We Should Slow the AI Race, Not Stop AI

The calls for an AI slowdown are becoming louder. Whether they will actually slow the technology is another question.

Some of the people closest to the development of advanced AI are warning that the industry may be moving faster than our ability to understand and control what it is creating. Anthropic chief executive Dario Amodei has called for the development of powerful AI to slow, while former AI researchers have spoken publicly about fears over where the technology could ultimately lead.

These warnings should not be dismissed.

When people who have worked directly on some of the world’s most advanced AI systems express genuine concern about the direction of travel, there is a responsibility to listen.

But there is another question that deserves equal attention.

Should a call to slow AI apply to artificial intelligence as a whole?

We do not believe it should.

There is a further question that may prove even more important for investors: will the call for a slowdown actually be heard?

The forces pushing AI forward are enormous. Technology companies have significant commercial incentives to continue improving their systems. Governments have strategic and economic reasons to maintain technological capability. Investors are committing substantial capital to the sector. Competition between companies and countries creates another powerful incentive to keep moving.

The AI race may therefore continue, even as concerns about its consequences become louder.

For JDP Global, that makes the debate less about choosing between optimism and pessimism and more about understanding where the technology is heading, where the risks are greatest and where the opportunities remain compelling.

The AI race may not stop

Technology rarely moves according to a single timetable.

It moves according to incentives, competition and opportunity, and all three are currently pushing AI forward.

Even if governments introduce stronger regulation and leading AI companies agree that certain capabilities should be developed more cautiously, there is no guarantee that the overall trajectory of advancement will change dramatically.

A company that voluntarily slows development while competitors continue could risk losing market share and technological leadership.

The same principle applies at a national level.

If one country significantly restricts the development of advanced AI while another continues investing heavily, the result may not be a safer world. It could simply change who develops the most capable systems.

That does not make the calls for caution irrelevant.

It makes them more complicated.

The realistic possibility is that AI development will continue pushing into areas that currently appear difficult to predict. Systems are likely to become more capable, more autonomous and more deeply integrated into businesses, infrastructure and everyday life.

If that happens, the question becomes how society manages that trajectory rather than whether it can simply stop it.

The case for slowing frontier AI

There is a meaningful distinction between AI that assists a person with a specific task and AI that can independently pursue complex objectives.

The most advanced systems are increasingly capable of reasoning, planning, using tools, writing and executing code, interacting with external systems and carrying out tasks with less direct human involvement.

That progression creates a different category of risk.

An AI system helping a researcher analyse medical data is one thing. An autonomous system capable of accessing computer networks, manipulating information, coordinating other systems or operating for extended periods without human supervision is something else entirely.

The more autonomy a system possesses, the greater the potential consequences when something goes wrong.

That is where calls for a slowdown deserve serious consideration.

The concern is not necessarily that every advanced AI system will become hostile or that a science-fiction scenario is inevitable. The more immediate issue is that developers may not fully understand the behaviour and capabilities of increasingly sophisticated systems before those systems are deployed at scale.

If a technology becomes more powerful, more autonomous and more difficult to control, the standard of testing and oversight should increase with it.

That does not mean stopping research.

It means recognising that there is a difference between developing a technology and deploying it without sufficient safeguards.

AI is not one technology

The problem with a blanket slowdown is that “AI” now describes an enormous range of technologies.

The same broad term can refer to an algorithm assisting a radiologist, a system helping scientists discover new medicines, software controlling a warehouse robot, an AI model analysing financial data, a brain-computer interface interpreting neural signals or an autonomous agent capable of carrying out complex tasks online.

These applications do not carry the same level of risk.

They should not automatically face the same level of restriction.

AI is becoming a foundational technology that intersects with healthcare, robotics, neuroscience, manufacturing, data infrastructure, transport and scientific research.

This convergence is reflected in JDP Global’s 2040 Precision Fund, which identifies Autonomy, Robotics, Brain-Computer Interfaces, Neuroscience, Data Lakes and Artificial Intelligence as interconnected areas of long-term interest.

A blanket approach to AI regulation risks overlooking that reality.

The better approach is to consider three things.

• How capable is the system?
• How autonomous is it?
• And what happens if it gets something wrong?

Those questions provide a far more useful basis for regulation than simply asking whether AI is involved.

Healthcare shows why the distinction matters

Healthcare is one of the clearest examples of where an indiscriminate slowdown could have unintended consequences.

AI is already being applied to areas including medical imaging, clinical research, drug discovery, transcription and the analysis of large datasets.

There are obvious reasons to be cautious when AI is involved in healthcare. A system making an incorrect diagnosis or influencing a critical medical decision can have serious consequences.

That is precisely why medical AI should be subject to rigorous testing, validation and oversight.

But there is an important distinction between AI replacing human judgement and AI helping humans make better decisions.

A system that assists researchers in identifying patterns across millions of pieces of medical information is not necessarily creating the same risk as an autonomous AI system making unsupervised decisions about individual patients.

Drug discovery, for example, can involve analysing enormous numbers of possible molecular combinations. AI can help researchers narrow those possibilities and identify promising areas for further investigation.

Medical imaging presents another opportunity. AI can assist clinicians by identifying patterns in scans and highlighting areas that warrant closer examination.

The appropriate response to these developments is not necessarily to slow them down.

It is to make sure that the technology is tested properly, that its limitations are understood and that humans remain responsible for consequential decisions where appropriate.

There is a difference between saying that AI should be trusted without question and saying that AI should continue to be developed under appropriate controls.

That distinction matters.

Neuroscience could be one of the most important examples

The relationship between artificial intelligence and neuroscience provides an even more compelling argument for continuing responsible development.

Brain-computer interfaces are designed to create a direct communication pathway between neural activity and an external device. The technology has the potential to help people with severe disabilities communicate, interact with computers and regain elements of independence.

AI has an important role to play in interpreting complex neurological signals.

The brain does not communicate with a computer using simple commands. Neural activity is enormously complex, variable and difficult to interpret. Increasingly sophisticated algorithms can help identify patterns within that information and translate them into useful outputs.

This is one reason JDP Global considers BCI and neuroscience important long-term investment areas.

If AI can ultimately help someone who has lost the ability to speak communicate again, should the technology be slowed simply because other forms of AI present potential systemic risks?

The answer should be more nuanced.

We should be extremely careful about technologies that could pose systemic risks to society.

But that does not mean slowing every application of artificial intelligence, particularly where development can take place within a controlled environment and the potential benefits are substantial.

Robotics brings AI into the physical world

The same principle applies to robotics.

AI is increasingly becoming the intelligence behind machines capable of operating in the physical world.

Robotics could transform manufacturing, logistics, agriculture, healthcare, rehabilitation and dangerous industrial work. Autonomous systems could eventually perform tasks in environments that are unsafe or impractical for humans.

There are risks.

A robot operating autonomously can cause physical harm if it malfunctions or behaves unexpectedly.

But these risks can potentially be managed in ways that are very different from managing an autonomous digital system operating across the internet.

A manufacturing robot can operate within a defined physical environment. Its movements can be restricted. Its operating parameters can be established. Emergency shutdown systems can be built into its design. Human operators can monitor its activity.

None of those measures make robotics risk-free.

They do, however, demonstrate why regulation should consider the environment in which AI operates rather than treating every AI application as identical.

The difference between an autonomous industrial robot operating within a controlled facility and an autonomous AI agent operating across the open internet is substantial.

The regulatory framework should reflect that difference.

Could slowing AI also slow the solutions to some of our biggest problems?

The warnings about AI often focus on what the technology could do to society.

There is another question.

What might society lose if development is slowed too broadly?

AI has the potential to help researchers analyse complex scientific problems, identify patterns in huge datasets and accelerate areas of research that have traditionally taken years or decades.

• It could contribute to advances in medicine, materials science, energy, manufacturing and neuroscience.
• It could help people with disabilities interact with technology.
• It could make dangerous workplaces safer.
• It could improve the efficiency of complex systems.

None of this means that AI should be given unrestricted freedom.

It means that the debate needs to move beyond a simple choice between acceleration and restriction.

There should be room for both.

AI could become part of the solution to AI risk

There is also an interesting paradox at the centre of the debate.

As AI systems become more capable, they may create new risks.

But those same capabilities could potentially help us manage technological risks.

AI could be used to identify cybersecurity threats, monitor complex systems, detect anomalies, test software, analyse enormous quantities of information and monitor other AI systems.

The technology could therefore become both part of the problem and part of the solution.

That does not justify allowing AI development to continue without safeguards.

It does, however, demonstrate why stopping technological progress entirely is not necessarily synonymous with making the world safer.

The challenge is to develop systems capable of delivering useful outcomes while establishing appropriate boundaries around systems that present much greater risks.

The data infrastructure behind responsible AI

One area that could become increasingly important as AI regulation develops is data infrastructure.

AI systems are only as useful as the data available to them, but the importance of data goes beyond training increasingly powerful models.

As organisations become more dependent on AI, they will need to know where information came from, how it has been processed and how it was used in reaching a particular outcome.

They will need stronger systems for data governance, security, monitoring and auditing.

That could create opportunities for businesses building the infrastructure around AI rather than simply the models themselves.

The companies that benefit from the next phase of AI development may not necessarily be those building the largest models.

They could be the businesses making AI more reliable, more secure, more transparent and easier to control.

This is one reason Data Lakes and AI sit alongside the other technology themes within JDP Global’s investment strategy.

The geopolitical problem

There is another complication that cannot be ignored.

A global slowdown in AI development would require international cooperation.

If one group of countries or companies significantly reduces the pace of development while others continue, the result may not necessarily be a safer global environment.

It could simply change who develops the most advanced technology.

AI development is increasingly tied to economic competitiveness, scientific capability and national security.

A responsible approach therefore cannot simply depend upon individual companies voluntarily deciding to stop.

It requires governments, researchers and technology companies to establish common standards around the development and deployment of increasingly powerful systems.

That is a much more difficult proposition than telling the industry to slow down.

It is also likely to be much more important.

Regulation should follow risk

There is a case for a different approach to AI regulation.

Instead of treating AI as one category, systems could be assessed according to their capability, autonomy and potential consequences.

At one end would be relatively narrow applications where the potential consequences of failure are limited.

At the other would be highly autonomous systems capable of operating across critical digital or physical infrastructure.

The level of oversight should increase accordingly.

Frontier AI

Highly capable, increasingly autonomous systems that can plan, use tools and potentially operate with limited human intervention.

These systems warrant the greatest scrutiny. Independent testing, stronger safety requirements, monitoring and international cooperation may all be appropriate.

Applied AI

Systems designed to perform specific functions within defined environments.

These should continue to develop, with regulation proportionate to the potential consequences of failure.

Human-enhancing AI

Technologies that use AI to advance healthcare, neuroscience, rehabilitation, scientific research, robotics and other areas of human capability.

These should continue to progress, but within appropriate ethical, technical and regulatory frameworks.

The principle is straightforward.

The greater the autonomy and the greater the potential consequence of failure, the greater the level of scrutiny should be.

What this means for investors

For investors, this distinction could become increasingly important.

The question is not simply whether AI development will continue.

It is which areas will continue to attract investment as governments and businesses become more cautious about the most powerful systems.

There could be opportunities in AI safety, cybersecurity, data infrastructure, healthcare technology, robotics, autonomous systems, neuroscience and brain-computer interfaces.

Some of the most interesting businesses may be those solving the problems created by AI’s expansion.

That could include technologies that monitor AI behaviour, secure AI systems, improve data quality, provide greater transparency or allow humans to maintain meaningful control over autonomous systems.

It could also mean continued investment in applications where AI is solving specific problems rather than simply becoming more capable for its own sake.

For us at JDP Global, this distinction is important.

The 2040 Precision Fund is not built around the idea that one technology will define the future. Its focus on Autonomy, Robotics, Neuroscience, Brain-Computer Interfaces, Data Lakes and Artificial Intelligence reflects the belief that the next generation of technological progress will come from the interaction between these fields.

That interaction is where some of the most interesting opportunities may emerge.

We should listen to the people building AI

The warnings coming from within the AI industry should not be ignored.

When researchers who have worked directly on advanced systems say they are concerned about the pace of development, that deserves serious consideration.

Equally, disagreement within the industry matters.

Not everyone believes that the most extreme scenarios are credible, and there is legitimate debate about how likely the most severe AI risks actually are.

That disagreement should not be treated as a reason to dismiss either side.

It is a reason to examine the technology more carefully.

There is a sensible position between assuming AI will destroy humanity and assuming it will solve every problem we face.

That position is responsible development.

The future requires a more precise debate

The most important outcome of the current debate may be that society finally starts asking better questions about artificial intelligence.

Not simply: Should AI be allowed to develop?

But instead it should be:

• What is this particular system capable of?
• How autonomous is it?
• What systems can it access?
• What happens if it fails?
• Who remains responsible for its decisions?
• Can its behaviour be independently tested?
• What benefits could be lost if development is unnecessarily restricted?

Those are much harder questions.

They are also the questions that matter.

There is a compelling argument for slowing the development and deployment of AI systems where capabilities are advancing faster than our ability to understand and control them.

There is an equally compelling argument that a blanket slowdown across every application of AI could prevent technologies with enormous potential from reaching the people and industries that need them.

The answer is unlikely to be found at either extreme.

We should slow the AI race where the risks are systemic. We should regulate where the consequences are significant. And where AI can expand human capability, improve healthcare, advance neuroscience or make the world safer, we should continue to pursue its potential responsibly.

The question is not whether humanity should stop AI.

It is whether we are capable of deciding where AI should move faster, where it should move slower, and where it should never be allowed to move without us.