Advertisement: Bradfield Centre mid
Advertisement: Mobas mid banner
Advertisement Cambridge China Centre
Advertisement: Lynch Wood Park
Barr Ellison Solicitors – commercial property
Advertisement: Wild Knight Vodka
RealVNC mid banner careers
ARM Innovation Hub
Advertisement: EY
RealVNC mid-banner general
Advertisement: TTP
Advertisement: Cambridge Network
Advertisement: Bridge Fibre mid
4 October, 2018 - 12:30 By Tony Quested

PolyAI targets $1.3 trillion market with conversational AI

Three Cambridge University PhDs who studied under five-star Dontrepreneur Steve Young – now full time with Apple – are scaling a transformational technology for conversational AI.

Co-founder Nikola Mrksic tells Business Weekly that the market PolyAI is focusing on is worth $1.3 trillion dollars – bigger than the world’s entire defence industry.

Headcount is scaling and thanks to funding by Amadeus Capital Partners in Cambridge, along with Passion, PolyAI has enough runway for the foreseeable future before having to put a figure on its killer round for global scale-up. Steve Young is advising the business and Dr Hermann Hauser of Amadeus Capital told me: “This is a prospect that I am really excited about.”

Mrksic said that the business had already received approaches from potential acquirers but says PolyAI has no intention of selling out. Mrksic was the first engineer at Cambridge speech technology startup VocalIQ which was bought by Apple for $100m.

He founded PolyAI with fellow Cambridge PhDs Shawn Wen and Eddy Su and they have been joined by an equally brilliant former PhD graduate, Matt Henderson.  

A source close to the business and an influencer in the field of conversational AI, revealed: “They have some very impressive technology for conversational AI. Firstly, they can train a generic query engine on large non-specific dialogue corpora and then map it onto a real application with very little application specific training data. 

“Secondly, they have some very smart software for efficiently collecting dialog model training data using crowd-sourcing. Thirdly they have a dataflow policy engine which allows traditional domain specific slot-based dialogue managers to be integrated with broad coverage Q&A style responses. 

“All of this allows them to build robust voice based application interfaces very quickly.”

Mrksic, who also worked at Apple for two years, said PolyAI had determined to cut through the hype surrounding Artificial Intelligence, focus on a game-changing technology and “pursue it relentlessly.”

PolyAI is already up to 15 staff from a standing start from 12 different countries and has set up a UK HQ in London while retaining its roots in Cambridge. Seven staff have been hired from Cambridge where Mrksic says the firm will remain “embedded.”

Mrksic says the team can build a world-beating AI business by using Cambridge, UK and EU talent and steal a march on any Silicon Valley rivals. “In the past few months we have worked on projects with major clients, including a leading phone company – not Apple – and we have a much different proposition to Apple’s Siri, for example. 

“We intend to use this tech platform to build a really big business with worldwide reach and totally disrupt the customer support market. Part of the vision would be to actually acquire call centres and drive the technology through the entire sector to huge public benefit.

“We have the capability to build the technology language by language – and scale across different languages. We had nothing like this at VocaliQ so we are thinking pretty big.”

PolyAI’s platform for conversational artificial intelligence is at the cutting edge of deep learning and natural language processing. Its models use language-independent neural networks to learn natural behaviour by imitating human conversation.

PolyAI supports virtual assistants including Alexa Skills, Google Actions, Facebook Messenger or Azure Bots. These platforms bring the middleware and PolyAI brings the voice app to life.
 
Whether it’s navigation, flight booking, shopping or playing music, the platform is sector-agnostic, continuously improving as it scales across different applications. 

The latest advances in deep learning also allow the models to scale across languages at almost no extra cost, making it easier to access new  markets.

Mrksic said: “Machine learning is the lifeblood of virtual assistants. Today, only tech giants have the AI arsenal to build good voice-based apps. With PolyAI, building great conversational apps is as simple as designing a website. Whether it’s navigation, flight booking, shopping or playing music, our backend models are the same.

“One assistant, many languages. Our models are polyglots – latest advances in deep learning allow them to scale across languages at almost no extra cost. New markets have never been closer.”

• PHOTOGRAPH: Steve Young of Apple (second left) pictured with his former PhD students who have co-founded PolyAI.

Newsletter Subscription

Stay informed of the latest news and features