Today we're diving into the techno-optimistic vision of Dario Amodei, the CEO of Anthropic. Published in October, Dario’s 15,000-word article, Machines of Loving Grace: How AI Could Transform the World for the Better, is an exciting read particularly if you’re bringing data science and machine learning to life.

The aspect of Dario’s article that made the biggest splash was his explicitly defining the term “Powerful AI”. He claims that he doesn’t like the term “Artificial General Intelligence” or AGI (although I happen to like it when it’s well-defined as it is by Google DeepMind; something I cover in Episode #748).

Dario defines quote-unquote “Powerful AI” as:

Smarter than a Nobel Prize winner across most fields – e.g., biology, programming, mathematics, engineering, and writing. We're talking about a system that can prove unsolved mathematical theorems, write exceptional novels, and develop complex software from scratch.
But here's where it gets interesting: this AI isn't just a "smart thing you talk to." like the Claude or ChatGPT UIs that AI predominates in today. Instead, Powerful AI has all the interfaces available to a human working virtually – text, audio, video, mouse and keyboard control, and internet access. It can take actions online, direct experiments, create videos, and communicate with humans, all at a level exceeding human capabilities.
This Powerful AI is autonomous too: It can handle tasks that take hours, days, or even weeks to complete, working independently like a smart employee would. While it doesn't have a physical body, it can control existing tools, robots, and laboratory equipment through computer interfaces.
Now, we can scale this system up: The resources used to train this model can be repurposed to run millions of instances of it. Each instance can process information and generate actions 10 to 100 times faster than humans. These millions of copies can either work independently on different tasks or collaborate like a team of humans would.

All in all, when this Powerful AI is created it would be what Dario referred to as a "country full of geniuses… in a datacenter."

While this Powerful AI is mind-blowing relative to today’s already tremendously impressive AI systems, there are limitations even to the Powerful AI – for example, biological experiments and hardware manufacturing still take time. Intelligence, while powerful, isn't magic fairy dust. You can’t instantly cure aging and all cancers just because you have vast intelligence. Humans and/or the Powerful AI can have a hypothesis as to cure, say, a specific cancer type, but then the hypothesis need to be tested for years in experimental animals and show promising, safe results before being tested on a small number of humans and then gradually more and more humans in well-controlled clinical trials. The timeline I just described is the best-case scenario where the hypothesis is proved correct by in vivo experiments, but most of the time breakthrough hypotheses are incorrect and lots of time-consuming iterations are required.

Nevertheless, the vision of Powerful AI is exciting. Potentially within as few as 5-10 years after developing this Powerful AI, we could see transformative changes across multiple domains.

In healthcare and biology, for example, we might compress a century's worth of progress into a decade or so. Many cancers could be eliminated, genetic diseases could be cured or prevented, and average human lifespan could potentially double to 150 years. Mental health conditions that have plagued humanity for millennia could become fully treatable, and we might gain unprecedented control over our biological and cognitive processes.

On the economic front, Dario envisions AI enabling developing nations to achieve unprecedented growth rates of up to 20% annually. This could lift billions out of poverty and help close the gap between developed and developing nations. We might see a second Green Revolution in agriculture, breakthrough solutions for climate change and dramatic improvements in food security worldwide.

Even in governance and social structures (which could be trickier to accelerate than technology) Powerful AI could potentially strengthen democratic institutions by improving government services, making legal systems more impartial, and helping create more informed and thoughtful citizenry.

So what's the timeline for this Powerful AI being realized? Is it decades away? Maybe. But Dario suggests Powerful AI could arrive as early as 2026 — that’s only two years away.

That could be a stretch (and Dario, as CEO of Anthropic, one of the leaders at the absolute cutting-edge of AI capabilities is highly incentivized to be highly biased in his optimism in order to encourage his investors and researchers), but by scaling up the AI capabilities we already have today (e.g., by scaling up the number of model parameters and scaling up inference-time compute like OpenAI’s o1 does) having a Powerful AI that exceeds human-level performance across virtually all cognitive tasks in only a few years is not wholly ungrounded science fiction. And, if we develop Powerful AI with effective safeguards, it could serve as a tool to create a more equitable, healthy, and prosperous world for everyone.

The SuperDataScience podcast is available on all major podcasting platforms, YouTube, and at SuperDataScience.com.