Computing

Why Aussie artificial intelligence start-up Strong Compute is getting backing from self-driving car pioneers

Why Aussie artificial intelligence start-up Strong Compute is getting backing from self-driving car pioneers
Written by admin

The round also attracted many angel investors who’ve worked in AI, including founders and engineers from Cruise, Waymo, Wayve, Lyft, Uber, OpenAI, Nueralink, SpaceX, NASA and Virgin Galactic.

‘Drag racing neural networks’

Mr Sand said many of the angels who participated in the capital raise had “felt the pain first-hand” of the problem Strong Compute is trying to solve.

Daniel Kan, the co-founder and chief product officer of self-driving car business Cruise participated in the round through the VC fund Rebel.

“When developing self-driving cars, compute for AI is essential and training times are an enormous pain point. I couldn’t be more excited by what Strong Compute is developing,” Mr Kan said.

It typically takes 30 to 60 hours to train a neural network – which are used to teach cars how to accurately identify objects such as pedestrians – creating a bottleneck for AI developers who need to wait for results.

Strong Compute has drastically cut that time down by making hardware and software optimizations that speed up neural network development. For example, the chief executive of MTailor, which makes custom clothes, said the training time for its code dropped from 30 hours to less than five minutes.

“That basically means these very highly paid engineers are able to actually make progress on their code much faster,” Mr Sand said.

He said the company is “drag racing neural networks” to help its customers reach their AI breakthroughs sooner.

“A major milestone for a medical imaging company, such as being able to detect a particular cancer with a sufficient level of accuracy, really takes the company to the next level,” Mr Sand said.

“Our engineers are basically sitting there, revving their engines, figuring out how they can hack the algorithm so that it gives the same answer as it did before, but a lot faster.”

About the author

admin

Leave a Comment