UK’s plans for “first global summit” on AI safety draw criticism


On Wednesday, UK Prime Minister Rishi Sunak announced that the nation will host “the primary main international summit on AI security” this autumn. It hopes to carry collectively “key international locations, main tech corporations, and researchers” to judge and monitor dangers from synthetic intelligence.

Over the previous yr, the perceived excessive fee of tech progress in machine studying has fostered issues about satisfactory authorities regulation. These worries had been just lately amplified by some AI specialists likening the potential threats posed by AI to these of pandemics or nuclear weapons. “AI” has additionally been a particularly buzzy time period in enterprise just lately. Alongside these traces, the UK authorities needs to step in and take a management function within the discipline.

“Breakthroughs from AI proceed to enhance our lives—from enabling paralysed folks to stroll to discovering superbug-killing antibiotics,” the UK authorities mentioned in a press release. “However the growth of AI is very fast-paced and this tempo of change requires agile management. That’s the reason the UK is taking motion, as a result of we now have a worldwide responsibility to make sure this know-how is developed and adopted safely and responsibly.”

The considerably obscure information launch didn’t specify a date, organizational format, or venue for the summit.

Within the lead-up to the potential summit, Sunak just lately performed talks with numerous trade leaders, together with CEOs from AI labs like OpenAI, DeepMind, and Anthropic. The UK says that the upcoming occasion is ready to construct on current discussions about AI security held on the G7, OECD, and World Partnership on AI. Moreover, the primary briefing of the UN Safety Council on the influence of AI on worldwide peace and safety is deliberate for July.

“Nobody nation can do that alone,” Sunak mentioned within the information launch. “That is going to take a worldwide effort. However with our huge experience and dedication to an open, democratic worldwide system, the UK will stand along with our allies to cleared the path.”

Who shall be invited?

In its summit announcement, the UK authorities didn’t formally reveal an inventory of invitees.  However the press launch enthusiastically discusses the businesses OpenAI, DeepMind, Anthropic, Palantir, Microsoft, and School as examples of machine learning-adjacent companies which have places of work within the UK. It additionally quotes executives from these corporations, together with Alexander Karp, the co-founder and CEO of Palantir:

“We’re proud to increase our partnership with the UK, the place we make use of practically 1 / 4 of our international workforce. London is a magnet for the most effective software program engineering expertise on the earth, and it’s the pure alternative because the hub for our European efforts to develop the simplest and moral synthetic intelligence software program options out there.”

This record of potential summit individuals has already drawn criticism from some quarters. Rachel Coldicutt, who runs the London-based fairness and social justice analysis consultancy Careful Industries, tweeted, “This press launch for the UK AI Security Summit options DeepMind, Anthropic, Palantir, Microsoft and School and never a single voice from civil society or academia, and nobody with lived expertise of algorithmic harms.”

Of the businesses talked about, Palantir, specifically, has been the subject of controversy, with its shut ties to the army and protection sectors raising questions in regards to the potential misuse of AI. The corporate’s know-how has been deployed in law enforcement and surveillance, resulting in privateness and civil liberties issues.

Dr. Sasha Luccioni of Hugging Face expressed comparable issues on Twitter: “The ‘first main international summit on AI security’ that includes corporations like Palantir (army/protection) and DeepMind (X-risk/AGI)—very promising for addressing the true and present dangers of AI (like its use within the army).” Her sarcasm underlines the perceived disconnect between the summit’s proclaimed security aims and the actions of among the companies touted within the announcement.

After a number of warning letters about AI danger this yr from large names in tech, “AI security” has turn out to be a sensitive subject, with a lot debate over whether or not machine studying techniques characterize an existential danger to humanity. Concurrently, AI ethics proponents like Luccioni, whom we now have interviewed previously on this subject, really feel that not sufficient consideration is being paid to dangerous functions of AI that exist already as we speak.

Whereas the UK authorities emphasizes common speaking factors about AI danger in its information launch, these critiques from AI ethics proponents spotlight a rising demand for extra complete and numerous views at occasions that search to craft authorities regulation over AI. The worldwide AI group will little question be carefully watching to see whether or not the UK-hosted occasion can ship an inclusive and productive dialogue in regards to the dangers of AI that reaches past the standard photo-ops with the brand new titans of trade.

Source link