Last Week in AI #9

Algorithm Trained to Detect Unsafe Foods

Researchers were able to train a machine learning algorithm that detects foods that could be recalled by the FDA by analyzing Amazon reviews for the same products. The model was able to predict which foods would be recalled 75% of the time.

Read more at: Scientific American

New Benchmark for AI Language Understanding is Launched

Snippet: Having already been surpassed several times over this previous year, Google’s BERT benchmark model has now been replaced by SuperGLUE, a new set of increasingly more challenging tasks for computer language understanding meant to further advance the field.

Read more at: VentureBeat

Machine Learning Helps Create Better Batteries

HolyGrail, a new startup launching from the prestigious Y-Combinator accelerator, is using machine learning techniques to build better batteries by analyzing how different cathodes and anodes will work together. Even though it may sound simple in principle, its founders claim the computational power to achieve this didn’t exist before.

Read more at: TechCrunch