Blockchain has the potentials of being used in virtually all segments of life. But how can it be used in Artificial Intelligence, a field humans still haven't fully developed?
These are views of professionals in the matters of the blockchain.
One of the biggest use cases for blockchain is in secure computing. Startups like Oasis Labs intend on using blockchain to enable privacy-first cloud computing, which will allow developers to easily deploy their applications in a secure environment. When you run machine learning or artificial intelligence tasks in such an environment, you'll find the intersection between AI and blockchain.
Another application is in other secure machine learning methods. Federated learning, which is being used by Google, distributes the training of artificial intelligence models across different devices, a process that blockchain can facilitate.
There are two immediate use cases I see for blockchain and AI. The first is to build faster and more efficient miners. The second is in the context of the Internet of Things technology. An AI assistant would process human vocal commands through natural language understanding which would then automatically execute blockchain payments or actions in your network.
One of the interesting aspects of blockchain technology is that network participants don't necessarily need to be human in order to initiate transactions. This creates the possibility of commerce between AI-powered machines, IoT devices, bots, etc. It's not too hard to imagine a future with not only self-driving cars but also self-owning cars capable of purchasing their own fuel, paying for their own maintenance, and generating their own revenue through ride-sharing
AI and Blockchain: One way we are using the two together is in situations where asynchronous learners submit to a central learning pool, and it is important to have an exact version of the truth.
Take yesterday. Let's say some AI driven model is deciding odds of whether Judge Kavanaugh will be successfully nominated next week. After the judiciary committee vote but before Senator Flake's condition to his committee vote became public, the likelihood was very high. It went down when the conditionality was revealed; kept going down as more and more senators gave statements; and went to zero when President Trump announced an FBI investigation and vote was postponed to a week after next.
If somebody were to poll this AI model, it's very important for its to know whether the model has ingested signals from all possible sources. Blockchain helps with this single version of the truth. This is a top of mind example but think of the various things that can be done within enterprises. For example, consider an AI planning system that controls a shop floor based on inputs from various machine/ tool specific models, which, in turn, depend on IOT Sensors. Or think of an investment bank with so many disparate teams that a simple question like What is GDP Forcast for 2019, confuses the user.
We use these asynchronous learners all the time for our cognitive computing applications in knowledge management. Someday, we have the ambition to ingest the whole internet. While alternatives are theoretically possible, it is very difficult to do such a thing without blockchain being embedded in the continuous flow of training set.
Arbot Solutions, Inc.
AI is primarily about data and how to gain value from it. A core component of this is dealing with the veracity and, in extension value, of the data and its application (the 5 Vs of Big Data: volume, velocity, variety, veracity and value). I will focus on the veracity and value here.
A central issue in machine learning and neural networks is gaining access to data sets to train the machines/applications. Important from this perspective is knowing where the data comes from, who has touched it, etc. in order to understand the biases of the underlying data. This will affect the ultimate value that can be gained from AI applications and how it affects business and society. For example, if the data is skewed it will be difficult to optimize along the correct vectors to maximize profit for a particular activity or to gain the expected efficiencies in terms of productivity. In extension, with the increasing proliferation of AI applications, there are a rising number of instances where these have deleterious consequences for society, including unintentional discrimination against women and minorities in job searches or loan applications, sending people to prison, and selecting for risk and other factors.
Blockchain, as a technology, can mitigate some of these factors. The naturally distributed nature of blockchain, the transparency and immutability that it provides, and the intrinsic value in creating an enhanced layer of security, are all central to moving the AI industry to a process and state where the underlying data for decisions are auditable and open. Since so many practical applications (e.g. image recognition and organizational decisions) are based on data, it is important to understand and be able to amend the processes and results from both an organizational and societal perspective.