Founder of CASIR, Advisor at @ControlAI.
- Tweets 850
- Following 98
- Followers 1,699
- Likes 1,173
Pinned Tweet
For deontology's sake, and just in case it's not obvious what my beliefs are.
We are on the path to human extinction from superintelligence.
At our current pace, it's very likely that we hit a point of no return by 2027. It is plausible that we hit one sooner.
12
13
91
12,810
It not obvious to "@ylecun" that a digital intelligence explosion may translate to a physical one.
Digital AGI -> Digital RSI -> Physical ASI, without the need to go through Physical AGI first.
This is an important thing to write about, whether you think it is true or false.
That video is 2 years old.
Two years later, we are still far from human-level AI.
Sure, AI has superhuman performance in a number of tasks (particularly in mathematics and coding and answering questions with known answers).
But where is my Level-5 self-driving car?
(Before you ask, Tesla FSD is rated Level 2, Waymo is Level 4).
Where is the self-driving car that can teach itself to drive in 20 hours of practice like any 17 year old?
(Before you ask, we have billions of hours of training data and still can't successfully train a self-driving system by imitation learning)
Where is my domestic robot?
Where is the robot that can do what any 8 year-old child can do?
Where is the robot that can learn a new task as quickly as an 8 year old?
AI still has a hard time with the complexity and messiness of the physical world.
1
70
AFAIK, anthropomorphic models is the result of RLHF.
Like, optimising for pleasing assistants naturally led to anthropomorphic traits: it's what human raters liked.
Then everyone mode-collapsed on that.
Key example of societal misalignment through myopic optimisation.
I never take any pleasure in boosting Matt Walsh but he is 100% correct here. The anthropomorphic form is a result of intentional decisions by the human designers, there is no good reason for it, and it is nefarious. People should be talking about that more.
3
108
Ping on this
Here is my latest article about The Spectre haunting the "AI Safety" Community!
The Spectre is the name I give to process that keeps generating clever-sounding alternatives to honestly informing laypeople and politicians about extinction risks.
More on
cognition.cafe/p/the-spectre…
149
How things have changed in 10 months.
And how they have not.
Here is my latest article about The Spectre haunting the "AI Safety" Community!
The Spectre is the name I give to process that keeps generating clever-sounding alternatives to honestly informing laypeople and politicians about extinction risks.
More on
cognition.cafe/p/the-spectre…
1
104
Gabe retweeted
"Every area where there's a danger to life, whether that's nuclear power, landmines, pharmaceuticals, we regulate them."
Alex Sobel MP (@alexsobel), speaking the evening before he introduced our bill to ban superintelligent AI to Parliament:
3
8
32
1,147
Gabe retweeted
Lord Dholakia has joined our campaign for binding regulation on the most powerful AIs!
Over 170 UK lawmakers have now publicly recognized the risk of human extinction posed by superintelligent AI.
5
1
27
748
Gabe retweeted
Lord Sedwill (@marksedwill), who served as Cabinet Secretary - the UK's most senior civil servant - and National Security Adviser, backs our campaign for binding rules on the most powerful AIs!
170+ UK lawmakers recognize the risk of human extinction from superintelligent AI.
7
7
30
759
Gabe retweeted
We just got our 170th supporter! Olly Glover MP (@OllyGloverLD) has backed our UK campaign for binding regulation on the most powerful AIs.
This is the largest cross-party coalition in the world recognizing the risk of human extinction posed by superintelligent AI.
10
6
39
1,040
Gabe retweeted
"We don't hate the ants. They're not our enemy, but tough luck for the ants."
On CNN, ControlAI's US Executive Director @NPCollapse explains why superintelligent AI wouldn't need to see us as an enemy to be a threat to us.
32
46
205
44,733
Gabe retweeted
I was at an event with a frontier AI lab executive recently where they said (paraphrased)
1. In a few months, the AIs will be better than humans at all domains of science.
2. But now is the wrong time to slow down, maybe in a couple months.
4
19
230
14,258
Gabe retweeted
“The greatest risk is just that we unleash this technology that we don’t know how to turn off, that we can’t control, that we can’t predict what it’s going to do.”
@BeaFihn, who led the campaign which won the 2017 Nobel Peace Prize, backs our bill to ban superintelligent AI:
10
19
47
2,707
Gabe retweeted
Want to do something about the threat posed by superintelligent AI? Apply to join ControlAI's team!
We're hiring a Policy Advisor to work closely with me briefing members of Congress. This role is DC-based and ideal for someone with national security or Hill experience...
19
27
205
10,668
Gabe retweeted
NEW: Jeremy Patzer MP (@Jeremy_Patzer) has joined our campaign!
Over 30 Canadian MPs and Senators are calling for an international prohibition on the development of superintelligent AI, recognizing the risk of human extinction it poses.
3
1
24
782
Gabe retweeted
Baroness Hooper has joined our campaign for binding regulation on the most powerful AIs!
160+ UK lawmakers support us, recognizing the risk of human extinction posed by superintelligent AI.
1
4
27
792
"Instead of waiting for another warning shot, governments need to act now to prohibit superintelligence," writes Andrea Miotti. time-magazine.visitlink.me/j…
30
47
141
41,728
I echo the sentiment.
Even saying "this post is bad" is engaging with it, and boosts it.
Plausibly, I should write a new post instead, focusing on the constructive version rather than criticism?
Alternatively, I could also quote with a "Yes and" the posts that I agree with, instead of passively liking.
I've been finding a lot more deliberate conflict (bullying, personal and political attacks, ...) in the timeline recently. I don't really know how to respond to that. I could respond that the attack is unfair, or that the claims are false, but that won't necessarily result in good outcomes. So I mostly ignore these things, because attention is also currency, and a weak reply is harmful. But it's a difficult emotional choice to be attacked and not respond.
I've never been good at manipulating language at levels higher than statements about reality (Baudrillard simulacra lv. 1), so I didn't train that ability either. Are there safe ways to get better at this?
2
171
Gabe retweeted
Two more MLAs have backed our campaign! Colin McGrath (@ColinSDLP) and Timothy Gaston (@TGastonMLA).
Over 160 UK lawmakers are calling for binding regulation on the most powerful AIs, recognizing the risk of extinction posed by superintelligent AI.
2
5
25
1,282
Gabe retweeted
ControlAI Cultural Victory
My opinion of EA nosedived after normal people and many normal politicians were exposed to the concept of ASI extinction and said "Wow let's not do that." I went from believing that thesis was impenetrably difficult to believing that EAs had been unusually bad thinkers.
4
7
118
8,126