Chinese universities now publish the largest volume of high-quality research on artificial intelligence in the world. In 1980 they published none, illustrating how far China has come in a relatively short space of time.

That’s according to the 188-slide The State of AI Report, an annual analysis of the sector by two investors, Nathan Benaich, founder of RAAIS and London.AI, a Cambridge PhD now working at Air Street Capital, and angel investor Ian Hogarth, a visiting Professor at UCL.

Among other findings in the report – which looks at skills, research, the political landscape, and industrial applications of AI – the ‘de-democratisation’ of the technology is continuing worldwide as so-called ‘Big Tech’ companies collaborate with elite universities, ignoring research done in lower-tier institutions.

This is affecting academia, as many institutions struggle to find sufficient computing resources to run large studies, while elite universities are increasingly receiving funding from Big Tech companies, such as Facebook, Amazon, Microsoft, Google, and Apple (FAMGA).

Overall, the data and AI ecosystem is maturing fast, with flotations signalling that the sector is moving firmly into the deployment phase.

Chipmakers are benefiting from the surge in AI’s popularity, as both nations and corporations look for supply-chain sovereignty.

Increasingly, AI-first products are being rolled out in critical use cases, say the authors, citing applications such as the UK’s National Grid – which is experiencing market instability post-Brexit – and employee health and safety in the wake of the pandemic.

However, AI is “literally an arms race” too, warns the report, with autonomous weapons now deployed on the battlefield and more testing of them happening regularly.

In some instances robots will be fighting robots, with Russia among those deploying anti-drone autonomous tanks, according to recent news reports.

Governments are amping up their defence rhetoric, explain The State of AI’s authors, and may use this to “smuggle” through greater adoption of AI and autonomous systems.

At the same time, AI safety is recognised as being of paramount importance, yet fewer than 50 researchers are working in this domain full-time at the major AI labs. Nevertheless, the industry is working towards tackling endemic problems, such as bias in historic data sets.

New experiments in AI governance are emerging too, as AI regulation begins in the European Union – which some innovators warn may leave Europe at a competitive disadvantage against China and the US.

Rules proposed in April this year aim to regulate any AI system, including from overseas, that is used within EU member states.

As with GDPR, regulators aim to protect consumers from any harms caused by their personal data being processed in this way, but some in the sector fear this may stifle research and force innovators to process data overseas.

In terms of the technology itself, the report observes that the Transformer architecture is emerging as a general purpose one for machine learning, large language models (LLMs) are becoming “nationalised” (in the sense that each country wants its own one), and JAX is establishing itself as a popular ML framework.