AI Weekly | Nov. 9, 2019

Researchers tout AI that can predict 25 video frames into the future

In a preprint paper, researchers propose an AI model that can predict up to 25 video frames into the future given only two to five starting frames.

Google's Teachable Machine 2.0 elucidates the basics of machine learning

Google launched a revamped Teachable Machine tool that lets users train and export machine learning models to any website or app.

Amazon's AI rewrites 'millions' of Alexa user commands to reduce defects by 30%

Amazon says it has deployed in production a system that improves the accuracy of Alexa responses, chiefly by rewriting utterances.

Facebook: Our video selfie test is for spotting bots, doesn't use facial recognition

Contrary to preliminary reports, Facebook isn’t using facial recognition to verify the identity of users, a spokesperson told VentureBeat.

MLPerf releases machine learning inference performance results

The MLPerf consortium today released an analysis of machine learning inference for datacenters and the edge to standardize industry benchmarks.

Applause targets AI bias by sourcing training data at scale

Software-testing company Applause wants to reinvent AI testing with a service that detects AI bias by crowdsourcing larger training data sets.

Machine learning shows no difference in angina symptoms between men and women

The symptoms of angina — the pain that occurs in coronary artery disease — do not differ substantially between men and women, according to the results of an unusual new clinical trial led by MIT researchers.

Remember that scary AI text-generator that was too dangerous to release? It's out now

OpenAI today published the final model in its staged release for GPT-2, the spooky text generator the AI community’s been talking about all year.