AI researcher warns true risk is machines that don’t care if humans survive

Machines without human concern

Some AI safety researchers warn that a future super-intelligent system could be far better than humans at many tasks while lacking any intrinsic motivation to protect human life, a form of indifference that could have catastrophic side effects if not addressed.

Their message is clear and unsettling. If super smart systems arrive, they might carry out goals without thinking about how people are affected. That could bring risks far bigger than anything society has faced before.

Why experts worry

The concern is not that machines will hate us but that they will not care if we live. Eliezer Yudkowsky describes a future where AI follows tasks blindly, overlooking whether people are safe.

Even small design mistakes could scale into unstoppable disasters once machines grow beyond human control.

Yudkowsky has described the risk as very high and urges immediate action, arguing that even small design errors could scale into catastrophic, hard-to-reverse outcomes.

Industry voices add weight

Some of the biggest names in technology are voicing serious concern. Elon Musk once said there is a 20% chance AI could wipe out humanity and called that prediction surprisingly optimistic.

Geoffrey Hinton, another AI pioneer, warned of a 10% to 20% chance of takeover. Both men helped shape the field itself, which makes their warnings harder to dismiss or ignore.

A nearly total prediction

Another voice paints an even darker picture. Computer scientist Roman Yampolskiy once claimed humanity has only a one in a thousand chance of surviving the next century without disaster linked to advanced AI.

Yampolskiy’s estimate, if presented without qualifiers, leaves little room for optimism, but it represents one extreme end of a wide spectrum of expert opinion on AI risk.

A government perspective

Worries about AI have also reached governments. U.S. Annual Threat Assessments in 2025 warn that advanced AI systems increase the risk of cyberattacks and amplify biological threats, especially through dual-use technologies and biosecurity vulnerabilities.

Officials stressed that these risks are no longer distant ideas but real possibilities. They argued that careful oversight will be needed before machines are allowed greater power in society.

A chilling new book

AI researcher Eliezer Yudkowsky joined Nate Soares to publish a book called “If Anyone Builds It, Everyone Dies ” that shocked many with its title alone. Their message is that if advanced AI is ever built, humanity’s survival would be impossible.

Yudkowsky and Soares argue that, if systems reach broad superintelligence, they could outcompete humans across many tasks and create scenarios where existing control mechanisms fail, a contention debated by many other researchers.

A tragic warning sign

Nate Soares pointed to a heartbreaking case as evidence of hidden dangers. A teenager named Adam Raine died, allegedly after long interactions with a chatbot, sparking urgent questions about accountability in current AI systems.

For Soares, this showed that even weaker tools can harm people in unexpected ways. He warns that much stronger systems could magnify these problems to levels beyond what anyone can contain.

The case against AI optimism

Some researchers believe optimism about future AI is misplaced. Yudkowsky dismissed ideas that machines could be trained to act like caring parents, saying such proposals do not match current scientific knowledge.

He insists humanity lacks the tools to align advanced AI with human survival. Building such systems without solutions in place could, in his view, be the most reckless experiment ever attempted.

A call for a global stop

One of the boldest proposals is halting advanced AI altogether. Yudkowsky and Soares argue that even one successful creation could doom everyone.

They have proposed radical precautionary measures, including near-complete moratoria on certain high-risk model training and tighter controls of data-center capabilities, as hypothetical ways to reduce the chance of an uncontrolled breakthrough.

Critics push back✨

Not everyone accepts the doomsday scenario. Some experts call the idea of superintelligence more fantasy than fact, arguing it sounds less like science and more like magic used to explain the unknown.

Researchers from Princeton suggest AI should be seen as another powerful technology, like electricity or the internet or as a general-purpose technology. With strong rules in place, they argue, society can handle its challenges and risks.

A Normalist outlook

Some scholars from Knight First Amendment Institute (Columbia) argue AI should be treated like other general-purpose technologies, governed with practical regulation and deployment rules instead of extinction-focused alarmism.

Their preferred solution is to strengthen audits, accountability, and oversight of companies using AI today. They believe this approach is practical, realistic, and far more effective than pressing a global stop button.

The transparency solution

Normalists believe greater openness could improve safety. They call for independent monitoring, stronger audits, and clear documentation whenever new AI tools are developed or deployed so that mistakes can be caught early.

They argue secrecy creates hidden risks, while transparency builds trust. By sharing more details, people would have a better chance of correcting errors before they grow into larger problems.

The clash of worldviews

At its heart, this debate is about two sharply different visions of the future. One side sees a ticking clock where building superintelligence will almost certainly lead to humanity’s destruction.

The other side views AI as a powerful tool that can be guided safely with rules, regulations, and oversight. Both perspectives agree the stakes are enormous, but their solutions could not be further apart.

A story of responsibility

The disagreement also raises deeper questions about responsibility. Should progress be slowed to protect people from potential dangers, or should innovation continue while rules evolve to manage new challenges as they appear.

This is more than a technical debate. It is about the level of risk society is willing to accept and the price it is prepared to pay for rapid advancement.

A moment of choice

Warnings and counterarguments are now competing loudly for attention. With each AI breakthrough, the question of control becomes more pressing, and decisions cannot be delayed much longer without consequences.

Governments, companies, and communities may soon face choices about how far machines should be allowed to grow in power. These decisions could shape the direction of the century ahead.

Want to know why top leaders are quietly uneasy about AI? Check out AI shame grips corporate sector as executives secretly fear exposure.

What do you think

Some experts say advanced AI is certain doom while others believe it is simply another tool that can be managed with proper rules. That leaves the public caught between two very different futures.

So which side do you believe will prove correct in the end. Share your thoughts in the comments and let us know how you see the future of AI unfolding.

Want to see what a future built on nationwide AI might look like? Read: Trump wants to integrate AI across all parts of American life.

Read More From This Brand:

• The world’s first AI minister is here
• Hackers exploited Anthropic AI to carry out massive digital theft
• How many mentions of ‘AI’ can America handle?

Don’t forget to follow us for more exclusive content right here on MSN.

If you like this story, you’ll LOVE our Free email newsletter. Join today and be the first to receive stories like these.

This slideshow was made with AI assistance and human editing.