stephen.news

hypertext, words and more

Lobe

  • It’s been an exciting year for Microsoft. Microsoft continues to make bets on thrilling components of software development. This was a huge missed opportunity for giants like Apple, Google and even Adobe. This morning, Microsoft announced they have acquired Lobe.

    If you don’t know what Lobe is, here’s a primer: Lobe, is basically software for building, training, and exporting custom deep learning models. The interface is very Quartz Composer-esque. I previously wrote about Lobe here.

    From Lobe.ai

    To be honest, I’m a little shocked Apple didn’t acquire Lobe first. Considering the boundless software opportunities and Lobe’s earnest genesis in Quartz Composer, it’s strange Apple didn’t beat Microsoft to the punch. Maybe there were meetings between Apple and Lobe, who knows. What I do know, is that Microsoft knows that ML tools are going to be huge.

    From Lobe’s announcement email:

    As part of Microsoft, Lobe will be able to leverage world-class AI research, a global infrastructure, and decades of experience building developer tools. This will allow us to rapidly scale Lobe to its full potential, making it more intelligent and available to more people. 

    We plan to continue building Lobe as a standalone service, supporting open source standards and multiple platforms. Together we are committed to making deep learning simple, understandable, and accessible to everyone. We deeply appreciate your support and look forward to sharing more with you soon! 

    Sounds familiar huh?

    I’m nervous Lobe will be assimilated into some obscure product, never to be heard from again. But, historically that hasn’t always been the case — LinkedIn and Github (at the time of writing this) are both still stand-alone services.

    On the other hand, I’m really happy for the Lobe team. They have a remarkable product on their hands. And Microsoft has essentially given them AI research carte blanche. Lobe lacked that world-class research ability until now. It’s entirely possible that joining Microsoft might have been the only real way to scale their shared vision of “deep learning for everyone.”

    So all-in-all, I’m into it. ✌️

  • Lobe

    Lobe, is a visual composer of sorts — for building, training, and exporting custom deep learning models. The interface is very Quartz Composer-esque. Check it out:

    If you’re in a hurry, watch from 6:12 for a walkthrough on how to create a project from scratch.

    Lobe is a start-up from Mike Matas, Markus Beissinger and Adam Menges. Matas, an ex-Apple and ex-Facebook Silicon Valley designer veteran — previously released published this demo called The Brain (see below), which was entirely built in Quartz Composer. A fucking awesome demo.

    No wonder Lobe looks like a Quartz composition. Keep in mind, this was published about one year ago:

    This is really really exciting stuff. Building and training deep learning models are simple in concept, but complex in reality. Lobe, may just be the first visual tool to bridge that divide.

    To quote John Gruber:

    Lobe is to CoreML what Illustrator was to PostScript — a profoundly powerful tool that exposes the underlying technology to non-experts through an intuitive visual interface.

    Gruber’s analogy is probably the best, and he’s right. When Desktop Publishing Software exploded in popularity the late 80’s and late 90’s, everything changed. Productivity skyrocketed, and progress in tooling, production and creative software just took of and to be honest, it never really cooled down. We’re still in living in that same epoch, but having different conversations about the same problems.

    Lobe truly exposes a whole new world of problem-solving to non-experts and that will lead to some really exciting tools. Hats off to everyone at Lobe. Remarkable work.