World’s most funded AI company, Sentient Technologies, launches online shopping product

Sentient AI, the world’s most-funded artificial intelligence company, wants to solve the world’s most complex problems, including financial trading, pancreatic cancer and e-commerce.

Chief executive and co-founder Antoine Blondeau, who helped invent the technology behind Apple’s Siri, announced it will launch its first retail partnership with US shoe retailer Shoes.com.

Sentient has built an intelligent system for online shopping that does not just provide a list of recommendations, like Amazon or Asos do, but lets you have a real-time dialogue in the virtual world, just as you might do with a particularly good salesperson in a bricks-and-mortar store.

In a video of the system in action, Mr Blondeau showed how the virtual storefront adapts its products as you browse through a visual catalogue, figuring out your intent in real time just like a shop assistant.

“It’s finding what you want, without having to Google it,” Mr Blondeau, 46, said.

Mr Blondeau said that Sentient's algorithms are already working on an unprecedented scale by using the power of idle computers around the world, including two million CPU cores and 5000 GPU cards distributed across 4,000 sites worldwide.

“No one had scaled AI to the level we were trying to scale it to […] there is no theoretical limit,” he said at the Wired conference in London in October.

The company, which was founded in 2008, came out of stealth mode in November 2014 and has raised more than $143m from investors such as Hong Kong-based Horizon Ventures run by business magnate Li Ka-Shing, and Tata Communications.

The products that Sentient has created are essentially machine learning algorithms on steroids. A regular machine learning system, something like the Siri assistant in your iPhone, learns as it is being used, and finds patterns so it can adapt itself accordingly in future.

Sentient’s AI doesn’t just crunch data to find patterns - it can make real-world decisions. These algorithms mimic the way biological life evolved on Earth - thriving by killing off the weakest members of a species, until only the strong survive and reproduce, for millions of generations.

Sentient’s first practical application for its AI system was financial trading, where it created an AI trader that is made up of trillions of software agents, or ‘species’ which compete against each other.

So far algorithmic trading has always needed a human analyst who comes up with a model using his or her own thesis. These systems use machine learning to optimise their portfolio.

“In our case, the whole strategy is self-evolved by the system. We are not asking it to compare assets, our AI picks its own assets, builds its own strategy; there are no humans in the system,” Mr Blondeau told the Telegraph. “It defines its own entry and exits, chooses its orders, runs its portfolio location. These are all very complex decisions.”

An early example of this type of system is the programmatic advertising exchange that Google uses to trade ads on its platforms. There is almost no human intervention in this industry anymore, with algorithms deciding within milliseconds whom to buy and sell to.

But what the Sentient AI trader does is far more complex, Mr Blondeau said. “We can make trades for hours, days, even months, so you need to understand much more of the world, crunching far more data.”

Mr Blondeau won’t disclose the algorithm’s trading performance, but said the company is considering spinning off a prop trading or an asset management business based on its success.

Financial trading and e-commerce are Sentient’s first two commercial products, but Mr Blondeau is also eyeing healthcare and biotech as possible future directions.

Sentient has already invented an AI nurse that was tested out with Massachusetts Institute of Technology to detect sepsis - a bacterial infection which is the number 1 killer in the intensive care unit. It kills 37,000 people in the UK every year at a rate greater than bowel cancer and breast cancer combined.

The artificially intelligent ‘nurse’ was built by continuously collecting blood pressure data from 6000 patients to spot patterns; it was then able to predict sepsis infection 30 minutes ahead of time, to an accuracy of 91 pc. “This is very meaningful, because it actually saves lives. And all that data was just not being used,” Mr Blondeau said.

The company is also working with Oxford University to predict the occurrence of pancreatic cancer in certain patients.

Will algorithms like Sentient’s ultimately replace humans completely? “Yes, we are trying to minimise the amount of human intervention. Our thesis is there will be more and more applications with limited human value,” Mr Blondeau said.

“But this doesn’t mean it should not be supervised. As long as humans can see what the system is doing, you can trace its steps and interpret the outcomes.”