Should AI development be paused or restricted?

Should we hit pause on AI? I'll map the debate — from calls for a moratorium to the case for continued development.

Should AI development be paused or restricted? - There is no consensus, and the debate is usually framed as pause vs. regulation rather than pause vs. unrestricted development.

1. The case for a pause or restriction

In March 2023 the Future of Life Institute published "Pause Giant AI Experiments: An Open Letter" calling on "all AI labs to immediately pause for at least 6 months the training of AI systems more powerful than GPT-4".

Motivations cited in the letter and its coverage include:

•Loss of control and existential risk: AI systems with "human-competitive intelligence can pose profound risks to society and humanity", including risks such as AI-generated propaganda, extreme automation of jobs, human obsolescence, and a society-wide loss of control.

•Out-of-control race: "recent months have seen AI labs 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."

•Need to refocus on safety: The letter asks to refocus research on making systems "more accurate, safe, interpretable, transparent, robust, aligned, trustworthy, and loyal" and recommends more governmental regulation, independent audits before training, tracking highly capable systems and robust public funding for technical safety research.

The letter received more than 30,000 signatures, including academic AI researchers and industry CEOs such as Yoshua Bengio, Stuart Russell, Elon Musk, Steve Wozniak and Yuval Noah Harari.

This view has continued in policy proposals:

•Sen. Bernie Sanders called for an immediate pause on AI development until clear safety rules are established by scientists, a permanent ban on the development of artificial "superintelligence" (AI systems that are more intelligent than humans in all domains), and an international treaty with China.

•The Ban Artificial Superintelligence Act would prohibit the development and deployment of artificial superintelligence, temporarily pause development of certain advanced AI systems, and create a new federal agency responsible for regulating advanced AI development. It would pause advanced AI development until a newly created Department of Artificial Intelligence establishes federal safety rules and a model-review process.

Supporters argue a pause would be "public and verifiable, and include all key actors. If such a pause cannot be enacted quickly, governments should step in and institute a moratorium."

2. The case against a broad pause

Most governments and many researchers did not adopt a pause, and as of July 2024, a pause has not been realized. Instead AI companies have directed "vast investments in infrastructure to train ever-more giant AI systems". Criticisms include:

a) Unworkable and unenforceable

•AI development is spread across different companies, research institutes and countries, making a slowdown difficult to achieve.

•The open-source nature means even if major players agreed, others could continue.

Bill Gates chose not to sign, stating that he does not think "asking one particular group to pause solves the challenges".

The letter was also criticized for diverting attention from more immediate societal risks such as algorithmic biases and for amplifying "some futuristic, dystopian sci-fi scenario" instead of current problems.

b) Geopolitical competition

> Even if one views a ban, pause, or slowdown as desirable, ask yourself what China will do? Beijing called Anthropic CEO Dario Amodei's warning "fearmongering," suggesting that the alarm about AI was a US plot to hinder China's own innovation. If China isn't on board, capable AI models will be available regardless.

House Speaker Mike Johnson said Tuesday that leaders of frontier AI labs should regulate themselves. He opposed a pause on AI development, saying the United States would lose its edge to China. The EU Commission has also said "There is no grace period. There is no pause" on its AI Act.

c) Innovation and safety trade-offs

•Opponents, including the President, argue that even targeted slowdowns could suppress innovation, delay scientific and economic benefits, increase compliance costs, and advantage large [companies].

•Many tech leaders argue regulation should not come at the expense of innovation and that a slowdown could lead to a significant loss of competitive edge.

•AI offers major benefits that would be delayed: Medical Advances: From early disease detection to personalized treatment plans, AI is reshaping healthcare. Scientific Discovery: Machine learning models help scientists analyze vast datasets, leading to breakthroughs in fields like climate science and genomics. Proponents also cite that AIs will help solve complex problems e.g., folding proteins, curing cancer and Alzheimer's disease, reversing global climate change.

•There are vast potential benefits of using AI to address climate and health challenges, but we need to implement policies that encourage sustainable energy use and transparency.

•A pause 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.

•Pausing development would also pause some of the work needed to make AI safer.

Markets also seem to believe risks are manageable - cyber-insurance premiums declined by about 4% globally, the 12th consecutive quarterly decline, despite alarms.

3. The middle ground: Restricted, not paused

Most expert proposals today reject both an indefinite, unfettered race and a blanket moratorium, in favor of governed acceleration or risk-based regulation:

> Rejecting a broad pause does not mean accepting uncontrolled development. The better approach is governed acceleration: continue advancing AI while imposing stronger controls as systems acquire greater capability. High risk systems should face mandatory evaluations, independent testing, incident reporting, identity and access controls, restricted tool permissions, evidence logging, human approval requirements, and tested shutdown and recovery mechanisms.

That model is reflected in:

•The EU AI Act, which proceeds without a pause but imposes tiered obligations by risk level.

•US proposals for export controls on advanced AI chips, kill-switches, and federal safety review before deployment of frontier models.

•Calls that regulation must be put in place before AI expands further into health and other sectors to maximize benefits and minimize harms, and that AI improves safety and efficiency only when guardrails are in place.

In short, the prevailing expert view is not to halt all AI development, but to restrict and regulate the most powerful, frontier systems while allowing continued research, application-layer development, and safety work.

> Answer to the question: A total pause is widely seen as impractical and potentially counterproductive, but targeted restrictions - mandatory safety testing, audits, transparency, liability, and international coordination for superintelligent or high-risk systems - are increasingly viewed as necessary and desirable.

Citations (3)

1wlrh.orgNot yet classified
Tech leaders urge a pause in the 'out-of-control' artificial intelligence race
Backs: “In March 2023 the Future of Life Institute published "Pause Giant AI Experiments: An Open Letter" calling on "all AI labs to immediately pause for at least 6 months the training of AI systems more powerful than GPT-4". Motivations cited in the letter and its coverage include: Loss of control and existential risk: AI…”
2wikipedia.orgEncyclopedias and referencequality 0.83
Pause Giant AI Experiments: An Open Letter
Backs: “Out-of-control race: "recent months have seen AI labs 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.". Need to refocus on safety: The letter asks to refocus research on making systems "more…”
3cepa.orgNonprofits and think tanks
AI: Let’s Not Pause Progress
Backs: “Beijing called Anthropic CEO Dario Amodei's warning "fearmongering," suggesting that the alarm about AI was a US plot to hinder China's own innovation.”