Dmitry Matskevich, founder of dbrain, argues that current AI is just a pattern recognizer and that data work comprises 80-90% of the value in AI projects. His platform uses blockchain to pay distributed workers globally who generate and label training data, with applications in factory surveillance, medical imaging, and self-driving cars.
Questions this video answers
Dmitry Matskevich.
For the past 10 years.
About 80% (he also says sometimes 80–90%).
About 2 billion people.
About 70,000 people and roughly $1.5 billion in cost (as quoted).
Chapters
This is Benjamin Davis with Cointelegraph tv. So actually, could you introduce yourself for our fans, please?
I'm Dmitry and I'm a founder of dbrain. It's like first human in the loop platform that works on blockchain for creating any artificial intelligence based solution.
And so when and how did you get started in this field and how have you kind of gone from where you began to where you are now?
Actually it's a long story. For the past 10 years I work in artificial intelligence field and I sold my previous company a year ago and gathered a small team of computer vision experts. And then they start to play these recent papers and try to apply to real life business problems. So we built a number of use cases like the craziest of them. We had like the biggest factory that produced meat and we have cameras. Cameras track every action of any human in the field. So for example, we can track whether you washed your hands properly before going to clean room. Yeah. We also do like quality control for pizza, like cancer detection, automatic retouch of any photo. And what we saw that in every AI solution, 80% of problem is data. So now we are focused on building this platform for people, for millions of people across the globe so they can work from India, Nigeria, Indonesia, generating data, cleaning data, cross validating data for our platform. So in every use case it's like 80 or 90% of value creation.
So I know that a lot of times, especially being here, I've noticed that people talk about how the public, there's a lot of misconceptions about AI, the way that people see AI and what they know about it. And I know some people like to talk about machine learning. And could you sort of clarify basically the power of it and what it is?
Yeah, it's really a good question. Yeah. You know like not everybody understands that current state of AI. It's not some like super intelligence, it's just like simple pattern recognizer, very stupid. You have some input and have output. So you just train for any use case those simple pattern recognizers and you just need a lot of data. So people think that AI is about computation, power or maybe algorithms, but there are no such a problem on the market. It's like commodity already. Like still to build high quality AI you just need a lot of data that generated by humans who earn like small money for that. Yeah. So AI is not about high paid engineers, but it's more like about millions of people working on data generation.
Yeah, maybe to give people a better idea of how AI is impacting their Lives. What are some things in the world right now that actually incorporate AI and so that people can see like oh this is actually kind of. I'm interacting with AI already.
Yeah, actually we focused on B2B use cases because whenever you have like 1% increase in efficiency in B2B you earn like millions of dollars. But to get some understanding I can come with some BTC examples. For example, in a new iPhone X you have this face recognition. So this is like really cool example of AI. So there's no need for any like touch etc. I just look at camera and camera understand this is like me and even it understands whether I like looking at my phone or. And I can give like maybe some more examples. For example, in the recent cameras from Google or Apple you have AI that enhance your selfies and. Yeah, and photos denoise so you can make photos in a dark environment and have AI like leverage recent technologies and make it lighter without any noise and self driving cars. Like some of us already tried it and it's gonna be huge for the like next five years. And in any like there's a huge use case in surveillance. Like the biggest company in US have like 70,000 people just watching cameras and they earn for example like 10 bucks per hour. It's like $1.5 billion just wasted. And AI gonna watch cameras and perform really better than any human can. What is cool. Also we have a lot of use cases in healthcare. So for example any pathologist that should analyze biopsy images, they can leverage AI to help them to do this job like really faster. So AI gonna show you like point of interest that you need to focus on. So it's going to like increase your capacity as a pathologist.
You know we're at block show, we're talking about, everybody's talking about blockchain and obviously like blockchain and AI coming together is a big thing. So how has the introduction of blockchain technology, how has it impacted the way that AI is being developed and how has it maybe sped it up or, or changed how, how the whole process is done?
Yeah, yeah, it's, it's a good one. You know what we are focused on? We are not focused on like some buzzwords like merging blockchain with AI. We just focused on how to leverage blockchain to make this like marketplace of people which work on our platform segmenting data more efficient. So what we allow, we allow like for example like 2 billions of people are unbanked globally. Not everybody understand that like 2 billion people, they are not allowed to earn like fiat currency digitally. But leveraging blockchain, we can reach those people and bring them to the platform so they can work and earn like real money, like doing real job. And we also leverage blockchain to eliminate any trust issues. So whenever you have some data on the platform, you're going to be sure that this data is not stolen or sold to anybody else. Because you know, AI is real like issue. Whenever you stole data, Basically you stole AI solution because it's like 90% of solution. Yeah, yeah. That's how we leverage blockchain here, I guess.
I guess some of the projects you're working on are really fun and really interesting and it would be really cool to see them. But in sort of what can people look forward to? So in say like 2018 is there are people are going to be or in a couple of years, what kinds of AI technology do you think are sort of going to come into the market and become very much a part of everyday life.
It's really hard to predict future now because current state of AI is really like primitive. So as I said it's just simple pattern recognizer. But we're going to see that more and more use cases. People going to delegate to AI and first frontier going to be like cognitive routine tasks. It's like any driver, any, anybody who employed in some like pattern recognition, even like pathologists or doctors who need like see something and make decision based on that, it's going to be implemented through AI. So the first frontier going to be like cognitive routine tasks that people do right now. And AI already perform better than humans in most of them.
Very cool. It was pleasure talking to you, nice meeting you.
