A study published in 2025 examining 303 accidents involving human–robot interactions notably highlights unexpected robot activations and sensor errors. With the emergence of Physical AI, artificial intelligence errors are gradually moving beyond screens and entering the physical world.

For several years, we have grown accustomed to artificial intelligence making mistakes.

A chatbot invents information. An image generator produces an anomaly. A model misinterprets a document.

In most cases, the error remains digital.

With Physical AI, this situation changes radically

When artificial intelligence controls a robotic arm, a mobile robot or, in the future, a humanoid robot, a misinterpretation of its surroundings can trigger a real physical action.

An error of just a few centimetres can have consequences far more serious than a ChatGPT hallucination.

303 accidents analysed

A study published in 2025 in the scientific journal Safety Science analysed 303 accident reports involving interactions between humans and robots.

The researchers identified seven major categories of incidents.

Two phenomena stand out in particular: unexpected robot activations and sensor-related errors.

This is precisely one of the challenges that Physical AI will need to address.

A robot must perceive its environment before it can act.

Cameras, proximity sensors, vision systems, contextual data and artificial intelligence models become its eyes and, to some extent, its decision-making system.

But what happens when that perception is incorrect?

A wrong decision becomes a physical movement

Consider an extremely simple scenario.

A robot equipped with a camera must identify a crate and move it.

Its vision system misinterprets the scene.

The software nevertheless decides that it has identified the correct object.

In a conventional computer system, this error might produce incorrect data.

In a robot, it can potentially produce an incorrect physical movement.

If an operator is standing in its path, the algorithmic error becomes a safety issue.

With Physical AI, an artificial intelligence
error no longer produces only an incorrect
answer. It can produce a physical
movement.

Accidents are already a reality

Data from the United States provide a tangible measure of this risk.

In its report on workplace accidents in the manufacturing industry, the US Occupational Safety and Health Administration identified 550 incidents involving robots.

In 33% of cases, the victims were treated in emergency departments, while 2% of incidents required hospitalisation.

Reported injuries included cuts, bruises and fractures.

Robot maintenance and cleaning also appear to be particularly sensitive situations.

Of course, not all these accidents can be attributed to artificial intelligence.

That distinction is essential.

However, they demonstrate what can happen when humans and machines capable of powerful movements share the same environment.

Physical AI is now adding a new variable: the machine’s decision-making autonomy.

From software bug to physical risk

This is probably one of the major changes that the robotics industry will need to address in the coming years.

Conventional software follows relatively deterministic logic.

New generations of robots will be increasingly capable of interpreting their surroundings, adapting their actions and making certain decisions in real time.

This capability is precisely what makes them valuable.

But it also creates a new problem.

How can we guarantee the safety of a machine when not all of its decisions can be predicted in advance?

The question becomes even more important with humanoid robots.

Unlike industrial robots, which have traditionally operated behind safety barriers, humanoid robots are specifically designed to work in environments created for humans.

• Factories
• Warehouses
• Hospitals
• Shops
• And potentially, in the future, our homes

Robot errors are becoming a new KPI

Until now, manufacturers have mainly measured availability, productivity, breakdowns and accidents.

Physical AI could introduce a new indicator:
the number of incorrect decisions made by a robot before they result in an incident.

A robot operating for 1,000 hours without an accident may appear extremely reliable.

But if an operator had to intervene 30 times to prevent an incorrect action, the risk assessment becomes very different.

These “near misses” could become an essential metric for assessing the true reliability of autonomous robots.

Accidents tell us what went wrong.
Near misses could show us what may
go wrong tomorrow.

Data remains insufficient to anticipate the risk

The problem is that organisations do not yet have a common language for measuring these situations.

An emergency stop, a trajectory corrected by an operator, an incorrect detection or a collision avoided by only a few centimetres are not always recorded in the same way.

Yet these events could provide the most useful warning signs before an accident occurs.

For example, a robot recording zero accidents over 1,000 operating hours does not present the same level of risk if it required zero, 10 or 30 human interventions to prevent a dangerous action.

The next generation of indicators will therefore probably need to distinguish between:

• Accidents resulting in injury
• Property damage
• Emergency stops
• Perception errors
• Unexpected activations
• Preventive human interventions
• Near misses with no consequences

Without this level of detail, companies risk underestimating the true danger posed by systems that appear to be functioning correctly.

The next battle for Physical AI may be about trust

Robot performance is advancing rapidly.

But if millions of autonomous machines are to work alongside humans every day, their intelligence will need to be more than just efficient.

It will need to be predictable, controllable and capable of failing without putting people at risk.

Artificial intelligence introduced us to a new category of error: hallucinations.

Physical AI now confronts us with a much more tangible question: what happens when a hallucination becomes a physical movement?

Anticipating the risk before the first major accident

The question is no longer whether autonomous robots will make mistakes.

They will.

The real question is whether we will be able to detect those errors early enough to prevent them from becoming accidents.

How should near misses be recorded? What thresholds should trigger a shutdown or a system reassessment? How can an AI model be tested in unpredictable physical environments? Who should be held responsible when an autonomous decision causes harm?

These issues already concern manufacturers, robotics integrators, safety managers, insurers, researchers and regulators.

At Robot Magazine, we want to open this discussion with the organisations and experts developing, deploying and regulating Physical AI.

If you work on autonomous robot safety, near-miss detection, embedded AI system certification or the prevention of human–robot risks, please contact us.

Together, we can begin defining the methods, indicators and best practices needed to anticipate this future challenge before it becomes a crisis.

Cet article Robotics: When an AI Error becomes a Physical Risk est apparu en premier sur Robot Magazine.

For decades, the technological singularity belonged more to the realm of science fiction than engineering. Today, the concept has become a subject of discussion in artificial intelligence laboratories, major industrial corporations, investment funds, and robotics companies around the world.

The spectacular acceleration of generative artificial intelligence, autonomous robots, foundation models, and AI agents has revived a fundamental question: what will happen when machines become capable of improving their own intelligence faster than humans can?

No one can say with certainty whether this “singularity” will ever occur. One thing, however, is certain: current advances in AI are forcing industrial companies to consider this possibility no longer as a philosophical curiosity, but as a genuine strategic issue.

A Sixty-Year-Old Theory

The origins of the technological singularity date back to 1965, when British mathematician I. J. Good, a former colleague of Alan Turing, imagined that an “ultra-intelligent machine” capable of designing even more powerful machines would trigger a genuine intelligence explosion.

In the 1990s, mathematician and author Vernor Vinge further developed this idea, arguing that once artificial intelligence reached a certain level, technological development would simply become impossible for humans to predict.

The concept was later popularised by Ray Kurzweil, now a principal researcher at Google, who believes that artificial general intelligence, or AGI, could emerge within the next few decades before gradually leading to the continuous self-improvement of intelligent systems.

Unlike today’s AI systems, which remain specialised despite their impressive performance, an AGI would be capable of learning, reasoning, and solving problems across virtually every intellectual field, with versatility comparable or even superior to that of a human being.

The singularity therefore does not simply refer to the moment when AI surpasses humans. It primarily describes the point at which intelligent systems become capable of improving their own algorithms, hardware architectures, research methods, and capacity for innovation.

The true AI revolution may not lie in its ability
to answer our questions, but in its ability to learn
without being shown every step.

Why Has the Debate Become So Serious?

Just five years ago, most industrial robots operated primarily through deterministic programming.

Today, the situation is changing extremely rapidly.

Large language models, vision-language-action models, autonomous agents, and foundation models are gradually enabling robots to understand natural-language instructions, interpret their surroundings, plan complex actions, and learn new tasks without task-specific programming.

Companies such as NVIDIA, Google DeepMind, Tesla, Figure AI, Boston Dynamics, Agility Robotics, Sanctuary AI, and Physical Intelligence are now investing billions of dollars in developing this new generation of intelligent robots.

At the same time, the global robotics market continues to grow.

According to the International Federation of Robotics, more than 4.2 million industrial robots are currently operating in factories around the world, while more than 540,000 new robots are installed every year. Robot density has now reached record levels, particularly in South Korea, Singapore, China, Japan, and Germany.

Goldman Sachs and Morgan Stanley estimate that the humanoid robot market could be worth tens of billions of dollars annually by 2035, with potentially several million robots deployed in manufacturing, logistics, healthcare, and service industries.

These figures explain why the singularity is no longer merely an academic subject.

The Computing-Power Revolution

One of the main drivers of this acceleration is the spectacular increase in available computing power.

The most advanced AI models are now trained on infrastructures composed of tens of thousands of graphics processing units, or GPUs.

Supercomputers dedicated to artificial intelligence now achieve performance measured in exaFLOPS while consuming several hundred megawatts of electricity.

The cost of training the most advanced models, which amounted to only a few million dollars several years ago, has now reached several hundred million dollars. Some analysts even estimate that future generations of models will require investments exceeding one billion dollars.

At the same time, every new generation of AI chips developed by companies such as NVIDIA and AMD significantly improves performance while reducing energy consumption per operation.

This simultaneous progress in hardware, algorithms, and infrastructure creates a virtuous cycle that continuously accelerates the capabilities of artificial intelligence.

When Robots Begin to Learn

For a long time, every new robotic task required weeks of programming.

That era is beginning to disappear.

Foundation models designed for robotics now make it possible to transfer knowledge from one task to another.

Instead of programming every movement individually, engineers can provide a few demonstrations or simply describe a task in natural language.

Platforms such as NVIDIA Isaac and the RT models developed by Google DeepMind perfectly illustrate this evolution.

Eventually, each robot could learn from the experiences of thousands of other robots connected to the same artificial intelligence model.

An improvement made in one factory could therefore benefit every other machine deployed around the world almost instantly.

Although this does not yet represent the singularity, this form of collective intelligence already constitutes an early form of large-scale continuous improvement.

The question is no longer simply how to build
more intelligent machines, but how to ensure
that they remain aligned with human interests.

Peter Thiel’s Perspective

Among Silicon Valley investors, PayPal co-founder Peter Thiel has adopted a particularly interesting position.

From an early stage, he financially supported organisations working on advanced artificial intelligence and invested in several major companies in the sector.

Unlike the most optimistic visionaries, Peter Thiel considers the singularity to represent both the greatest economic opportunity in history and the greatest technological risk humanity may ever face.

He has sometimes compared the arrival of artificial superintelligence to a first encounter with an extraterrestrial civilisation: the real issue would not be its power, but its ability to remain compatible with human interests.

Sceptics Point to the Limitations

Not all researchers share this optimism.

Current systems continue to produce errors, struggle to reason over very long periods, consume enormous amounts of computing resources, and still possess only an imperfect understanding of the physical world.

Many specialists believe that simply increasing the size of models will probably not be enough, on its own, to produce genuine general intelligence.

Others point out that physical limitations remain very real, including global semiconductor production, energy consumption, data-centre cooling, the availability of critical metals, and infrastructure costs.

The development of AI could therefore be slowed as much by economic constraints as by scientific challenges.

Alignment: The Real Challenge

Ultimately, the greatest challenge may not be building a more powerful intelligence.

It may be ensuring that it remains aligned with human objectives.

This is the central purpose of AI alignment research, which seeks to make artificial intelligence systems transparent, interpretable, controllable, and safe.

In robotics, this issue is even more critical.

Unlike conversational models, robots interact directly with the physical world.

A logistics robot, surgical robot, or autonomous construction robot will have to make decisions continuously in unpredictable environments, sometimes while working in direct contact with humans.

Safety could therefore become a competitive advantage just as important as the performance of the artificial intelligence itself.

A Transition Rather Than a Sudden Breakthrough

Will the technological singularity occur in twenty years? In fifty years? Or never?

No one can answer that question today.

One reality, however, is already clear: artificial intelligence is transforming robotics at an unprecedented speed.

Robots are becoming adaptive rather than programmed.

Factories are becoming software-driven.

Humanoid robots are gradually leaving laboratories and entering their first industrial sites.

Engineers now work every day with AI assistants capable of generating code, designing mechanical components, optimising trajectories, and analysing thousands of simulation scenarios.

Perhaps history will ultimately remember the singularity not as a sudden event, but as a long transition during which machines gradually stopped being simple tools and became genuine engineering partners.

From this perspective, the real question is no longer whether artificial intelligence will one day surpass human capabilities, but how industrial companies will learn to collaborate with systems capable of innovating alongside them.

Cet article What Is the Technological Singularity? est apparu en premier sur Robot Magazine.