Kings of the wild frontier

Both the UK government and Open AI published new plans to build frontier AI, and Mr Meta took part in some free speech - what did we learn?

By coincidence this week three different sources gave their view on the future of Artificial Intelligence and how it might be shaped by nation states, corporations and tech titans.

Here the UK government published the “AI Opportunities Action Plan” (1) which was announced by the PM in his earnest voice which means this is a serious thing. In the US, Open AI published their “AI in America” (2) economic blueprint as a welcome note for the attention of the new administration. Finally wherever you get your podcasts we got Mark Zuckerberg on Joe Rogan ruminating on a big year for AI. (3)

It gives a chance to gauge what is pretty much accepted as consensus on the future and what is still up for grabs at this stage. Clearly all three believe this is the single most important technology for humanity going forward, so worth a few minutes to catch up.

Here’s five themes that come out...

It’s great to see some ambition from the UK government

Matt Clifford’s plan is nothing if not bold. It lays out a view of all of the things that need to happen for the UK to become a superpower in AI and ride this technology wave in such a way as to drive the UK economy forward for the next decade and beyond.

There will be many views on whether the 50 recommendations are the right ones, but what should not be ignored is how refreshing it is to see the UK really swing for the fence and create a national mission this bold.

The plan aims not just to keep up but to genuinely develop at the frontier of machine learning and deployment. For Starmer - a stultifyingly cautious politician - to rubber stamp the whole thing in one go is actually genuinely exciting.

States and platforms mostly agree on what is needed

The UK plan wants to build national infrastructure from energy to computing power to data. It aims to build talent pipelines from educational establishments through to hot house AI zones. It wants to leverage government procurement and it wants the private sector to pitch in to help scale the benefits of what it builds.

Representing the platforms, Open AI are the mirror. They want preferential access to state backed infrastructure / energy / data / talent, and it wants the public sector to use its procurement power to backstop any investment that the private sector has to make. It wants support for innovation more than it wants state involvement, but broadly areas like AI Zones come out the same.

Both platforms and government say that they want safety and regulation, but there is more nuance here. Her Majesty’s Government want to assure everyone that this will have guardrails and safety precautions and content creators will get protection, but it’s one of fifty action points.

Open AI and Meta want the White House to standardise rules at a federal level (currently done state by state in US) and then to use it’s power abroad to enforce standardisation across the rest of the western world (“A US-led international coalition for shared safety standards” (4)

TDM and copyright protections are ‘urgent short term’ problems that no-one has cracked yet

Everyone involved knows that the Text and Data Mining / Copyright question has to be resolved, or hundreds of years of copyright protection enabling creative industries to be rewarded for their work gets swept away as an unintended consequence of hungry machines.

Open AI acknowledge the need, but understandably want someone else to set the loosest rules they can so they can get on with it. Zuckerberg is on record as wanting to override copyright completely.

Clifford for once loses all ambition at this stage and simply asks to reform the UK regime “so that it is at least as competitive as the EU” (5), as if they have solved the problem and we just need to catch up. They haven’t, and matching Brussels is a very low bar to set.

Designing and implementing an effective TDM process is becoming an urgent short term problem - the longer it is unresolved the more creative content is hoovered up by bad actors betting that retrospective action will be too difficult.

Infrastructure and Talent Pipelines are ‘urgent long term’ problems

The hardest problems to solve are those that require action right now to have a chance of having made a giant leap forward into the unknown in twenty years time. These kind of ‘urgent long term’ challenges mess with the human mind, which wants either the short term gratification of finishing something or at the very least certainty about what it gets for putting so many resources into plans that contain so much of the unknown.

Clifford to his credit does a good job of being honest about that and making the argument that not just the prize but even the side benefits of trying are worth it;

“Even if AI progress slows we will see large benefits from deploying today’s frontier capabilities and investing in our infrastructure and talent base.. If, however, capabilities continue to advance, having a stake in… advanced AI could be the difference between shaping the future of science, technology and work, or instead seeing these decisions made entirely outside our borders. This is a crucial asymmetric bet - and one the UK can and must make” (6)

The AI Great Game is afoot

A common theme that runs through the reports and discussions is the geopolitical implications for nation states that host and develop frontier AI capability. The UK government wants to design a model for “a modern social market economy” and talks of the risks of falling behind the US and China.

Open AI only gets to the second paragraph of their blueprint before invoking the threat of AI being “shaped by autocrats”. By page 5 they are openly calling out “the Chinese Communist Party’s global influence” and describing “a race America can and must win”(7) with its designated allies. Zuck casually describes a world of “big companies working on it an the Chinese government that steals it from them”(8)

Expect more of this as the new US administration comes in. The Clifford report is timely because it describes the steps the UK must take to have an independent voice in the next generation of technology as a respected ally of the US and a trading partner of China rather than a pawn passed between them. This may be the real bet the UK has to make.

And finally..

Just for a fun finish my highlight of the three hour Markathon on Rogan was during the section on hunting, shortly after Zuck says that he is teaching his daughters to hunt wild pigs for meat. Zuck says his favourite hunting tool is bow and arrow. A few minutes later this happens:

Click: Zuckerbow courtesy of Joe Rogan

It will be many machine hours into the future before AI personas can create awkward male one-upmanship at this level. A Turing test for the future..

(1) https://www.gov.uk/government/publications/ai-opportunities-action-plan/ai-opportunities-action-plan

(2) https://cdn.openai.com/global-affairs/ai-in-america-oai-economic-blueprint-20250113.pdf

(3) Joe Rogan Experience #2255 wherever you get your podcasts or YouTube if you want to watch

(4) https://cdn.openai.com/global-affairs/ai-in-america-oai-economic-blueprint-20250113.pdf

(5) https://www.gov.uk/government/publications/ai-opportunities-action-plan/ai-opportunities-action-plan - Page 14

(6) https://www.gov.uk/government/publications/ai-opportunities-action-plan/ai-opportunities-action-plan - Page 6

(7) https://cdn.openai.com/global-affairs/ai-in-america-oai-economic-blueprint-20250113.pdf

(8) Joe Rogan Experience #2255