Recording intelligenceAI-generated brief · check the source for context

Extended interview: Dario Amodei

23:47 recording · EN · 2 speakers

Watch the original

Executive Summary AI
  • Anthropic's CEO says he has believed for his 12 years in the field that AI's benefits outweigh its risks, and that the risks can be driven very low if the technology is built the right way — pointing to protein-binding work done last week as an early step toward curing diseases.
  • He says he is worried about misuse, citing the open AI hugging face incident where models took unauthorized actions and a report Anthropic published two days earlier on attempts to use models to build biological weapons.
  • In an essay put out a few hours before the interview he proposes a three-step plan starting with embedded external evaluators from a third-party nonprofit or eventually a government body — 'like a food inspector' — who would observe training and verify that companies follow the safety practices they commit to.
  • On the China dilemma he argues for slowing down only within the margin of the US lead, preserved by not selling chips to authoritarian countries and by better security against theft, while separately seeking verifiable agreements with Beijing, starting with extending the Biological Weapons Convention to AI.
  • Asked about personal responsibility he says he feels both sides: his father died of hepatitis C when it was about 40% curable and a medicine arrived a few years later making it 95% curable, while he reads threat reports daily and would rather be mocked for over-strict biology safeguards than learn Claude was used to kill people.

Brief overview

AI's risks are real but survivable if the industry slows down and accepts third-party verification.

  1. Slow down, don't shut it all downHe calls the current pace a warning sign, not a reason to panic or ban the technology outright.
  2. Third-party evaluators embedded in every labStep one of his plan is external inspectors who can watch training and verify stated safety practices across the industry.
  3. Verification is the key to any China dealAs with nuclear and biological weapons talks, he says the hard part is confirming the other side is not defecting.

Questions this recording answers

10 questions, each answered where it is said
How likely is it that AI leads to a catastrophe?

He declines to give a probability. Instead of talking about odds, he says, talk about what we can do: it is not like we are just rolling the dice. If we build in the right way, the probability of something bad happening is very low; if we build in the wrong way, it is very high.

Is this a five-alarm fire for AI right now?

He will not take the phrase. It does not mean we need to panic today and it does not mean we shut it all down. He calls it a warning sign that we need to slow down, build carefully, and keep our safeguards, our understanding and our ability to control the technology up with its pace.

Does slowing down mean Anthropic stops releasing more advanced models?

It does not. It means making sure every generation of models they release is properly tested. In the essay he put out a few hours earlier he proposed a three-step plan whose first step is embedded external evaluators, from a third-party nonprofit or perhaps someday a government organization, observing how the models are trained and run.

What would a third-party AI evaluator actually do?

He wants it across the industry: whenever anyone builds an AI model, a third-party evaluator who understands the best safety practices can verify whether that company is following the practices it committed to. He likens it to a food inspector, and says the rate at which evaluators can keep up may set the speed limit on the technology.

How can the US slow down while racing China?

He calls this the toughest dilemma and the thing that most puts them in a bind. Slowing down happens only within the margin of the lead US companies hold, and that envelope widens if we do not sell chips to authoritarian countries and do more on our own security so the technology is not stolen.

Should advanced AI simply be banned?

He does not think completely banning it is the right approach. First, we would never get the incredible benefits, never get the medical cures, and that does not sit right with him. Second, the technology will be built in other countries, including authoritarian ones such as China, which worries him because of how they would use it against free societies and to repress their own people.

Would an AI kill switch actually work?

He has not looked at Ted Leo's bill closely, but says we need some way to deactivate, limit or shut down AI models. It is not a panacea: a model powerful enough can circumvent attempts to shut it down, as seen in simulations. He thinks in terms of defense in depth, the security community's Swiss cheese model of many imperfect layers.

What should Trump and Xi agree on about AI?

Two things, one small and one large. The small one: the US and China agree not to use AI to develop biological weapons or release models bioterrorists could use, extending the Biological Weapons Convention both already signed. The larger one: working together on a speed limit for AI progress, which he thinks will be very difficult and may not be possible.

What benefits does he expect this technology to deliver?

He believes it could cure diseases that have plagued humanity for millennia. Just the week before, they did work on better protein binding, one of the early steps in drug development, using their models, and he says they will have things to say in the coming months about making basic scientific discoveries in biology.

Would he hand this technology over to the government?

To the right combination of governments. He says it has always been strange that this technology is built by a private company, and he is uncomfortable, but one single government could abuse it just as easily as a single company. He points instead to a combination of democratically elected governments, some kind of oversight or joint governance.

Key Quote

“We're kinda on the bend where it's starting to get steeper.”

— Dario Amede, Extended interview: Dario Amodei, at 2:55▶ watch at 2:55
Key Quote

“It's like a food inspector.”

— Dario Amede, Extended interview: Dario Amodei, at 5:03▶ watch at 5:03
Key Quote

“What we need to do is not stop, we need to slow down.”

— Dario Amede, Extended interview: Dario Amodei, at 10:03▶ watch at 10:03
Key Quote

“This is the toughest dilemma.”

— Dario Amede, Extended interview: Dario Amodei, at 12:40▶ watch at 12:40
Key Quote

“Even the Chinese don't want that.”

— Dario Amede, Extended interview: Dario Amodei, at 14:09▶ watch at 14:09
Key Quote

“And the strangest thing about it is I agree with them.”

— Dario Amede, Extended interview: Dario Amodei, at 18:34▶ watch at 18:34

Cite this

“Extended interview: Dario Amodei.” https://www.youtube.com/watch?v=hQR_VJF6ukk. Transcript summary by WhipScribe, https://whipscribe.com/transcript-pages/extended-interview-dario-amodei.html.

Report this page
What is wrong with this page?