Microsoft announced today that it has purchased Lobe, a boot that allows you to build machine learning models using a simple drag and drop interface. Microsoft plans to use Lobe, which was first released in beta earlier this year, to build on its own efforts to make building AI models easier, but Lobe will currently work as before.
"As part of Microsoft, Lobe will be able to utilize world-class AI research, global infrastructure and decades of building tool experience, writes the team." We plan to continue to develop Lobe as a standalone service that supports open source standards and multiple platforms. "
Lobe was co-founded by Mike Matas, who previously worked on iPhone and iPad, as well as Facebook's paper and instant articles. The other co-founders are Adam Menges and Markus Beissinger.
I In addition to Lobe, Microsoft also recently purchased Bonsai.ai, a deeply reinforcing learning platform and Semantic Machines, a call AI platform. Last year, it bought the Disrupt Battlefield participant Maluuba. It's no secret that machine training talent is hard to pass, so that it is no surprise that all major technology companies acquire as much talent and technology as they can.
"In many ways, however, we only realize that get the full potential AI can provide, "writes Microsoft's EVP and CTO Kevin Scott in today's announcement. "This is largely due to the fact that AI development and the construction of deep learning models are slow and complex processes, even for experienced computer scientists and developers. Until now, many people have had a disadvantage regarding access to AI and we are committed to changing it. "
It is worth noting that Lobe's approach complements Microsoft's existing Azure ML Studio platform, which also offers a drag-and-drop interface for building machine learning models but with one more utilitarian design than the smooth interface that the Lobe team built. Both Lobe and Azure ML Studio aim to make machine training easy to use for all, without having to know in and out of TensorFlow, Keras or PyTorch. These approaches always come with some limitations, but just like low-key tools, they serve a purpose and work well enough for many user cases.