Hardik's Notes · AI, simply
How an AI found a new antibiotic, and won a prize
Two MIT scientists won a 2026 prize for using AI to find antibiotics. Here is how the AI did it, in plain words, and what it has not done yet.
What happened?
On 2 November 2026, MIT News said Regina Barzilay and James Collins won the 2026 Pritzker Prize for AI in Science Research Excellence. The University of Chicago and Caltech give it.
The prize is for their work on finding antibiotics (medicines that kill bacteria) with AI.
How did the AI find one?
The team first taught a neural network (a program that learns from examples). The examples were about 2,500 molecules, each marked as killing the bacterium E. coli or not.
- Train the model on about 2,500 molecules, including about 1,700 approved drugs and 800 natural products.
- Ask it to score a library of about 6,000 compounds it had not seen.
- Pick the top molecule, one with a different shape from known antibiotics.
- Test it in the lab on bacteria taken from patients.
The molecule was named halicin, after the computer HAL 9000 in the film 2001: A Space Odyssey. Doctors had earlier studied it as a diabetes drug.
Did it work?
In the lab, halicin killed many drug-resistant bacteria, including Clostridium difficile and Acinetobacter baumannii. The 2020 report says it did not work on one lung germ, Pseudomonas aeruginosa.
30 days
E. coli did not become resistant to halicin in the lab test. Against the older drug ciprofloxacin, resistance began in one to three days.
The same model then screened more than 100 million molecules in three days. A lab testing them one by one could not do that.
In 2023 the team found a second one, abaucin. It works against A. baumannii only. A narrow target spares the helpful bacteria in the gut.
Why should you care?
Bacteria change and learn to survive old medicines. Few new antibiotics have been made in recent decades. A model that ranks molecules before anyone mixes a chemical saves months of lab work.
It is also a plain example of what AI is good at: learning a pattern from examples, then searching a huge space fast.
What’s the catch?
- Halicin is not an approved medicine. The 2020 results are from lab dishes and mice.
- Barzilay says AI can automate parts of drug discovery, not all of it, and that we are still early.
- People still test every pick in a real lab.
Try it
Imagine 1,000 cards, each a molecule. You label 20 as “kills bacteria” or “does not”. Write down what the 20 have in common. Now guess the rest. That is what the model did, with 2,500 examples.
Sources
- MIT News: Regina Barzilay and Jim Collins win Pritzker Prize for AI and Science Research Excellence, MIT News, 2 November 2026
- MIT News: Artificial intelligence yields new antibiotic, MIT News, 20 February 2020
- MIT News: Using AI, scientists find a drug that could combat drug-resistant infections, MIT News, 25 May 2023
Keep reading
Three more from the notebook.
1 November 2026
An AI small enough to fit in a PC's boot sector
A maker has made a tiny neural network that runs before any operating system loads. Here is how small an AI can really be.Read it31 October 2026
Four small AI machines, all built on one board
Arduino named the winners of a contest for its UNO Q board. Four of them use AI to sense water, eggs, rides and light.Read it30 October 2026
A tiny chip that dreams up its own video games
A maker runs two small neural networks on a Raspberry Pi Pico 2 W. One invents each game frame. The other plays it.Read it
Beyond the notebook
This is what we do in schools.
A year-long programme for classes 6 to 12. Creativity, taught as a skill, inside the classroom you already have, led by your teachers.