Unlike blockchain technology and IoT, AI — which, in one
sense, is about creating computer applications that act as smart as humans — is
not a new concept. Research began in academia in the 1950s, and the subject was
popularized in the 1968 science fiction movie “2001: A Space Odyssey,”
featuring the humanlike HAL 9000 computer. Usable computing systems running AI programs
emerged in the 1980s, in the form of expert systems that were able to apply
pre-programmed knowledge and make rules-based business decisions. In 1997, an
IBM AI system called “Deep Blue” beat reigning world champion Gary Kasparov at
chess.
Today, AI — and a related focus on machine learning (allowing computers to
learn based on supplied data) — is rapidly evolving due to the development of high-performance
microprocessors that are able to work quickly with very large amounts of
in-memory data.
Data is a key ingredient of approaches to developing AI and machine
learning, which are now being applied to a wide variety of uses, from stock
trading to chatbots to self-driving cars. There is barely a business or human
activity that is not considered as a target for AI in future years and decades.
This is where blockchain technology comes in, which has its own role to play in
the world of data. While work on aligning blockchain and AI technology is still
emerging, a few development threads now exist and 2018 looks to be a year when
progress toward convergence will accelerate.
Enter Blockchains
Blockchain technology’s ability to guarantee the accuracy of data makes it
useful for a number of AI applications, both for feeding data into AI systems
and for recording results from them.
For example, CognitiveScale, an AI startup that’s backed by IBM, Intel,
Microsoft and USAA, among others, is leveraging blockchain technology to
securely store the results of an AI application that it built for regulatory
compliance in the financial markets world. That’s an industry with a lot of
regulatory scrutiny, so being able to store AI-derived decisions securely helps
market participants stay on top of onerous reporting requirements.
IBM is marrying both its blockchain offering (based on the open-source
Hyperledger Fabric codebase) and its Watson AI platform for a range of
industries. One early project involves Everledger, which is applying blockchain
technology to track the provenance of luxury items, including diamonds.
Leveraging Everledger’s data store of individual diamond characteristics (more
than a million of them, secured by IBM’s blockchain), Watson is applying knowledge
of thousands of regulations to ensure that diamonds comply with United
Nations edicts preventing the sale of conflict minerals.
Some see certain implementations of blockchain technology as
benefiting the development of AI applications, which improve as more data is
made available to them for training the machine-learning models upon which they
are built.
The Importance of
Data
In a lengthy blog
post, Trent McConaghy, founder and CTO of BigchainDB, set out his reasoning
for why blockchain-enabled, decentralized networks encourage the creation of available
data — essentially because individual data silos are replaced by a shared and
accurate ledger — which can be leveraged to train better AI models. For
example, a consortium of banks sharing credit card usage data via a shared
ledger might be expected to lead to improved, faster and more accurate fraud
detection.
BigchainDB is
currently rolling out a database that combines the capacity and performance of
NoSQL database technology with the immutability and consensus of blockchain
approaches. Already, the company is working on enterprise implementations in
partnership with the likes of Capgemini, Daimler, Porsche and Toyota.
Moreover, the company is also deploying its technology to underpin the Interplanetary Database (IPDB), a publicly owned and
governed database network that it sees as becoming the database for a “decentralized
world computer,” much like the platform the Ethereum community is creating.
Most recently, BigchainDB began to evangelize the open Ocean Protocol, designed to facilitate
data exchange as an enabler for AI applications. Core software that supports
the protocol leverages the IPDB. A test network for the protocol should be
launched by the middle of 2018.
Other recent initiatives to develop
transactional interoperability between blockchains should also feed into the
availability of shared data to drive better AI applications. Last month, the
Blockchain Interoperability Alliance
announced
its formation with a mission to connect blockchain networks being created
by Aion, ICON and Wanchain.