British AI start-up Weave becomes Silicon Valley target

Silicon Valley is once again raiding British expertise in artificial intelligence, with early stage start-up Weave.ai the latest target of takeover talks for US tech groups.

Founded last year in London, Weave.ai has only eight employees and has raised just $200,000 to date, from seed investors including start-up accelerator Techstars. Its team includes four AI engineers, whose goal is to build a system that makes smartphone assistants like Siri or Google Assistant more human-like.

British AI talent has been particularly attractive to the world’s biggest companies, including Google, Microsoft and Apple. In 2014, Google paid $400m for DeepMind, a London start-up that did not have a product for sale at the time, but is now arguably a world leader in AI, with a team of more than 250 academic experts. Last year, Apple acquired UK-based speech technology start-up VocalIQ.

Microsoft paid $250m in February this year for Swiftkey, the London-based maker of a predictive keyboard. In June, Twitter paid $150m to acquire London-based Magic Pony, a 14-person start-up that had only raised seed funding.

Weave.ai is in discussions with “multiple parties” on the US west coast to close a deal imminently, according to sources close to the talks.

Last year, Weave, which is still in undercover “stealth” mode, demonstrated a beta service that could analyse messages on a smartphone and provide contextual assistance. For instance, a tweet about the Age of Ultron film brought up links to buy tickets at a cinema chain and information about the film — very similar to Google’s Now on Tap feature.

Chief executive and co-founder Rodolfo Rosini wants it to move on to addressing a problem he calls “explainable artificial intelligence”, which organisations like Darpa, the US Department of Defense’s research arm, are also working to solve.

Currently, AI algorithms are limited by their inability to explain their decisions and actions to human users. Weave.ai wants to introduce context and transparency into the process.

“Today there’s no way to trust the algorithm,” says Mr Rosini. “If a computer tells a doctor you have cancer, the doctor doesn’t know why. You can’t ask it why it thinks that, it could be a false positive.”

“Our system would work for anything where the user has to communicate with an AI,” he says.

Prospective buyers could be from among a number of Silicon Valley companies that are developing voice-based AI assistants. The best known is Apple’s Siri, with Microsoft’s version known as Cortana and Google represented by Google Assistant. Amazon’s Alexa assistant works through its Echo speaker, a product the company plans to launch in Europe this week.