Opinion

April 3, 2023

Europe can't match OpenAI — but Silicon Valley could learn from us

The continent's array of medical AI companies provide a model for ethical growth


Eleanor Warnock

2 min read

Picture: DALL-E/Stable Diffusion

The world is still losing its shit over each successive OpenAI release.

But let’s take a minute and get excited about some other AI news. AI biotech unicorn Owkin said on Friday that it was leading a new €33m consortium — backed by the French government — to improve cancer diagnosis and better tailor treatments to individuals.

While Europe is, undoubtedly, lagging in the generative AI race, it’s performed incredibly well when it comes to innovations in AI-driven medicine.

In 2020, Oxford’s Exscientia brought an AI-discovered drug molecule to human clinical trials for the first time ever (though the study eventually failed). In 2021, London’s DeepMind spun out a unit, Isomorphic Labs, to work on AI drug discovery using protein-folding AI model AlphaFold. And in 2023, European AI drug discovery startups continue to attract funding.

Advertisement

Still, OpenAI’s recent progress makes AI drug development seem frustratingly slow. The latest GPT-4 Technical Report shows how that model could be used in pharma, either to look for existing patents or look for compounds with similar properties to a certain medicine.

No doubt that these kinds of models will be applied more and more as one tool in the arsenal of AI drug discovery companies — despite how terrible they are at science (hello, made-up academic papers).

But wouldn’t it be great if AI medicine developed at the same speed and attracted the same hype that we see around large language models like GPT-4? And wouldn’t it be great if new large language models had the same guardrails — the equivalent of clinical trials — that are required in medicine? Each have something to learn from the other.

Eleanor Warnock

Eleanor Warnock was Sifted’s deputy editor and cohost of Startup Europe — The Sifted Podcast. Find her on X and LinkedIn