Data Access for AI: An Under-Explored Use for Blockchain
There is a vast amount of computing power around the world that is not used efficiently. There are an estimated 4 billion personal computers in the world and 90% of them have free capacities at any given moment. This is to say nothing of the idle capacities of other personal devices, like smartphones and tablets. In effect, this excess computing power is wasted.
Lots of startups are trying to use blockchain technology to take advantage of this inefficiency to meet different economic needs. The blockchain is exciting in this context because it provides an infrastructure for distributed computing power, while at the same time providing an incentive for individuals to link their idle devices to the network.
Many of the startups I’ve come across are focused on providing web hosting and/or data storage solutions via blockchain. However, the use case I think is the most exciting involves artificial intelligence or “AI”.
The idea of applying the blockchain to artificial intelligence is attracting a lot of attention. Similar to the goal of distributed data storage and web hosting, some experts argue the blockchain could encourage a broader distribution of the data and algorithms that will determine the future development of artificial intelligence.
However, many artificial intelligence experts are concerned that IBM, Facebook, Google and a few other big companies are monopolizing all the talent in the field. These giant corporations also control the massive silos of digital data necessary to create and refine the best machine learning programs. Many argue there is a need to democratize data if A.I. is to develop in a direction to benefit all of humanity.
This is where blockchain technology comes in.
At the most basic level, just as the blockchain allows money to be transacted between private parties without a bank or central authority in the middle. Artificial intelligence experts are hoping blockchain technology can allow artificial intelligence networks to access large stores of data and computing power without a large company or other centralized entity in control.
Browser-Enabled AI Calculations
Several start-ups are setting up blockchain-based marketplaces, where people can buy and sell data. The goal, says one expert, is to “decentralize access to data before it’s too late.” I’ve encountered multiple projects looking to accelerate artificial intelligence development in this way. Serious entrepreneurs with impressive academic credentials are behind several of them (e.g. Skynet and Kaasy).
In this article, I want to focus on Hadron. Hadron is creating a marketplace to harness otherwise unused and wasted computing power to solve real-life problems with artificial intelligence. Specifically, Hadron processes and categorizes images and sounds in ways that can be tailored to various uses, such as face detection and image classification. Users who link their devices to the Hadron network perform a piece of this work and in return, they get rewarded in Hadron tokens.
The Demand for AI Computation
Hadron gets its supply of artificial intelligence. tasks from the marketplace it creates, which enables people looking to processes and categorizes images to put in a bid for the task. Hadron assigns the tasks to devices linked to its network and pays out a percentage of the bid proportional to demand. The demand refers to the number of artificial intelligence tasks in the marketplace and supply refers to the number of linked devices. Together these determine the price of computations on Hadron’s platform. For example, if the demand for computation is high, the bid prices will be higher. If the demand is low, the bid prices will be lower.
Centralized vs Distributed AI Computing
In contrast to centralized artificial intelligence computing, Hadron’s distributed computing model has several advantages. The most notable is that scaling the computational resources based on demand is easy. When the demand for computation increases, the price increases, which provides an incentive for existing and new users to link their devices to Hadron.
Hadron recently announced it will be the first dService launched by Aikon, which is a blockchain-enabled API marketplace. This will let anyone use Hadron’s new “Computer Vision API” to recognize and categorize any image.
While still very much in its early days, blockchain technology seems to hold much promise for accelerating artificial intelligence development. More than merely accelerating this development, however, blockchain also provides a decentralized infrastructure that can side-step corporate and government controlled data siloing while also ensuring that the value created through artificial intelligence can be shared in a democratic way.