Is a benign digital superintelligence possible? What can we do to ensure a positive outcome as we continue towards developing AIs that are as smart and eventually much smarter than us?
In this second blog based on my recent Moonshots podcast with Emad Mostaque (who recently stepped down as CEO of Stability AI), I'll share his answers to these questions. I'll also discuss what he plans to do next, his vision for revolutionizing healthcare and education with AI, and why the fight for open and decentralized AI is so important and urgent.
Let's dive in…
The Promise and Peril of Digital Superintelligence
When Emad and I were on stage at the Abundance Summit just a few weeks ago, we talked about the possibility of a digital superintelligence—a future in which we've got AI that is a billion times more capable than a human, which, looking at it just from a ratio of neurons, is the ratio of a hamster to a human.
Does Emad believe that someday we could have a benign superintelligence that is supporting humanity?
"Yes," he says, "and I believe that it should be a collective intelligence that is made up of amplified human intelligence…versus a collected intelligence and an AGI that is top-down and designed to effectively control us."
As he explains it: "If you look at OpenAI's statements on the road to AGI, they say this technology will end democracy, end capitalism, and maybe kill us all. I don't like that."
Instead, Emad envisions a future where AI is distributed and reflects local cultures:
"If you have datasets for nations that are built on enhancing the capability of the nation…and you push for data transparency on models, which I believe we must have, especially language models, then you're more likely to have a positive thing."
Navigating Cultural Differences
To better understand Emad's views on decentralized and distributed AI, it helps to keep in mind his Massive Transformative Purpose (MTP): "I want every kid to achieve their potential and give them the tools to do that."
Here's how Emad explains the role of cultural traits in shaping our perceptions of AI:
"When you look at the Japanese concept of a robot, the robot is your equal and your helper. You look at the Western concept of the robot, it's Terminator and Skynet … This is where cultural norms become very interesting. What do we want to build? Do we want to build an AI god, or do we want to build that AI helper that helps us and we help it?"
This concept of taking into account how different countries and cultures will navigate this coming wave of AI development relates directly to what Emad is planning to do next now that he has stepped down as CEO of Stability AI…
Emad's Next Chapter...
Emad's plan is to help every country navigate the coming AI wave by setting up AI champions in each nation, working with the brightest people to guide them through this transformative period. He also believes that every sector should have a "generative AI-first infrastructure company" to help healthcare, finance, and other industries through this transition.
To coordinate all of this, Emad envisions a Web3-type protocol—not about tokens or speculation, but about creating a standard for intelligence. As he describes it:
"AI is an infrastructure, but it should be the rocket ship of the mind…I want to create organizations that [the smartest people in the world] can come to, the chefs and the cooks, the thinkers and the doers, and think: What is the future of finance? What is the future of education?"
A Revolutionary Vision for Healthcare
Emad and I are both passionate about the positive disruption and effects that generative AI can have in the healthcare industry. Here's Emad's vision:
"We're going to build a GPT-4 for cancer…Nobody is going to be alone on that journey and lose that agency because they'll have comprehensive, authoritative, up-to-date knowledge. AI models already outperform human doctors in empathy. They're not going to be alone anymore."
What Emad is suggesting here is critical. There are so many people left Googling to find information about the cancer or other diseases their loved ones are suffering from. The idea of having a model with the most cutting-edge information, which also incorporates a patient's medical data and gives them tailored, accurate advice in an empathic way, would be revolutionary.
And in building "Stable Health," or whatever it's ultimately called, the amount of capital required is de minimis compared to what's spent on a single human trial of any drug.
Transforming Education
Emad also has his sights set on disrupting education. In fact, that's how I originally met him—he was one of the winners of the Global Learning XPRIZE that Elon Musk and Tony Robbins had co-funded.
Now, with the advancement of large language models, Emad believes we're on the cusp of a transformative shift in education:
"We can have a GPT-4 level AI from us or someone else on a smartphone or a tablet by next year…You have an AI that teaches the child, learns from the child. Are you visual, auditory, dyslexic? That's the best data in the world for a national model, but also to teach these models how to be optimistic, how to be encouraging."
This really is the Young Lady's Illustrated Primer—Neil Stephenson's vision from his science fiction novel. And Emad is the only person I know of in the entire world working towards this vision.
Here's how he puts it:
"What happens if we have an entire nation of children that have this helper that brings in the right information at the right time and tells them they can always believe, that supports them? What can you do?…Some of the contributors to our open codebases for our models are 15 years old. They just taught themselves and happen to be wizards. In this new age, you don't know. And again, they should contribute to the whole because once something goes into this model or this system…it can proliferate to everyone using that system."
The Dream of Decentralized AI: Our Time is Now
One of the things I love about Emad is his passion and commitment to fighting for the cause of decentralized AI. And as he describes it, we are indeed in a fight:
"[Centralized AI companies] are building single systems that take our collective intelligence…and package it up, sell it back to us, but they don't care. These organizations are trying to build a system that will take away our freedom, liberty, and potentially kill us all. Let's be quite fair about that. Let's be direct."
Emad believes the only way to counter this threat is through decentralized, collective intelligence:
"When I resigned from Stability, I said you can't beat centralized intelligence with centralized intelligence. The only way you can beat it is to create the standard that represents humanity: decentralized intelligence. It's collective intelligence—datasets and norms that help children, that help people suffering, that reflect our moral upstanding and gather the best of us to do it."
The stakes couldn't be higher. As Emad warns:
"If we build an AGI as a centralized thing, is Windows or Linux safer as infrastructure? A monolith is likely to be crazy…Geniuses are not mentally stable. Why would you expect an AGI to be so? You're putting all your eggs in one basket versus creating a complex hierarchical system that is a hive mind. That's the intelligence that represents us all. We should be working towards building that because it's safer, it's better."
Ultimately, the question comes down to this: Do you believe in amplified human intelligence or artificial general intelligence? Do you believe in collective intelligence or collected intelligence? Is AI infrastructure or a product?
As Emad puts it:
"The time is now. It's our last chance to do it."
We don't have much time to figure this out. But with passionate visionaries like Emad leading the charge, I'm optimistic we can create a future of AI that uplifts and empowers humanity—not one that diminishes or destroys us.
The choice is ours. Let's make it count.