View from Washington: the great digital game

The US National Security Commission on Artificial Intelligence (NSCAI) published its final draft report towards the end of January. It makes worrying reading, but nevertheless highlights an opportunity for the UK to align itself with the Biden Administration through technology as it looks to rebuild the post-Brexit ‘special relationship’.

The most worrying aspect is that given the report’s focus on defence, the commissioners align themselves with warnings from senior military leaders that “if current trend lines are not altered, the US military will lose its military technical superiority within the next five years”. There are concerns about Russia’s ability to challenge the US in AI, but greater fears inevitably surround China.

It must be said that many of the report’s main conclusions reiterate warnings and urgings contained in broader AI analyses already undertaken and published during both the Obama and Trump administrations.

There are identified “pockets of excellence”, but the draft says not enough is happening at the whole-of-government level in terms of either interoperability or coordinated research. Far more needs spending on government research: some $63bn (£46bn) by one measure between now and 2026. The US talent pool is thinning while China’s is growing (China now accounts for 29 per cent of the global top tier of AI researchers). The Middle Kingdom is then also proving more aggressive when it comes to setting international standards, raising obvious comparisons with 5G.

We’ve been here before. However, the commission still believes the US can provide the foundations for the West to maintain its edge, technologically and politically, if it moves quickly and decisively. This is where the UK could aim to punch above its weight.

The NSCAI has issued and previewed recommendations several times since beginning work in March 2019. Many have already been adopted, but one group persists into its final draft and these concern internationalism.

The report persuasively argues that while the US can lead, achieving goals in terms of interoperability and research will mean working alongside various alliances.

Harmonising the weapons and defences used across physical or cyber theatres will inevitably require encouraging AI systems adoption and agreement across groups ranging from the Five Eyes (US, UK, Australia, Canada, New Zealand) to NATO.

Beyond that, AI still needs to be sold to the public as a benefit rather than seen as either a Skynet-like risk or a tool that enables techo-authoritarianism through large-scale surveillance and profiling. Doing that will require a transparent, democracy-led response, the report argues.

“The United States must also demonstrate how a democracy should use AI to protect the security of its citizens in ways that uphold liberal democratic values,” it says, arguing that this task should include forging “a coalition of democracies”.

Then, at a technical level but arguably no less important, the report repeatedly raises the challenge of 'TEVV' – test, evaluation, verification and validation – for AI systems. The issue is particularly sensitive when it comes to so-called 'LAWS' (lethal autonomous weapons systems) which not only need to remain under human direction and authorisation, but provide confidence that they will behave as directed when authorised. The US has specifically claimed that its rivals have applied less regard to rigour and fewer ethics when it comes to LAWS’ development and deployment.

It is debatable whether the Trump administration would ever have paid that much heed to the inherently international nature of these recommendations – or even if it had, whether its swaggering approach to diplomacy could have wooed political and military allies in the way the commission thinks necessary. President Biden and his officials are likely to be much more outward-looking and measured in their approach (you do wonder whether his victory in November placed the report on a more global path).

Whatever Whitehall is making of all this right now is being left typically opaque, but the UK does seem to be in a good position to take advantage of the US adopting a more internationalist outlook on AI.

Based on the NSCAI’s recent open meeting, the commission seems to regard Five Eyes as one good place to start the tasks around interoperability, R&D and global leadership in standards and ethics. The clue is in the name. Getting five countries to align on a course is likely to be easier and quicker than for the 30 in NATO or the 27 in the European Union - and time is short. Moreover, Five Eyes is already a largely digital alliance built upon signals intelligence.

Of the Five Eyes members, the UK is ranked as the most advanced member in AI after the US, albeit just ahead of Canada based on the latest Tortoise AI index, and has long been seen as a world-leader in standards-setting. In that second area, it can now operate more independently of the EU on ‘greenfield’ issues (though some cross-referencing would seem sensible).

It looks like a chance for the UK to push against an open door. A handful of other issues remain to take account of and it is also worth looking in a little more detail at how the NSCAI thinks the US should address its global alliances and partnerships in AI. More on this to come.