Spare5 founders, from left to right: Patrick O’Donnell, Founder and Chief Architect; Matt Shobe, Founder and Chief Product Officer; Matt Bencke, Founder and CEO; Daryn Nakhuda, Founder and CTO. Photo via Mighty AI.

On the surface, artificial intelligence is sexy. The technology, which has become one of the hottest trends, helps turn products and services human-like, accomplishing computational tasks that were either previously impossible or took far too long.

But underneath the AI lays less flashy, yet equally crucial infrastructure that makes everything tick. It’s why computer vision and natural language model companies like Mighty AI are becoming more and more important — and also why investors like Intel, Accenture, and Google just put their money behind the Seattle startup.

Matt Bencke
Matt Bencke.

Mighty AI, previously known as Spare5, today announced a $14 million round led by Intel Capital, the investing arm of Intel, and also included participation from Accenture and GV, the investing arm of Google. Previous investors like Foundry Group, Madrona Venture Group, and New Enterprise Associates (NEA) also participated; total funding in Spare5 is now $27 million.

Mighty AI originally launched in 2014 as Spare5, debuting an app that enables everyday people to perform short tasks on their smartphone — like photo tagging, price guessing, and sentiment analysis of text —  in exchange for payment.

Then late last year, the startup narrowed its focus on what it calls “Training Data as a Service,” as CEO and co-founder Matt Bencke told GeekWire. The service provides the human insights AI engines need to “think” like people and helps companies across a variety of industries scale their computer vision and natural language models.

Mighty AI, the first spin-out from Madrona Venture Labs, is essentially making artificial intelligence more intelligent by using human labor to re-tune those models. Bencke said his team is solving a “big data problem” at enterprise scale by turning “blank slates into simulations of human knowledge,” as the company describes.

An example of Mighty AI helping identify objects in a photo and training computer vision models. Photo via Mighty AI.

Now Spare5 is having a re-launch of sorts with a new name and a more focused product. Bencke was careful not to call the strategic change a “pivot,” but more doubling down on an opportunity.

“Everybody knows AI is probably the biggest wave in technology innovation right now,” explained Bencke, formerly a senior vice president at Getty Images. “Yet there is a lot of anticipation and frustration about seeing it applied successfully across a number of specific use cases. I think our investors are looking at us as providing picks and shovels to the AI goldrush. No one knows exactly where the gold is, but everyone knows you need picks and shovels.”

Bencke added that if AI engines don’t have good training data, then you “basically have a really expensive data science team.” He said the AI models need to understand how humans think; Mighty AI provides this via its community of “Fives,” the human users who get paid by the company to perform short tasks that improves training data.

“Training data isn’t the sexiest part of AI,” Bencke said. “But it is vital.” 

Personal digital assistants like Amazon’s Alexa or Apple’s Siri, for example, have the benefit of using conversations with users to improve their data sets. But other applications that aren’t consistently interacting with end users don’t have this benefit — that’s where Mighty AI comes in with help.

“A computer is only as smart as the people who train it or teach it,” Bencke said.

One customer is GumGum, a computer vision company that sells advertising and visual intelligence solutions for brands and publishers. GumGum uses Mighty AI to help advertisers understand the context of where their advertisement will be placed, either on websites or social media.

“They need to train their computer vision models to recognize logos in all kinds of different context,” Bencke noted.

Mighty AI, which has also developed its own AI technologies to help improve the quality of the “Fives” community, does have competition in the form of services like Amazon Mechanical Turk and CrowdFlower, but Bencke said those require more management from employees.

“We are agnostic about domains,” Bencke added. “We are just trying very hard to abstract human knowledge and make it available to our customers on demand.”

As a result of the funding, Mighty AI will partner with Intel and Accenture, who will promote and sell the company’s training data services to customers worldwide. Ken Elefant, managing director at Intel Capital, told GeekWire that Intel sees AI as the “next major turning point in human history” with applications like automated driving, robotics, healthcare, and IoT.

Elefant said Mighty AI has the best CEO and team compared to other competitors.

“Unlike other traditional business outsourcing or crowdsourcing sources, Mighty AI can provide high quality, domain specific, on-demand data annotation, gathered via a community of thousands of experts,” he said.

The 30-person company, named a Seattle 10 startup in 2015, will use the fresh funding to grow its team and build out its platform. Bencke called the new investor group a “dream team of strategic partners who are also investing.”

Along with Bencke, the Mighty AI leadership team is impressive. CTO Daryn Nakhuda is a former CTO at TeachStreet and Director of Engineer at Porch; Patrick O’Donnell co-founded Urbanspoon; and Matt Shobe helped start of Feedburner.

Robin Murdoch, a global managing director at Accenture, called Mighty AI “a leader in the critical area of training data.”

“It’s an area that is particularly challenging for our clients, and their technology will help us develop and implement AI solutions that help our clients’ businesses compete more effectively,” he said.

Mighty AI is one of many new AI/machine learning startups in Seattle, which is becoming somewhat of an epicenter for companies developing technologies related to machine learning, artificial intelligence, natural language processing, algorithms, data analytics, and more.

Like what you're reading? Subscribe to GeekWire's free newsletters to catch every headline

Job Listings on GeekWork

Find more jobs on GeekWork. Employers, post a job here.