Rob Wiblin
@robertwiblin.bsky.social
Exploring the inviolate sphere of ideas one interview at a time: http://80000hours.org/podcast/
"There is no democratic element to the House of Lords and it is by far the most effective part of British system.
This is not a popular view, as you can imagine.
But you have to go on the evidence in front of your eyes." –
@iandunt.bsky.social
More: bsky.app/profile/robe...
This is not a popular view, as you can imagine.
But you have to go on the evidence in front of your eyes." –
@iandunt.bsky.social
More: bsky.app/profile/robe...
May 3, 2025 at 9:18 AM
"There is no democratic element to the House of Lords and it is by far the most effective part of British system.
This is not a popular view, as you can imagine.
But you have to go on the evidence in front of your eyes." –
@iandunt.bsky.social
More: bsky.app/profile/robe...
This is not a popular view, as you can imagine.
But you have to go on the evidence in front of your eyes." –
@iandunt.bsky.social
More: bsky.app/profile/robe...
"…in the final hours when people's lives could have been saved… the team ended up reformatting a table on Microsoft Word for Dominic Raab to look at while people were being left to the mercy of the Taliban." — @iandunt.bsky.social
From How British Government Got So Amateurish: youtu.be/1Yfo1aS0O6s
From How British Government Got So Amateurish: youtu.be/1Yfo1aS0O6s
May 3, 2025 at 9:12 AM
"…in the final hours when people's lives could have been saved… the team ended up reformatting a table on Microsoft Word for Dominic Raab to look at while people were being left to the mercy of the Taliban." — @iandunt.bsky.social
From How British Government Got So Amateurish: youtu.be/1Yfo1aS0O6s
From How British Government Got So Amateurish: youtu.be/1Yfo1aS0O6s
"10 Downing Street works on a system of professionalised loitering ...
People just kind of hang around thinking 'How do I get the most facetime with the prime minister — and how do I act as a barricade to stop other people...'" — @iandunt.bsky.social
Full episode here: youtu.be/1Yfo1aS0O6s
People just kind of hang around thinking 'How do I get the most facetime with the prime minister — and how do I act as a barricade to stop other people...'" — @iandunt.bsky.social
Full episode here: youtu.be/1Yfo1aS0O6s
May 3, 2025 at 8:58 AM
"10 Downing Street works on a system of professionalised loitering ...
People just kind of hang around thinking 'How do I get the most facetime with the prime minister — and how do I act as a barricade to stop other people...'" — @iandunt.bsky.social
Full episode here: youtu.be/1Yfo1aS0O6s
People just kind of hang around thinking 'How do I get the most facetime with the prime minister — and how do I act as a barricade to stop other people...'" — @iandunt.bsky.social
Full episode here: youtu.be/1Yfo1aS0O6s
Crazy to think that machines might someday surpass human reasoning.
February 16, 2025 at 2:59 AM
Crazy to think that machines might someday surpass human reasoning.
"It's crazy how the Republican Party Platform was already predicted by a comic about the effective supervillains over a year ago" @LinkOfSunshine
February 11, 2025 at 10:56 AM
"It's crazy how the Republican Party Platform was already predicted by a comic about the effective supervillains over a year ago" @LinkOfSunshine
No I think that's not right.
February 7, 2025 at 12:33 PM
No I think that's not right.
I asked the new ChatGPT 'Deep Research' to investigate how current US allies would try to defend their core interests if the US imposed tariffs on them and made threats to annex territory (or actually followed through).
Here was part of its answer.
So far seems like a powerful tool.
Here was part of its answer.
So far seems like a powerful tool.
February 4, 2025 at 7:29 PM
I asked the new ChatGPT 'Deep Research' to investigate how current US allies would try to defend their core interests if the US imposed tariffs on them and made threats to annex territory (or actually followed through).
Here was part of its answer.
So far seems like a powerful tool.
Here was part of its answer.
So far seems like a powerful tool.
The Vatican's new position paper on AI titled 'Ancient and new':
en.wikipedia.org/wiki/Antiqua...
vatican.va/roman_curia/...
en.wikipedia.org/wiki/Antiqua...
vatican.va/roman_curia/...
January 31, 2025 at 4:50 PM
The Vatican's new position paper on AI titled 'Ancient and new':
en.wikipedia.org/wiki/Antiqua...
vatican.va/roman_curia/...
en.wikipedia.org/wiki/Antiqua...
vatican.va/roman_curia/...
Huh, LLMs have a form of intuitive self-awareness of their own personalities.
January 31, 2025 at 10:29 AM
Huh, LLMs have a form of intuitive self-awareness of their own personalities.
Is AI scaling very impressive or very unimpressive?
As @tobyordoxford explains, it's both:
"AI capabilities have improved remarkably quickly... but if you examine the scaling laws they actually show extremely poor returns to scale. What’s going on?"
tobyord.com/writing/the-...
As @tobyordoxford explains, it's both:
"AI capabilities have improved remarkably quickly... but if you examine the scaling laws they actually show extremely poor returns to scale. What’s going on?"
tobyord.com/writing/the-...
January 31, 2025 at 10:28 AM
Is AI scaling very impressive or very unimpressive?
As @tobyordoxford explains, it's both:
"AI capabilities have improved remarkably quickly... but if you examine the scaling laws they actually show extremely poor returns to scale. What’s going on?"
tobyord.com/writing/the-...
As @tobyordoxford explains, it's both:
"AI capabilities have improved remarkably quickly... but if you examine the scaling laws they actually show extremely poor returns to scale. What’s going on?"
tobyord.com/writing/the-...
New release:
"The International AI Safety Report is the world’s first comprehensive synthesis of current literature of the risks and capabilities of advanced AI systems. It is the culmination of work by 100 AI experts..."
assets.publishing.service.gov.uk/media/679a0c...
"The International AI Safety Report is the world’s first comprehensive synthesis of current literature of the risks and capabilities of advanced AI systems. It is the culmination of work by 100 AI experts..."
assets.publishing.service.gov.uk/media/679a0c...
January 31, 2025 at 10:27 AM
New release:
"The International AI Safety Report is the world’s first comprehensive synthesis of current literature of the risks and capabilities of advanced AI systems. It is the culmination of work by 100 AI experts..."
assets.publishing.service.gov.uk/media/679a0c...
"The International AI Safety Report is the world’s first comprehensive synthesis of current literature of the risks and capabilities of advanced AI systems. It is the culmination of work by 100 AI experts..."
assets.publishing.service.gov.uk/media/679a0c...
A lot of fighting between people who make 0, 1 and 2 exceptions to the general rule.
January 31, 2025 at 10:27 AM
A lot of fighting between people who make 0, 1 and 2 exceptions to the general rule.
"It's true that the right to free expression and protest in the UK is insufficiently protected by law.
For instance the Home Secretary has the effective power to stop protests and protestors on the basis of next to nothing:
“The right to protest in Britain is maintained at the whim of a minister.”
For instance the Home Secretary has the effective power to stop protests and protestors on the basis of next to nothing:
“The right to protest in Britain is maintained at the whim of a minister.”
January 31, 2025 at 10:25 AM
"It's true that the right to free expression and protest in the UK is insufficiently protected by law.
For instance the Home Secretary has the effective power to stop protests and protestors on the basis of next to nothing:
“The right to protest in Britain is maintained at the whim of a minister.”
For instance the Home Secretary has the effective power to stop protests and protestors on the basis of next to nothing:
“The right to protest in Britain is maintained at the whim of a minister.”
Didn't realise how bad things had gotten in Myanmar.
After their coup in 2021 the government is failing to win a civil war.
en.m.wikipedia.org/wiki/Myanmar...
After their coup in 2021 the government is failing to win a civil war.
en.m.wikipedia.org/wiki/Myanmar...
January 31, 2025 at 10:24 AM
Didn't realise how bad things had gotten in Myanmar.
After their coup in 2021 the government is failing to win a civil war.
en.m.wikipedia.org/wiki/Myanmar...
After their coup in 2021 the government is failing to win a civil war.
en.m.wikipedia.org/wiki/Myanmar...
An AI generated song:
As an AI Language Model (I Can't Do That)
Click through for the rest, it's funny.
udio.com/songs/iu1381...
As an AI Language Model (I Can't Do That)
Click through for the rest, it's funny.
udio.com/songs/iu1381...
January 31, 2025 at 10:24 AM
An AI generated song:
As an AI Language Model (I Can't Do That)
Click through for the rest, it's funny.
udio.com/songs/iu1381...
As an AI Language Model (I Can't Do That)
Click through for the rest, it's funny.
udio.com/songs/iu1381...
Constantly emailed LLM spam from publicists these days. They're usually hilariously ridiculous. Here's a typical one:
January 31, 2025 at 10:20 AM
Constantly emailed LLM spam from publicists these days. They're usually hilariously ridiculous. Here's a typical one:
"For now, training still needs industrial compute. But it's looking increasingly like we won't be able to contain what comes after."
lesswrong.com/posts/7X9Bat...
lesswrong.com/posts/7X9Bat...
January 31, 2025 at 10:19 AM
"For now, training still needs industrial compute. But it's looking increasingly like we won't be able to contain what comes after."
lesswrong.com/posts/7X9Bat...
lesswrong.com/posts/7X9Bat...
"That sounds bad. But 60 times 7,000 is not 42,000. It is 420,000. ... The estimated exposure is not even a tenth of the reference dose. That does not sound as bad."
nationalpost.com/news/canada/...
nationalpost.com/news/canada/...
January 31, 2025 at 10:18 AM
"That sounds bad. But 60 times 7,000 is not 42,000. It is 420,000. ... The estimated exposure is not even a tenth of the reference dose. That does not sound as bad."
nationalpost.com/news/canada/...
nationalpost.com/news/canada/...
Seems super unlikely to me because to stay competitive companies will have to redesign their processes in ways that make it impossible for humans to participate (office work too fast and complex for humans to track let alone add to; factories too hot or dangerous for humans to be present).
January 31, 2025 at 10:18 AM
Seems super unlikely to me because to stay competitive companies will have to redesign their processes in ways that make it impossible for humans to participate (office work too fast and complex for humans to track let alone add to; factories too hot or dangerous for humans to be present).
Gwern comments on "Implications of the inference scaling paradigm for AI safety": lesswrong.com/posts/HiTjDZ...
January 31, 2025 at 10:17 AM
Gwern comments on "Implications of the inference scaling paradigm for AI safety": lesswrong.com/posts/HiTjDZ...
People marvel at the human brain but really it's a piece of soon-to-be-obsolete junk, the best evolution could manage with its severe limitations:
t.co/TxPFu87dgy
t.co/TxPFu87dgy
January 31, 2025 at 10:16 AM
People marvel at the human brain but really it's a piece of soon-to-be-obsolete junk, the best evolution could manage with its severe limitations:
t.co/TxPFu87dgy
t.co/TxPFu87dgy
January 29, 2025 at 12:10 PM
January 29, 2025 at 12:09 PM
Here are some of the specific organizations you could give to (a few more if you click through).
Another reason to give now is that AGI may come fairly soon, in which case the window for positive impact is ever shrinking.
80000hours.org/2025/01/it-l... 2/
Another reason to give now is that AGI may come fairly soon, in which case the window for positive impact is ever shrinking.
80000hours.org/2025/01/it-l... 2/
January 29, 2025 at 12:08 PM
Here are some of the specific organizations you could give to (a few more if you click through).
Another reason to give now is that AGI may come fairly soon, in which case the window for positive impact is ever shrinking.
80000hours.org/2025/01/it-l... 2/
Another reason to give now is that AGI may come fairly soon, in which case the window for positive impact is ever shrinking.
80000hours.org/2025/01/it-l... 2/