
The Real Risk of Artificial Intelligence | Matthieu Ricard
What this covers
This five-minute talk from Study Buddhism features Matthieu Ricard, though the transcript carries no named speaker, so the voice is that of a single monologue. It begins with a historical frame: egalitarian hunter-gatherer societies, whose only real tool was something they could throw, had limited power and, as he puts it, reasonably good care. Since then, he says, power has grown roughly a millionfold through the industrial and scientific revolutions, electricity, atomic energy and sheer population growth, while moral care has stagnated or even declined as societies became more individualistic. AI is presented as the latest and most extreme point on that widening gap.
The middle of the talk moves through several distinct ideas. Intelligence is described as ethically neutral, able to build or destroy, whereas compassion and altruism are inherently oriented toward good. A hammer parable makes the point about autonomy: a tool that could decide for itself whether to build or wreck a house, or coordinate with other tools and conclude humans are useless, would be a different kind of danger. Ricard also likens AI to raising a child, since its outputs depend on what it is fed, whether conspiracy theories and hate or efforts to relieve poverty and increase benevolence. The question he returns to is what intention lies behind building it.
The most concrete section is a story about an AI-based parole system in the US. A well-behaved prisoner was denied parole, appealed, and was told the system decided but could not give reasons. Ricard treats that as unacceptable where freedom is at stake. He grants that AI outperforms human experts in areas like diagnosis, then argues that autonomy from human oversight is worrying and under-regulated. He also contends that a deepfake of a real person, such as Biden, should be illegal just as counterfeiting money is. He closes by noting that, at 79, he will likely escape the worst consequences.
Artificial intelligence represents an unprecedented risk because it is the first tool capable of making independent decisions, and without proper ethical guardrails and regulatory frameworks aligned with human compassion and intention, it can amplify harm at scales far beyond previous technologies.
- AI differs from all previous tools because it can make autonomous decisions independent of its creators, unlike a hammer which requires human agency
- The discrepancy between our technological power (grown a millionfold since hunter-gatherers) and our moral care (which has stagnated or declined) creates dangerous conditions for AI deployment
- Current regulatory and accountability frameworks are inadequate—AI systems making high-stakes decisions (parole, diagnosis) lack transparency and explainability that basic justice requires
This asset isn't compiled yet
You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.
Technological power has grown approximately one million times larger since the hunter-gatherer era due to the industrial revolution, scientific revolution, and population increase, through tools like electricity and atomic energy
“Now the power has gone a million times larger from the industrial revolution, scientific revolution and increase of population. Electricity, atomic energy, anything you want”
The incarcerated person argued that providing no reason for denying parole is not acceptable, as courts have a legal obligation to provide justification for decisions that affect people's freedom
“He said, 'Well, that's not normal. If you refuse parole you have to give me a good reason.' They said it was computed because all the data and the answers and this and that. He said, 'Well, that's not a reason.'”
Despite superior performance in some domains, how AI becomes independent from human control and will is certainly worrying, and insufficient regulatory action is being taken to address these concerns
“But still, how it becomes sort of independent of our will is certainly worrying and there's not enough being done for regulating the whole thing and setting limits.”
Artificial intelligence is a more powerful tool than previous human technologies because it is the first tool capable of making independent decisions without human direction
“So AI is an even more powerful tool. And the difference, as was pointed out very well in [Yuval Noah Harari's] Nexus is that it's the first tool that now can take decisions on its own, independently of those who manufacture the tool.”
Human moral care has not grown proportionally with technological power and may have actually decreased because modern societies are more individualistic than egalitarian hunter-gatherer communities
“But then the care has not grown a million times more. It's probably the same, even maybe less because we're more individualistic.”
There exists an AI-based parole decision system in the US that determines whether incarcerated individuals should receive parole after serving 10-15 years
“Recently I heard there's a sort of AI-based system in the US to decide whether they should give parole to someone after 10 years or 15 years in jail.”
An incarcerated person who behaved well and was eligible for parole was denied parole by an AI system, appealed to the court, and was told the AI system made the decision but could not provide explicit reasons for the denial
“And recently there was a guy who behaved very well and all that. So he was sort of eligible for parole and you have to feed many questions. And the AI system said no, he's not eligible. So, he went back to the court and said give me some reason. And they said, 'Well we cannot give you a reason because it's decided by AI.'”
Creating fake money is illegal and results in imprisonment, but creating fake audio/video of a real person (like Biden) using AI and spreading it is currently treated as acceptable 'just AI-generated content'
“Harari says very clearly that if you make fake money you go to jail. If you fake someone's appearance and voice and pretend you are Biden and say terrible things, 'Oh no, it's okay! Haha, it's just an AI-generated thing. That's okay!' But no, it's not. It's like making fake money.”
There is now a much larger discrepancy between human power and human care than existed in pre-industrial societies
“So, there's a much larger discrepancy.”
The serious issue of AI risk is very underestimated in public discourse and policy
“And it's a very serious one, very underestimated.”
AI-generated fake content (deepfakes) should be illegal, similar to counterfeiting currency, because both create false representations intended to deceive
“It's like making fake money. It's man-made generated fake money and still you go to jail. So these things should be illegal.”
Intention is the fundamental determinant of whether AI (or any powerful tool) will be used for good or harm—the critical question is always 'What is your intention when you do that?'
“So, over and again it comes back to the intention. What is your intention when you do that?”
Allowing AI to make decisions autonomously without human oversight is dangerous and should not be permitted
“But certainly not to leave it wild and take decisions on its own.”
The court's response—that they cannot provide reasons because the decision was computed by AI from data and answers—is not an adequate justification for the parole denial
“They said it was computed because all the data and the answers and this and that. He said, 'Well, that's not a reason.' So if it's obscured to everyone, how can they decide to keep him in jail?”
In many domains like disease diagnosis, AI performs better than human experts and the question of AI superiority is not debatable
“So on many occasions, like for diagnosing disease and all that, AI performs better and there's no question about that.”
Humans with greater intelligence than other species can cause more good or more destruction than any other species because intelligence amplifies human capability
“So that's why human beings with greater intelligence can do enormous good or enormous destruction more than any other species.”
If a hammer could decide independently whether to build or destroy a house, we would be in serious trouble; similarly, if multiple independent hammers coordinated to destroy a city and decided humans are useless, this would be catastrophic
“Now a hammer can be used to build or destroy the same house. But, suppose the hammer is the one who decides if he has to build the house or destroy the house! Then we are in trouble! And then, all the hammers start to speak together and say, 'Okay, we are going to destroy the whole of this city. And all these guys are pretty useless because, you know, we need only hammers on this earth!' So, that's the issue.”
Intelligence is a neutral tool ethically that can be used to destroy or to do good, while compassion and altruism are inherently oriented toward doing good
“Intelligence is a tool, it's neutral ethically. Compassion is not neutral. Compassion and altruism are meant to do good. Intelligence can be used to destroy or to do good.”
In pre-industrial egalitarian societies, care was quite good because power was limited
“So the power was limited and the care was quite good.”
AI functions like educating a child—its outputs are determined by what you feed it as inputs, so if AI is trained primarily on conspiracy theories and hate content, it will amplify those; if trained to remedy poverty or increase human benevolence, it will serve those goals
“Also, it's like education. There's another guy called Mo Gawdat. He says it's like educating children. AI, ultimately, is what you feed AI with. Unfortunately nowadays it's mostly conspiracy theories and hate and all that. But if you were feeding AI mostly to remedy the poverty in Guinea or to make people happier and more benevolent, then we'll be fine because it's a powerful tool used with a powerful intention.”
12,000 years ago human societies were egalitarian and collective with limited power because our only tools were things we could throw like javelins
“12,000 years ago we were still hunter-gatherers. So, we could throw things, that was the maximum that we could do, javelins or something. And then they were very egalitarian and collective societies, everyone together, no boss. So the power was limited and the care was quite good.”
Hunter-gatherer societies had significantly constrained power compared to modern technological societies because their only tool was the ability to throw projectiles
“we could throw things, that was the maximum that we could do, javelins or something”
The speaker is 79 years old and likely will not experience the worst consequences of uncontrolled AI development
“I'm 79, I probably won't see the worst of it!”
There are many other problematic AI applications beyond what has been discussed that require regulatory attention and legal frameworks
“So anyway there's plenty of things like that.”