AI in 2020
5 events, oldest first. Each links to a page with details and sources.
- OpenAI publishes 'Scaling Laws for Neural Language Models' ★★★★★
Kaplan et al. showed language-model loss falls as a smooth power law in parameters, data and compute over many orders of magnitude, giving a quantitative case for building ever-larger models.
- Deep learning discovers halicin, a structurally new broad-spectrum antibiotic ★★★★
MIT's Collins and Barzilay labs (Cell, Feb 2020) trained a message-passing neural network on ~2,300 molecules. It identified halicin, a diabetes drug candidate, as a potent antibiotic that killed M. tuberculosis, carbapenem-resistant Enterobacteriaceae and pan-resistant A. baumannii, and cleared…
- GPT-3 (175B) shows in-context few-shot learning ★★★★★
OpenAI's 175-billion-parameter GPT-3 could perform new tasks from a few examples in its prompt, without fine-tuning; it was offered via the OpenAI API from June 2020.
- Liverpool's mobile robot chemist runs 688 experiments in 8 days and finds a 6× better photocatalyst ★★★
Andrew Cooper's group (Nature, July 2020) built a mobile robot that moved around a standard lab and ran 688 experiments over 8 days in a 10-variable space, guided by batched Bayesian optimisation. It found photocatalyst formulations about 6× more active for hydrogen production from water than the…
- AlphaFold 2 solves protein structure prediction at CASP14 ★★★★★
AlphaFold 2 achieved a median GDT score of 92.4 at CASP14, accuracy competitive with experimental methods, widely seen as solving the 50-year-old protein folding problem for single chains.