“The next big thing from the same guys” - Sebastian Moss on AlphaFold

Welcome to another episode of the AI Business podcast, sponsored by data collection and annotation specialist Lionbridge AI.

This week we look at the major scientific breakthrough achieved by DeepMind’s AlphaFold, consider some of the some of the first commercial use cases for neuromorphic computing, and learn the secrets of dataset creation with Aristotelis Kostopoulos, VP for AI product solutions at Lionbridge.

AI Business · AI Business Podcast 13: Data for science, and the science of data

We start with the story about AlphaFold, an artificial intelligence system that can predict the structure that proteins will fold into. Why is this important? Because we’re all made of proteins, and their shape defines their function. Understanding these shapes could make drug development much easier, and set humanity on a path towards producing enzymes that can eat plastic, or absorb carbon.

The only barrier to utopia is the fact that the tech was developed by DeepMind, a part of Google – a for-profit corporation that’s been known to have issues with wielding its power responsibly.

Next, we find out where the carefully labeled voice recordings, handwriting samples and video clips that power AI products and services actually come from.

These datasets are created by armies of employees around the world, in major crowdsourcing projects run by companies like Lionbridge AI – a data prep specialist that emerged out of machine learning operations run by its parent, translation company Lionbridge.

Aristotelis Kostopoulos walks us through the typical data collection process, and shares a fascinating case study.

In November, Lionbridge announced that the AI division was being acquired by digital customer experience company Telus in a deal worth $935 million. The acquisition is expected to close at the end of December.

And finally, we look at GrAI Matter Labs, which has just raised $14m to develop AI inference chips based on the principles of neuromorphic computing – attempting to replicate the physical structure of the human brain in silicon.

The company says neuromorphic chips – originally proposed 40 years ago – are a perfect fit for certain AI applications, able to deliver superior latency and very low power consumption. The first chip is already on the market, under the impressive moniker of GrAI One.

Also in this episode: Zombies! Queen’s Gambit! Detecting COVID-19 by smell!

As always, you can find the people responsible for the circus podcast online: