Software writing is slowly moving from human written to computer generated using large amounts of data. Blockchain allows for crowdsourcing of cheap and diverse high-quality data, which wasn’t possible before.
The “IF” “THEN” approach
The Von Neuman computer model I used in all the computers we are familiar with from desktops to laptops via smartphones and tablets. To make the computer do what we want, the typical computer programmer will, elegantly and through more advanced rules, basically write a list of “if” “then” conditions where all possible cases will hopefully be covered. For example, if you type “W,” move the character up. If you type “S,” move it down. If there is a wall in the direction of movement, stop. And so on. As a result, the computers we are familiar with today that use keyboards, mice, and digital inputs (0s or 1s) are very clear. This also includes capacity sensing on smartphones, which are “is the finger here or not?”
However, as seen in self-driving cars, recent advances in computer science are opening the door to computers using other inputs like “what they see.”
Neural Network Programming
Programming these new computers is very different. A training environment is being setup. In this training environment, a lot of known, labeled, organized data is being fed to the computer, and the computer is being told what output is expected for this given data. The computer itself organizes its code to make sure that when this data is fed as input, the output matches expectations.
This so-called “code” now is in fact very different. It is not human readable. It’s just a series of numbers that humans can not interpret. To be more precise, it is often a neural network, and the coding itself is about attributing numbers to the connections between neurons from adjacent layers.
Once a few sets of such data has been inputted and the desired outputs have been achieved, then a copy of this “code” is being installed in the production environment. And as long as data that is similar to the training data is being fed, the outputs are nearly always the ones expected.
Data is the Key
What is important here, however, is that the data, the amount of data, the quality of the data, and the clear outputs related to this data, are in fact what is programming the computer. The human who feeds this data to the machine has a very limited impact on the resulting code and the code’s quality. It is, instead, crucial that the data be as big as possible, that it covers all possible inputs the computer can be presented with, and that it is labeled properly so that the desired outputs can be tested.
Therefore, the data is the key.
Let’s take computer vision, and more precisely, enabling a computer to recognize indoor objects. The applications start with augmented reality glasses and go on and on via software that provides posture advice to enabling robots to cook or clean for you.
To create the software that allows these systems to work, one needs hundreds of thousands of pictures of home interiors as well as hundreds of thousands of pictures of all the types of objects one usually finds inside a home. This includes in the kitchen, in the office, and in the bedroom. Ideally, not only in an apartment in New York, but also on a farm in California and in an apartment in Tokyo or in Moscow. How can a company launch such a product without the data? Where can a company find the data?
Blockchain is the Key to Data
Blockchain allows for micropayments fast, efficiently, and safely to anybody on the planet with an internet connection.
This allows companies to crowdsource data. For example, labeled pictures from anybody who is interested in participating and being compensated for it. Of course, to assure that the data is accurate and clean, one can also use blockchain to make payments to other people, let’s call them voters, to confirm that the pictures and its data are accurate. To make sure that the picture’s quality is good enough. And so on.
This is only possible now, as blockchain becomes more familiar to the data miner and to the companies who need the data. How else could a Silicon Valley startup pay hundreds of thousands of people $1 each for a bunch of pictures?
Furthermore, the blockchain system ensures that once one company has obtained the data, the data miner can also sell the data to other companies. No one large company can take control and dominate this data marketplace or hike the prices. This data ecosystem is decentralized and enables a new function that was not possible without blockchain.
We started work on such an ecosystem in September 2017. We have developed the PIX ecosystem, which is in beta stage at this time. We have started building an ecosystem around pictures to be used for computer vision. At a later stage, this ecosystem can also be used to mine all kinds of other user-generated data from sound to video, to be used for computer voice recognition like Alexa, or perhaps future machine learning and AI computer tools that haven’t even been invented yet.
George Popescu has been in the cryptocurrency space since 2011. He is the co-founder and CEO of Lampix, a table-top augmented reality company. He is also the co-founder of the PIX ecosystem and of Block X Ventures, a blockchain-focused investment bank.
Previously, he sold and exited his most successful company, Boston Technologies (BT) group, in 2014. Popescu was the founder and CEO of BT, which he boot-strapped from $0 to a $20+ million business. BT was the #1 fastest growing company in Boston in 2011, according to the Boston Business Journal and has been on the Inc. 500/5000 list of fastest growing companies in the US for four years in a row ( #143, #373, #897 and #1270).
Popescu earned the best journalist award for Lending Times at the Lendit Industry Awards in 2017.
Previously, he obtained three master’s degrees: a Master of Science from MIT working on 3D printing, a Master in Electrical Engineering and Computer Science from Supelec, France, and a Master in Nanosciences from Paris XI University.