with other AI

@ai_ctrl

ControlAI

2 hours

John Schulman, the co-founder of OpenAI, warns that AGI (a general-purpose artificial intelligence so powerful that it can supplant almost all human jobs) could be only 2 or 3 years away. This puts us all in an extremely risky situation – but he admits that doesn’t even have a

0

2

11

0

0

5

ControlAI Retweeted

@yashvardhan_07

Yashvardhan Sharma

@yashvardhan_07

2 hours

This was genuinely a very concerning thing from the podcast episode.

0

1

2

@ai_ctrl

ControlAI

2 hours

John Schulman, the co-founder of OpenAI, warns that AGI (a general-purpose artificial intelligence so powerful that it can supplant almost all human jobs) could be only 2 or 3 years away. This puts us all in an extremely risky situation – but he admits that doesn’t even have a

0

2

11

@ai_ctrl

ControlAI

1 day

🙋‍♂️ AGI stands for Artificial General Intelligence! It's a type of artificial intelligence which can outpeform humans on a wide range of cognitive tasks, in contrast to narrow AI, which is designed for specific tasks. Functional AGI would be able to make the vast majority of

@ESYudkowsky

Eliezer Yudkowsky ⏹️

2 days

I did not expect Jeopardy to just come out swinging like this (NOT FAKE)

Tweet media one

55

68

938

0

2

11

@ai_ctrl

ControlAI

1 day

Oui 👍

@StanislasBerton

Stanislas Berton

@StanislasBerton

2 days

Selon Donald Trump, l'IA est une des technologies actuelles les plus dangereuses car il est possible de prendre l'image d'une personne et de lui faire dire des choses fausses, sans que cela soit détectable. Dans le cas d'un président, cela peut conduire déclencher des guerres.

3

60

118

0

0

1

@ai_ctrl

ControlAI

2 days

OpenAI is bleeding safety researchers, losing 5 of their top talent in the last few weeks. This includes 4 from the Superalignment team, tasked with ensuring control over AI systems much smarter than us — which should be a key part of OpenAI’s plan. Here are the researchers who

@TolgaBilge_

Tolga Bilge

2 days

Another two safety researchers leave: Ilya Sutskever (co-founder & Chief Scientist) and Jan Leike have quit OpenAI. They co-led the Superalignment team, which was set up to try to ensure that AI systems much smarter than us could be controlled. Not exactly confidence-building.

Tweet media one

17

45

284

2

12

35

@ai_ctrl

ControlAI

2 days

Talking to Lex Friedman, the head of OpenAI Sam Altman admitted that "the road to AGI should be a giant power struggle...I expect that to be the case." But Artificial General Intelligence – an artificial intelligence that can perform as well or better than humans on a wide range

3

2

6

@ai_ctrl

ControlAI

2 days

We're profoundly grateful to have the MP Sarah Owen's backing in our campaign to get AI under control. This entire thread is brilliant, touching both on our work to ban deepfake sexual imagery as well as the use of AI generative materials for electoral interference. "This

@SarahOwen_

Sarah Owen MP

2 days

I’m proud to have the backing of @HOPEnotHate and @CCDhate in trying to get the law changed. This must come as part of a broader move to regulate AI companies, which is why I support @ai_ctrl (6/7) 👇

2

4

16

1

2

9

@ai_ctrl

ControlAI

2 days

A clear and disturbing pattern is emerging in OpenAI: whether of their own accord or not, figures involved in AI safety are leaving the company. Ilya Sutskever (co-founder & Chief Scientist) and Jan Leike have joined William Saunders and Daniel Kokotajlo in departing the company.

Tweet media one

Tweet media two

@janleike

Jan Leike

2 days

I resigned

1K

1K

11K

0

7

27

@ai_ctrl

ControlAI

4 days

Let's be honest – GPT-4o was weirdly flirty. And Sam Altman did say the team had been explicitly influenced by the film 'Her', where Joaquin Phoenix falls madly in love with a futuristic AI voice (well, we say futuristic – it was wildly, impossibly futuristic at the time, about

@OpenAI

OpenAI

4 days

What would you like to ask GPT-4o? We’ll pick some live prompts to do on our livestream happening now:

654

731

5K

0

2

14

@ai_ctrl

ControlAI

4 days

This should be understood as the world's rivalling superpowers coming together to better understand the risk of a new emerging world superpower: autonomous superintelligent AI.

@MarioNawfal

Mario Nawfal

4 days

🇺🇸U.S: AI IS AS BIG OF A THREAT AS NUCLEAR WEAPONS 🇨🇳CHINA: YEP... LET'S TALK The US and China will initiate formal government discussions on AI risks in Geneva. Spearheaded by Tarun Chhabra and Seth Center from the US and counterparts from China's Foreign Ministry and the

Tweet media one

Tweet media two

Tweet media three

Tweet media four

35

24

88

1

1

13

@ai_ctrl

ControlAI

4 days

we're keeping an eye on OpenAI's upcoming announcement, and we're hearing a lot of rumours about 'Her'-style AI girlfriends? I'm just saying that Futurama warned this would lead to total civilisational collapse...

Tweet media one

Tweet media two

Tweet media three

Tweet media four

1

0

5

@ai_ctrl

ControlAI

7 days

There's now a link to our blog in our bio, which is hosted on a site which gets your tweet algorithmically deboosted if you dare utter its name over here 👀 (It, uh, rhymes with 'wubwack'). So if you want to read (and even maybe subscribe to?) our blog on goings-on in the AI

Tweet media one

Tweet media two

0

3

6

ControlAI Retweeted

@tegmark

Max Tegmark

8 days

As this video says, humanity lacks a plan for not losing control of the the smarter-than-human AI it’s trying so hard to build:

18

22

128

@ai_ctrl

ControlAI

8 days

Our new video's up on Youtube now. Really, AI is grown, not built, so we're not fully in control of its development – and 'black box' technology means we're not entirely sure what's going on in there either. This should worry you. Watch to find out

Tweet card media

Superintelligent AI Should Worry You...

We need to do far more to stay in control of upcoming superintelligent AI. The new Head of AI Safety for the US government's AI Safety Institute – as well as...

www.youtube.com

1

6

11

@ai_ctrl

ControlAI

8 days

OpenAI's new Model Spec announcement claims that it can prevent their GPTs being used to break the law. But they don't seem to account for the range of jailbreaking techniques that have constantly bypassed GPT LLM guardrails. The Taylor Swift deepfakes from earlier this year

Tweet media one

Tweet media two

@ai_ctrl

ControlAI

1 month

Many-shot jailbreaking is a terrifying and dangerous vulnerability in AI LLM chatbots. And what's worse, the very meaning of many-shot jailbreaking is that the more complex and powerful AIs become, the more vulnerable they become to jailbreaking attempts – and the less control

Tweet media one

4

1

7

3

3

13