Jan 28th 2014, 12:27, by Steve O'Hear
Most headlines surrounding Google’s acquisition of DeepMind focused on the $500 million+ price tag, what its new owner may or may not be planning to do with the AI talent and tech it’s buying, and the fact that rival Facebook appears to have, at one time or another, been in the bidding. However, another interesting footnote is that DeepMind is British.
The UK is well-known for being a producer of startups that punch above their weight in the creative industries, such as fashion and music, as well as in fintech, but might AI or ‘deep learning’ be another area of strength?
The challenge with assessing this thesis based on DeepMind’s exit alone is that little is known about the company. It rather vaguely describes itself as a “cutting edge artificial intelligence company” that combines “the best techniques from machine learning and systems neuroscience to build powerful general-purpose learning algorithms.” Based on this tech it cites commercial applications in simulations, e-commerce and games.
What’s also interesting is that there are other noteworthy ‘Made In Britain’ AI startups
I’m also told that DeepMind’s co-founder, Demis Hassabis, who I understand may have made as much as ~$100 million out of the exit, is extremely charismatic in person — one industry insider described the former chess child prodigy as having “Reality Distortion”-styled abilities — which is always an asset on the part of the seller. And certainly a $500 million-plus exit to Google for a three-year-old company undeniably makes for a great British success story, likely making it one of the top 10 UK tech exits in the last 10 years.
However, what’s also interesting is that there are other noteworthy ‘Made In Britain’ AI startups, on paper at least, many of which are chronically underfunded compared to DeepMind. Two examples that piqued my interest are Storybricks, the gaming AI startup founded by serial entrepreneur Rodolfo Rosini, and Existor, maker of a number of AI-assisted bots and smartphone apps that create natural language two-way communication between a human and computer.
In contrast to DeepMind’s $50 million, Storybricks has taken a very modest $800,000 in funding from, mostly, European and U.S.-based angel investors. Meanwhile, Existor is thought to have taken no funding at all. That’s intriguing considering that Facebook may have been outbid by Google in an attempt to acqu-hire DeepMind.
Largely in stealth since it was founded in 2010, Storybricks started out life with the super-ambitious mission to create a new browser-based MMO that would let users turn stories into games. It’s since pivoted to build an engine to communicate with artificial intelligence via understanding the context and use of symbols, which can be applied to things like giving characters in games emotions, and licensing this engine to third parties. Or, in the London/San Diego-based startup’s own words, it claims to be the “pioneering market leader of Emergent AI.”
In 2013 it caused the gaming/AI industry to sit up and take notice, announcing that it had partnered with Sony Online to help power the next version of EverQuest. However, although it’s applying its tech to the online gaming space, it sees itself very much as an AI company first. Co-founder Stéphane Bura has a degree in Computer Science/AI, while members of the team have written two of the leading books on gaming AI. Storybricks has also filed three patents, and I’m hearing it’s recently poached an impressive CTO that will surprise some in the industry when it is officially announced.
Aside from being British, what both Storybricks and Existor have in common is that their AI technology is about understanding context
Its tech is being applied to business, education and entertainment, but it’s not hard to imagine lots of use-cases for an interface that feels conversational. Think Apple’s Siri, for example, which in contrast took five years, $200 million+ in investment, 300 researchers and the help of 25 universities to pull off. One of Existor’s apps, Cleverbot, can be seen in action here.
Existor also demands particular attention because its co-founder and CTO, Carpenter, has won the Loebner prize (like a Nobel prize for AI) twice, an achievement that can be matched only by a team funded by DARPA.
Aside from being British, what both Storybricks and Existor have in common is that their AI technology is about understanding context — which is also where companies like Google and Facebook’s interest in artificial intelligence comes into focus. As more and more data points are created, AI is needed to figure out that a person wants one piece of information or action over another.
“The idea is that your software can learn from context, big data and usage. Basically it’s about creating more complex UIs that are simpler to use and ‘know’ what you need,” Storybricks’s Rosini says. He notes that those data points were originally in the browser, then mobile platforms such as Android, and now Internet of Things (IoT) devices, such as Google-acquired Nest, the smart thermostat.
It’s not difficult to see how Facebook could also benefit from AI-driven pattern matching and behavioural targeting. Right now advertisers need to specify what type of customers they want to target on the social network. With enough AI, says Rosini, it would be the opposite. Facebook could preempt the match between users and a product “just by analyzing all the data points and behaviours.”
In researching this post I also talked to Ian Hogarth, co-founder and CEO of gig search engine and recommendation platform Songkick, who has a background in machine learning. Hogarth also knows the London startup scene well.
“My take on [the DeepMind acquisition] is that it showcases the depth of talent in the UK around Machine Learning, we have some of the world’s leading research teams doing amazing stuff in computer vision and deep learning,” he says. As another example, Hogarth points to one of Google’s biggest AI hires to date, Geoff Hinton, who has a background at Cambridge University.
More broadly, on the topic of Google’s interest AI and the DeepMind acquisition, he had this to say:
“I think that machine intelligence will gradually start to complement many tasks that have historically been performed by humans. The near term ones are extensions of robotics and navigation (e.g. self-driving cars) but long-term I think more and more computers and humans will pair up to solve tasks. For my Masters I worked on a system that could help process breast cancer biopsy scans faster – the idea was that a machine could pair with a pathologist and help to do some of the more mechanical aspects of diagnosis, thus saving the human time in processing the more edge cases. I think Google is probably betting on that trend as a whole, and in the meantime there are plenty of ways to use machine learning to optimise ads, so unlike many companies they have a business model that justifies these kinds of acquisitions.”
The same, of course, could well be said of Facebook. And, perhaps, the UK’s AI talent pool could once again provide part of the answer.
Image credit: Dimitry B.
No comments:
Post a Comment