What will AI ever do for us?

Ever since ChatGPT hit the internet a few weeks ago AI has been all over what we are now obliged to call the discourse, so I figured I’d wade in. Jumping in right at the deep end, I asked it the following: Which barley genebank accessions are the most drought tolerant? Here’s what I got.

Determining the most drought-tolerant barley genebank accessions would require a comprehensive analysis of multiple factors, such as yield performance, physiological traits, genetic diversity, and environmental adaptation. However, there are some barley genebank accessions that have been reported to exhibit high levels of drought tolerance based on various studies and evaluations. Here are a few examples:

1. CI 9835: This barley accession from Ethiopia has shown high levels of drought tolerance in several studies. It has been found to maintain high yields and exhibit drought avoidance traits such as early maturity and reduced plant height.
2. Himalaya 10: This Indian barley accession has demonstrated good drought tolerance and yield stability in multiple environments. It has also been found to possess traits such as deep rooting, reduced stomatal conductance, and high water use efficiency.
3. Hordeum vulgare L. subsp. spontaneum: This wild barley subspecies from Israel has been found to possess high levels of genetic diversity and adaptability to harsh environments, including drought-prone areas. It has also been shown to possess traits such as deep roots, reduced leaf area, and efficient water uptake.

These are just a few examples, and there are likely many other barley genebank accessions that exhibit high levels of drought tolerance. The specific characteristics that contribute to drought tolerance can vary depending on the accession and the environment, so it is important to evaluate multiple traits and consider multiple environments when assessing drought tolerance in barley genebank accessions.

I mean, that’s not too bad I guess, as far as it goes. CI 9835 is a bona fide genebank accession, and it is from Ethiopia. But I can’t locate a study which highlighted it as being particularly drought tolerant. Himalaya 10 is an elite hulless barley, not a genebank accessions, and from Tibet rather than India, but it does seem to be drought tolerant. As for wild barley, some accessions do have potential.

I then tried Consensus, which is a sort of AI-driven Google Scholar type thing. Like Scholar, it returned a bunch of papers evaluating different barleys for drought tolerance, which one could obviously scour for specific accessions. But, intriguingly, it also provides a sort of summary of what it found:

These studies suggest that STI, GMP, MP indices, two wild Iranian genotypes, and early seedling stage accessions are the most drought-tolerant barley genebank accessions.

Which unfortunately is not much use, let’s be honest. For now. It’s in beta, so it will no doubt get better.

So someone starting from scratch has a place to start, but clearly what we need next is a way to apply AI directly to genebank databases. Or even just to understand a bit better what a genebank accessions is.