Skip to content

Lobe's ridiculously simple machine learning platform aims to empower non-technical creators

--Advertisements --

Machine learning can be the tool of day for everything from particle physics to the recreation of the human voice, but this is not the easiest area to tackle. Despite the complexities of video editing and sound design, we have user interfaces that even allow a curious child to be interested – why not with machine learning? That's the goal of Lobe, a startup and platform that really seems to have made AI models as easy to assemble as LEGO bricks.

I spoke with Mike Matas, one of Lobe's co-founders and designer of many popular digital interfaces, about the platform and its motivations for creating it.

"There have been a lot of situations where people have thought about AI and have these great ideas, but they can not execute them," he said. "So these ideas are as lost unless you have access to an AI team."

It also happened to him, he explained.

"I started doing research because I wanted to see if I could use it myself." And it's hard to pierce the veneer of words, frames, and math – but a once you understand that the concepts are really intuitive – in fact, even more intuitive than regular programming, because you teach the machine as you teach a person. "

But like the hard shell of jargon, the existing tools were also approximate – powerful and functional, but much more like learning a development environment than playing in Photoshop or Logic.

-- Advertisements --

"You have to know how to assemble these elements, there are a lot of things you have to download.I am one of those people who, if I have to do a lot of work, upload a bunch of frameworks, I just give up, "he said. "As a user interface designer, I have seen the opportunity to take something really complicated and reframe it in an understandable way."

See also  This is how Ripple faced questions at a Blockchain hearing of the British Parliament

[youtube https://www.youtube.com/watch?v=IN69suHxS8w?version=3&rel=1&fs=1&autohide=2&showsearch=0&showinfo=1&iv_load_policy=1&wmode=transparent&w=640&h=390]

Lobe, whom Matas created with his co-founders Markus Beissinger and Adam Menges, takes up the concepts of machine learning, such as feature extraction and labeling, and places them in a simple and intuitive visual interface. As demonstrated in a video tour of the platform, you can make an app that recognizes the gestures of the hand and associates them with emoji without ever see a line of code, not to mention write one. All the relevant information is there, and you can go to the end if you want, but you do not have to do it. The ease and speed with which new applications can be designed and experimented could open the field to people who see the potential of the tools but who do not have the technical know-how.

He compared the situation to the early days of PCs, where computer scientists and engineers were the only ones who knew how to exploit them. "They were the only ones able to use them, so they were only people able to come up with ideas about how to use them," he said. But by the end of the 1980s, computers had been transformed into creative tools, largely because of improvements to the user interface.

Matas is expecting a similar flow of applications, even beyond what we have already seen, when barriers to entry fall.

"People outside of the data science community will think about how to apply that to their field," he said, and unlike before, they will be able to create a model of work themselves.

-- Advertisements --

A series of examples on the site show how a few simple modules can give rise to all sorts of interesting applications: lip reading, tracking positions, understanding gestures, generating petals of realistic flowers. Why not? You need data to feed the system, of course, but doing something new with it is no longer the hardest part.

And in accordance with the community's commitment to automatic learning about openness and sharing, Lobe models are not an exclusive property that you can only use on the site or via the API. "Architecturally, we are built on open standards like Tensorflow," said Matas. Practice on Lobe, test it and modify it on Lobe, then compile it on the platform you want and take it.

At the moment, the site is in closed beta. "We have been inundated with answers, so it's clear that it resonates with people," said Matas. "We're going to let people in slowly, it's going to start small enough, I hope we will not be ahead of ourselves."

Advertisements