Gamalon emerged from stealth this week after getting $7.7M from DARPA to speed up artificial intelligence
Gamalon is a Cambridge, MA-based startup that has received $7.7 million from DARPA to create an advanced machine learning and artificial intelligence platform that the company says is more time and computationally efficient than others on the market.
Gamalon uses a new type of machine learning it has developed named Bayesian Program Synthesis, which the company says can accelerate machine learning by more than 100X. The basis of the BPS system is that it uses probability statistics to determine potential connections among the data. By doing so, it drastically reduces the amount of data that it needs to conduct artificial intelligence tasks, the company says.
+MORE AT NETWORK WORLD: 6 Cool new Internet of Things startups | +
The company says it ran a test in which it took Google’s Tensorflow approximately 500 data examples to achieve the same accuracy as Gamalon’s BPS system, which conducted the exercise using three examples.
Sam Charrington, AI-industry analyst and host of the ‘This Week in Machine Learning & AI podcast’ says such comparisons among machine learning platforms should be taken with a grain of salt though because of details of how tests are run. Gamalon says an independent body named Galois conducted the test.
In emerging from stealth this week, Gamalon has released two alpha-version products named Gamalon Structure and Gamalon Match. They’re meant to be used by organizations with large amounts of unstructured data. Structure takes streams of text paragraphs from databases or documents and converts them into what the company calls “clean, structured data rows.” An accompanying product named Gamalon Match can deduplicate and link the data rows created by the Structure product.
The products are available in private alpha versions as APIs that integrate with cloud platforms from Amazon, Microsoft and Google.
“We’re still in the early days of machine learning and AI, and there’s a tremendous amount of innovation happening at the framework and platform levels,” Charrington noted. “This is a good thing, because different techniques are best applied to different types of problems. At the same time, there’s still a lot of work to be done in making these technologies more broadly accessible to the non-Ph.D.’s among us. That’s where the real prize lies.”
Gamalon is led by Ben Vigoa, who has his PhD from MIT and whose previous startup Lyric Semiconductor produced processors that were designed specifically for AI use cases.
In addition to the funding from DAPRA, Gamalon has also received $4.4 million in seed venture capital financing from Felicis Ventures, Boston Seed Capital and Rivas Capital.
