- Advertisement -

Ex-Intel executives’ startup aims to tackle spiraling costs of AI

9

Oct 13 (Reuters) – After Intel Corp (INTC.O) purchased Naveen Rao and Hanlin Tang’s startup in 2016, their job was to assist the chip large earn cash off synthetic intelligence, which requires huge computing energy and had change into profitable for the chip business.

But after leaving Intel final yr, the pair on Wednesday revealed a startup known as MosaicML. It aims to assist companies and researchers perform synthetic intelligence extra effectively – in different phrases, through the use of much less computing energy and fewer chips. It has raised $37 million in funding from DCVC, Lux Capital, Future Ventures, Playground Global and different enterprise traders.

MosaicML’s founders stated that synthetic intelligence fashions, which have to be “trained” utilizing big troves of information, have jumped in complexity since 2018. Training a mannequin can now require million of {dollars} value of computing, placing it out of attain for all however giant firms.

“Businesses can’t do this anymore – they can’t have access to the latest methods,” Rao instructed Reuters in an interview. “It’s not good for our field.”

Rao and Tang linked up with Professor Michael Carbin and Jonathan Frankle, a doctoral candidate in his analysis group, respectively, on the Massachusetts Institute of Technology, who had revealed work on how to make machine studying extra environment friendly. What the group discovered is there isn’t any single expertise for lowering AI computing costs, however as a substitute a mosaic – therefore the corporate’s identify – of completely different approaches rising amongst researchers.

MosaicML places these present concepts right into a extra polished kind that’s straight usable by companies. It then additionally gives paid providers, reminiscent of instruments for estimating the tradeoffs between the pace, value and accuracy of utilizing completely different varieties of computing {hardware} for coaching an AI mannequin.

David Kanter, govt director of MLCommons, an AI business group that’s not concerned with MosaicML, stated that predicting and controlling computing costs have change into a key difficulty for companies wanting to faucet into the most recent advances in AI expertise, making work like MosaicML’s necessary to watch.

“Removing hassle, making it faster, making it more affordable, all these things are potentially beneficial to the industry,” Kanter stated.

Reporting by Stephen Nellis in San Francisco; Editing by Steve Orlofsky

- Advertisement -

- Advertisement -