Claude founder: AI will completely surpass humans in just 2-3 years

I am confident that in the next 2-3 years, AI systems stronger than humans will emerge in various fields.

In a recent interview, Dario Amodei, CEO of Anthropic, the maker of Claude, discussed his views on the future of AI.

He believes that AI could indeed replace some human jobs in the short term, but making AI complementary to humans is the most beneficial way for productivity.

For today's young people, Amodei also offered his advice, encouraging them to learn how to use AI tools and develop critical thinking skills.

At the same time, Amodei also teased some recent updates about Claude.

“AGI is more like a marketing term”

Note: This article is based on an interview with Amodei conducted by Joanna Stern of The Wall Street Journal and CNBC's "Squawk Box."

Since the release of OpenAI's o1, reasoning models have become a hot new product, and many believe that o1 has opened a new paradigm for large models.

However, Amodei has a different perspective; he believes that reasoning models and "test-time computation" are not entirely new paradigms but rather the result of scaled reinforcement learning.

People have been talking about reasoning models and test-time computation as if there are ordinary models and reasoning models, with completely different ways of working.

This is not our view. We believe it is more like a continuous spectrum, where models have the ability to think and reflect on their thinking, ultimately arriving at a result.

But I think the change we will see is that we will see reinforcement learning used on a larger scale, allowing models to think and reflect more.

So reasoning or test-time computation is not a brand new method; it is more like an emergent property, the result of training models in a more outcome-oriented way on a larger scale.

Regarding the key to competition among large models, Amodei emphasized the importance of model differentiation, and he also gave a shout-out to his own Claude, stating that Claude's selling point is its unique personality traits.

Regarding Claude, the feedback we hear is that people really like its personality.

People enjoy interacting with Claude, seeking relationship advice and help with problem-solving, which occurs much more frequently than with other models.

Not only does the model have personality, but Amodei's views on AGI, which vendors including neighboring OpenAI particularly like to mention, are also quite different

To me, AGI has never been a clearly defined term; I have always considered it a marketing term.

But you know, my thought is that at some point, we will have artificial intelligence systems that are better than almost all humans at nearly all tasks.

“AI Will Surpass Humans in the Next 2-3 Years”

Of course, Amodei is very optimistic about the future development of AI.

He believes that within the next 2-3 years, AI systems stronger than humans will emerge in various fields.

He also believes that the speed of such progress is consistent with public perception.

My view when I first entered the AI field a few years ago was that we would have models that are stronger than almost all humans in nearly every aspect.

I guessed this would happen at some point in the 2020s, possibly in the mid-2020s, but I always said I don’t know, I’m not sure. We could be wrong.

Until about three to six months ago, I still thought there was considerable uncertainty, and even now there is, but that uncertainty has greatly decreased.

I am confident that in the next two to three years, we will indeed see such models—they are assistants to humans but will gradually become stronger than us in almost every aspect.

But does surpassing mean replacing? Amodei believes that in the short term, the real productivity improvement relies on using AI as a complement to humans rather than a replacement.

About a year ago, economist Eric Bernyolfssin conducted an interesting study, stating that general AI companies tend to deploy products in a replacement mode.

However, when these companies think more deeply, they often find ways to deploy AI in a complementary mode, which leads to greater productivity improvements.

Therefore, we can also think from the product perspective about how to design our products to achieve complementarity rather than replacement.

But in the long run, Amodei believes that once AI surpasses humans in almost all fields, many fundamental assumptions about economic organization, human self-worth, and more will be challenged.

In the long term, I have a strong feeling that AI systems will be stronger than humans in almost every aspect.

When that happens, how do we organize our economy? How do humans find meaning?

When humans are the most intelligent species on Earth, we have made many assumptions that will become invalid due to the development of AI.

In addition, Amodei also gave advice to young people, suggesting that they learn to use AI tools and develop critical thinking skills to discern the truth of information.

Claude plans to launch "Virtual Collaborator" this year

During the interview, Amodei also teased some recent product plans for Claude, stating that enterprise-level services are a key development direction.

Regarding the much-anticipated connectivity feature, Amodei mentioned that it is "coming soon, but the specific timing is undecided"; bidirectional voice is also in development, but it is not a high priority.

At the same time, he revealed that Anthropic is developing some new large models, expected to be released in the next 3-6 months.

These models will adopt a strategy different from other companies, aiming to better coordinate reasoning capabilities with other abilities of the model.

Another development in progress is a "Virtual Collaborator."

This is an AI system that can perform various tasks on a computer (such as writing code, testing, communicating, drafting documents, etc.), allowing users to engage in long-term task-oriented interactions with it.

As for image and video generation, it is not a focus of Anthropic's research and development; if necessary in the future, they will also choose to "borrow" from specialized image generation companies.

Quantum Bit, original title: "Claude Founder: AI Will Fully Surpass Humans in Just 2-3 Years"

Risk Warning and Disclaimer

The market has risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment goals, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investment based on this is at their own risk