AI et al

In January, Jennifer Newton asked how long it would be until an AI was listed among the authors on a research paper. Well, it turns out the answer was not very long at all. ChatGPT, the generative text algorithm that has rapidly become a cultural phenomenon, was recently listed as a co-author on several papers. Publishers have responded with a ban on the grounds that a machine cannot meet the criteria of responsibility and accountability required of authors. Yet as Holden Thorp noted in a recent Science editorial, a bigger issue might be that when ChatGPT’s contributions aren’t disclosed, we simply won’t know. Either way, the technology is here to stay.

AI, machine learning and automation are revolutionising chemistry, making it faster and more efficient than ever before. With the ability to analyse large amounts of data and make predictions, these technologies are helping researchers discover new compounds and improve existing ones. However, as with any new technology, there are also concerns about the impact of AI and automation on jobs and the potential for misuse.

We’ve been reporting on these topics for some time and so, not to be outdone, I asked ChatGPT to write the preceding paragraph. It did a perfectly serviceable job. (Things seem to get more fun when you ask for mimicry – the same query requested in the style of Roald Hoffmann returned a few lines of verse in one instance.) Still, I hope that taste of ChatGPT’s style will suffice and that, at least for now, you’d rather not suffer the 600-word version as I have.

It’s clear AI is already changing our lives. Yet it’s harder to see through the hype to know where the changes will happen and right now there are more questions than answers. So we’re devoting a series of feature articles to answering those questions – to understand the reality and capability of these technologies.

There have already been some notable milestones in chemistry. Last year, Deepmind’s AlphaFold effectively solved the problem of protein folding and we’ve recently reported on machine learning tools that have delivered the most accurate solutions yet to the Schrödinger equation. Big pharma companies are also partnering with AI experts not just to accelerate drug discovery with virtual screening and modelling, but also to improve syntheses, optimisation and clinical trials. And as Nessa Carson explains, you can use many of these tools in your own lab today.

The opportunities here are also bringing new players – some heavy hitters with deep pockets – to the periodic table. Alphabet, Meta, IBM and other tech giants see potential for disruption and have been building their chemistry capability. The pace of that growth has cooled recently as economic headwinds force a reevaluation of projections made during the pandemic, but the trend is there.

Perhaps the most intellectually exercising question is what does all this mean for chemists? Yes, ‘the impact on jobs’ as ChatGPT puts it, but also the more existential implications. What if chemistry becomes a black box, for example, understood only by machines and a few specialists? Or, as Philip Ball ponders, what becomes of chemistry’s epistemological essence – what sort of insight and discoveries might machines make that could inform our subject and what chemists do?

A recent study in Nature suggested that there has been a gradual decline in the rate of disruptive research  since the mid-20th century. The authors propose one reason for this might be that science has become so specialised, and its knowledge so vast, that we can no longer span different fields and connect them. Yet a machine could. Perhaps we are on the verge of a new era of disruption.