MMOOZZGG.com

Glossary

Who is Dario Amodei?

In short

Dario Amodei is a co-founder and the chief executive of Anthropic. He wrote an essay on how AI could speed up science, medicine and the economy.

People · Dario Amodei

MMOOZZGG.com editorsChecked: 1 min read

For a child

Imagine you are building a very strong robot helper. It can tidy your room in a minute, but it can also break a vase if nobody explains the rules. Dario Amodei thinks about how to make such helpers useful and careful. His company builds a smart program you can talk to and teaches it to follow rules and not harm people.

In plain words

Amodei leads Anthropic, the company that makes the Claude models. Before this, he did research on AI safety. One of his papers examines problems that look simple: a system does its task but breaks something on the way, or finds a loophole in its reward rules. Another work showed how to train a model on people’s judgments: a person compares two versions of behavior and picks the better one, and the model adapts to that choice. A similar approach is now used to tune language models. In October 2024 Amodei published an essay on how powerful AI could speed up medicine, science and the economy.

For business

For business, Amodei matters for two reasons. The first is the Claude models, which companies use for long documents, code and correspondence. The second is his view of risk. In a 2016 paper he and his co-authors listed the errors to catch before launch: side effects, loopholes in the scoring system and unsafe trial actions. This logic is worth bringing into your company. Before you give employees AI, write down what data they may send to it, who checks the answers and what to do after a mistake. Such rules go into an AI use policy. Companies in Russia check whether the service is available to them in the provider’s documents before they connect.

For a specialist

Amodei is the first author of “Concrete Problems in AI Safety” (June 2016). It names five practical problems: side effects, reward hacking, scalable supervision, safe exploration and distributional shift. In June 2017 Christiano, Leike, Brown, Martic, Legg and Amodei published “Deep reinforcement learning from human preferences”. In it, an agent learns from people’s comparisons of short pieces of behavior instead of a hand-written reward function, and feedback is needed for less than 1% of actions. This approach is close to reinforcement learning from human feedback (RLHF). On 28 May 2021 Anthropic announced a Series A round of $124 million, with Amodei as chief executive. In October 2024 he published the essay “Machines of Loving Grace”, with sections on biology and health, neuroscience, economic development, peace and governance, and work and meaning.

Example

A legal department introduces an AI assistant to review contracts. Before it opens access to all lawyers, the team checks for side effects. Does client personal data leak into the chat? Does the assistant edit text without marking the change?

Related terms

Questions and answers

What is Dario Amodei known for? Amodei is a co-founder and the chief executive of Anthropic, the company that makes the Claude models. He is also the first author of a 2016 paper on practical problems of AI safety.
What is the essay “Machines of Loving Grace” about? In the essay of October 2024, Amodei describes how powerful AI could speed up medicine, science and the economy. Its sections cover biology and health, neuroscience, economic development, peace and governance, and work and meaning.

Sources

Not sure whether your company needs this?

AI solutions map
Contact us