So Cloudera announced its second quarter earnings on Thursday, and they were fine
, but the real news is that it also announced the acquisition of Fast Forward Labs
. The acquisition of a research company would be interesting enough on its own, but this deal is made even more interesting because FFL was founded by Hilary Mason—the former chief scientist at Bitly and a very big name in the world of data science. She’s now Cloudera’s vice president of research.
Mike Olson of Cloudera does a good job explaining FFL and Cloudera’s plans for it in the blog post linked to above:
[Hilary] and her team join us to deepen our expertise in applying machine learning to practical business problems, and to give Cloudera a much clearer view of the future of the field.
Fast Forward Labs’ tag line is “reporting on the recently possible.” This is key to FFL’s value to Cloudera: They continually survey academic and industrial research for new techniques. They take those that are newly available and use them to attack actual business problems, building code and developing expertise in applying those techniques to real-world problems. Their customers, and now ours, benefit early from the latest advances in applied AI.
And so, today, we are launching Cloudera Fast Forward Labs. Hilary and her team, working closely with others here, will continue to investigate and report on the state of the art in machine learning and applied artificial intelligence. We’ll continue to apply those techniques to critical business challenges for large enterprises. We’re adding the research subscription to our portfolio. Each of us has a solid customer base that can benefit from the products we offer together.
While we don’t pre-announce offerings, I expect Cloudera Fast Forward Labs to extend both our product and our services portfolios over time.
I’m very curious to see how this plays out, but the idea seems solid in theory. If Cloudera can adequately fund FFL, it could serve as a smaller-scale, and applied, version of the research labs that drive so much innovation at companies like Google and Microsoft.
FFL was doing targeted research on new technologies, and helping clients figure out how to incorporate them into their businesses. Cloudera builds products in these same spaces. While there’s certainly money to be made selling research and advising, that might pale in comparison the long-term benefits of figuring out what works and what types of products will have broad market appeal.
If the acquisition pays dividends, I’d expect to see research divisions popping up in a lot more companies that compete with Cloudera or want to follow in its footsteps.
From my POV, the other big announcement on Thursday is that Facebook and Microsoft released the Open Neural Network Exchange standard
for porting AI models from one framework to another. This has the potential to be pretty big because, as Facebook’s Joaquin Candela explains in the blog post, different teams often use different frameworks. And it would be great if that work was more transferable between, say, research and production:
When developing learning models, engineers and researchers have many AI frameworks to choose from. At the outset of a project, developers have to choose features and commit to a framework. Many times, the features chosen when experimenting during research and development are different than the features desired for shipping to production. Many organizations are left without a good way to bridge the gap between these operating modes and have resorted to a range of creative workarounds to cope, such as requiring researchers work in the production system or translating models by hand.
We developed ONNX together with Microsoft to bridge this gap and to empower AI developers to choose the framework that fits the current stage of their project and easily switch between frameworks as the project evolves.
Of course, there’s also a business angle here, which would help explain Microsoft’s participation. ONNX currently supports PyTorch, Caffe2 and Microsoft’s Cognitive Toolkit, but noticeably does not yet support Google’s TensorFlow or Amazon favorite MXnet. It’ll be a different story if those companies get involved and brings their frameworks into the fold, but for now Microsoft—which is doubling down on open source on Azure—gets to look like the company enabling open AI, too.
Finally, there was IBM and MIT announcing the new MIT-IBM Watson AI Lab
, which IBM is promising to fund with $240 million over 10 years. Given all the flak IBM has taken over Watson during the past few years—and, really, over the past few months—partnering with MIT could be a very smart idea. Some fresh blood might be good for IBM in terms of thinking outside the Watson box, and the lab’s focus on creating companies to commercialize its work could help give IBM a strategic M&A advantage.
Partnerships like this have worked out pretty well at UC-Berkeley over the years (e.g., with AMPLab and now RISELab
), but those don’t have the tight connection with one specific company. If this works, IBM could come out looking really good. If not, well, IBM’s growing list of critics won’t be surprised.