Should AI development be paused or restricted?

Should AI be paused? I'll survey the current debate to map arguments for and against restrictions.

There is no consensus on whether AI development should be paused or restricted - the debate reflects different assessments of risks, benefits, and feasibility. The positions can be grouped into three camps: for a pause/restriction, against a pause, and for targeted regulation rather than a blanket pause.

1. What "pause" proposals have called for

The most prominent call for a pause was the open letter published on March 28, 2023 by the Future of Life Institute:

> "We call on all AI labs to immediately pause for at least 6 months the training of AI systems more powerful than GPT-4"

The letter argued that "human-competitive intelligence" systems could pose serious threats to society and humanity, citing risks such as "AI-generated propaganda, extreme automation of jobs, human obsolescence, and a society-wide loss of control".

It was signed by more than 30,000 people including Elon Musk, Steve Wozniak, Yoshua Bengio and Stuart Russell, and stated that AI labs are locked in an "out-of-control race to develop and deploy ever more powerful digital minds that no one – not even their creators – can understand, predict, or reliably control".

The letter called for:

More recently, Anthropic CEO Dario Amodei has argued development should be decelerated so that adequate guardrails can be deployed, saying that if a product is not under adequate control or may not be safe, "development or release should be paused until the problem is addressed".

2. Arguments FOR pausing or restricting

a) Existential and catastrophic risk: Some researchers argue the scale of danger "really could cripple civilization or cause extinction", pointing to risks ranging from "automated assembly of bioweapons to the unpredictable disruption of the social and environmental systems we depend on". Similar concerns include AI systems "with human-competitive intelligence can pose profound risks to society and humanity — from flooding the internet with disinformation and automating away jobs to more catastrophic future risks".

Additional sources of risk cited include "losing control of autonomous weapons systems, deliberate misuse of powerful AI to destabilize society, accidents in powerful AI systems" and a race toward AGI triggering conflict.

b) Governance lag: The argument is that capabilities are advancing faster than society's ability to understand and govern them, and that "level of planning and management of AI is insufficient".

c) Need to build safety: Supporters say a pause would allow making "today's powerful, state-of-the-art systems more accurate, safe, interpretable, transparent, robust, aligned, trustworthy, and loyal" and to "dramatically accelerate" AI governance.

d) Public support: Polling suggests public concern is significant - a Wall Street Journal survey of 1,500 registered voters found that "63% supported pausing AI development and construction of supporting data centers, while 57% favored immediate government safeguards", with concerns including "potential job disruption, economic inequality, concentration of power" and safety of increasingly autonomous systems.

The strongest "doom" position argues AI is so dangerous that "AI development should be stopped (or 'paused' indefinitely)".

3. Arguments AGAINST pausing or restricting

a) Unworkable globally and counterproductive for defense: Critics argue "an AI slowdown will prove unworkable". Even if the US paused, "If China isn’t on board, capable AI models will be available regardless of what the US does". Beijing called Amodei's warning "fearmongering," suggesting it was a US plot to hinder China's innovation, and a recent US-China summit produced "No deal to slow or pause AI progress".

Pausing "risks accentuating asymmetry between able defenders and rogue actors, whether they be nation states, criminals, terrorists, or simply misaligned models. AI defense can reduce or offset the risk of AI offense. If we delay access to the most capable models, defenders' hands would be tied". Proponents of this view note Chinese and American models help harden software by identifying vulnerabilities.

b) Loss of benefits: AI "may help reduce risks to large-scale threats from asteroids, volcanos, solar storms, floods, and wildfires by improving prediction and detection" and "The economic growth dividend from AI could also be used to manage and help society adapt". Critics of restriction also argue that "Potential benefits: Self-replicating AI could have significant benefits, such as solving complex problems in areas like medicine, energy" and that restricting "stifles innovation and prevent[s] the development of potentially beneficial AI technologies".

AI promises "massive consumer benefit, estimated at $97 billion annually in the US alone".

c) Innovation, competition and concentration: Opponents argue "regulation stifles innovation and reinforces monopolies", that stringent regulation "could stifle innovation, slow economic growth, and cause the U.S. to lose its competitive edge", and that "Regulation may slow down development and limit the benefits to society, be too rigid, and hinder competition".

Some, like Yann LeCun, argue restricting development to a few elites "increases the real danger—monopolistic control and misuse" and advocate for "open access to AI research and global collaboration".

Others argue we should "regulate applications rather than the technology itself" and that legislation "moves too slowly to effectively govern rapidly evolving AI technologies".

High costs and risk of a black market are also cited: regulating AI "could lead to a ‘black market’ in AI, with criminals using it for nefarious purposes".

d) Safety research needs progress: Some argue "Pausing AI development is not a realistic solution" because it "could not stop development in China or the spread of open-source models" and "slowing AI research could slow the safety research needed to make AI systems more secure". Similarly, "Regulation risks boomeranging. It could slow researchers’ attempts to identify and address safety challenges".

4. The middle path: Restriction vs. regulation

Most current policy does not propose a blanket pause, but risk-based regulation.

The debate is often not "pause vs. no pause" but what kind of restriction:

A balanced perspective often stated is: “Pausing AI development ensures ethical safeguards are embedded before irreversible impacts occur” versus “A pause on AI advancement risks stifling innovation and losing competitive advantage,” with a third view that “Advancing AI cautiously, with iterative ethical assessments, could strike the right balance”.

Conclusion

Whether to pause depends on what risk model you accept:

In practice, no six-month pause ever materialized after the 2023 letter — "Six months later, companies had not slowed development" — and the trajectory has been toward differentiated, risk-based regulation rather than a wholesale halt.