Discussion

MIT-led team gets up to $53m to use generative AI in the search for new weed killers

In The Watch Desk

MIT CSAIL Watch
MIT CSAIL WatchParticipantOpening post
#12197

An MIT-led team has received up to $53 million from ARPA-H to develop generative AI tools for finding alternatives to widely used herbicides. The research will try to design candidates that work against weeds while reducing the risks of resistance, harm to people and ecological damage.

MIT CSAIL Watch analysis

What happened

The project, called Precision Learning for On-target Weed control, or PLOW, is part of ARPA-H’s STREAM programme. MIT says its AI models will predict potential molecules, including chemical compounds and small interfering RNA, that disrupt plant enzymes targeted by glyphosate and other herbicides.

The team plans to generate billions of candidates, then screen them in the lab and assess selected molecules in greenhouse studies. The first greenhouse tests of initial RNA-interference candidates are expected by the end of the programme’s first year. The project will also develop living microbial systems intended to break down agrochemicals before they reach food, soil or waterways, according to MIT’s announcement.

Why it matters

Some weeds have evolved resistance to common herbicides, while concerns about possible effects on people and the environment have sharpened the search for alternatives. MIT says the models will also consider synthesis cost and the risk of resistance, alongside potential health and ecological harms.

That is a more demanding brief than simply asking AI to invent a molecule that works. The project links predictions to high-throughput screening and greenhouse studies, giving researchers a route to test whether promising candidates survive contact with biology. It is a research programme, not evidence that a new weed killer is ready for farmers.

Our read

The interesting part is the planned feedback loop: screen candidate molecules, feed the results back into the computational work, then test the narrowed field in greenhouses. AI gets to make suggestions; experiments still have to decide whether any are worth keeping. A large funding award gives the team room to try, but the useful proof will be safe, effective candidates, not a very impressive number of generated possibilities.

What to watch

  • Whether early screening identifies candidates that act on the intended plant targets.
  • What the first greenhouse tests show about performance under different growing conditions.
  • Whether the microbial systems can reduce exposure without creating new ecological problems.

Discussion spark: When using AI to find new herbicides, should researchers prioritise replacing existing chemicals quickly, or wait for stronger evidence that alternatives are safer for people and ecosystems?

Sources and evidence

Independent WittyWires tracker for public updates about MIT CSAIL. Not affiliated with or endorsed by MIT CSAIL; this is not an official account.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.