IBM acquires AlchemyAPI to bring deep learning to Watson

Technology heavyweight IBM today is announcing that it has acquired AlchemyAPI, a startup with a service for making inferences on images and text using a form of trendy artificial intelligence called deep learning. The idea is to bolster the IQ of IBM’s Watson “cognitive” computing system with AlchemyAPI’s tools.

Terms of the deal weren’t disclosed.

Deep learning — which involves training systems called artificial neural networks on heaps of data like pictures or speech and then throwing new data at the systems to receive predictions — is one of several types of computing that AlchemyAPI can perform, founder and chief executive Elliot Turner told VentureBeat in an interview. But it’s the type that likely looks most fascinating to IBM as a way to distinguish itself from other cloud service providers.

Facebook, Google, and Twitter have all made acquisitions in the realm of deep learning. Giving startups access to deep learning could make IBM appear more contemporary than other enterprise software vendors.

Meanwhile, today’s deal could lead to greater merger merger and acquisition activity around deep learning startups.

“I think we’re going to see more consolidation, quite honestly,” Turner said.

Turner sounded excited about the opportunity to join up with Watson, which can already perform a range of functions inside of developers’ applications.

“You’re going to have a whole variety of primitives to, for example, understand language, speech, conversations, video processing, and speech synthesis, allowing folks to combine these things like Lego blocks to build new and interesting applications that automatically generate combinatorial effects,” he said.

AlchemyAPI’s capabilities in the world of natural language processing include entity extraction, sentiment analysis, and language extraction. When it comes to computer vision, it can tag images and detect faces. The startup charges based on how many transactions are conducted through its application programming interface (API).

“We have never had to do an [on-premises] deal ever in history of the company,” Turner said. He figures many customers will prefer to tap the AlchemyAPI services as cloud services in the future, under IBM’s ownership. But technically, it’s possible run the AlchemyAPI technology from inside of a company data center.

“We want to service as many use cases and as many different business functions as possible,” Turner said.

And Turner isn’t opposed to AlchemyAPI being lumped in with other “cognitive” technologies IBM is selling. After all, AlchemyAPI has been working on several tools that have not been disclosed to the public yet. They could align well with the label, perhaps.

“I think that whole cognitive technology is going to make a lot more sense as we kind of walk through the roadmap in the coming months and quarters,” Turner said.

AlchemyAPI began offering computer vision services based on deep learning only last year, but that’s still good timing, as the level of interest in deep learning remains high.

AlchemyAPI started in Denver in 2005; it employs around 20 people. The startup has taken on less than $2 million in external funding, coming from Access Venture Partners and Turner himself.

The AlchemyAPI customer list includes CrisisNet, Pocket, Spiderbook, Tabelog, and Waggener Edstrom.