Who needs lab rats when you’ve got AI?

MACHINE LEARNING CAN MAKE ANIMAL TESTING A THING OF THE PAST.

It’s time for the rodents in the lab to retire. Well, nearly. A paper published in the journal Toxicological Sciences claims scientists can use machine learning algorithms to predict toxicity of chemicals to the same degree as animal testing (which, most often than not, results in the death of the poor creatures). To test this theory, the scientists put together a list of nearly 90,000 known harmful chemicals, then used machine learning to label them under categories like ‘dangerous’, ‘corrosive’ and so on. Using these as a base, the algorithm was able to learn to categorise new compounds. While its results matched those found during animal testing 70–80% of the time, it was unable to account for the fact that not all animals react to a particular chemical the same way.