We need to democratize AI and Data now!
Artificial Intelligence (AI) is all the hype nowadays. Turn, however, your gaze away from the media limelight, focus your eyes on the enterprise, and you’re likely to get a different picture. Adoption of AI is slow. According to Gartner only one in 25 CIOs had a live AI implementation by 2017. At the same time, AI is virtually owned by the big tech giants whose business models require machine learning to operate, as they live by targeting users and customers with personalized messaging and recommendations using big structured and unstructured data. And yet business analysts agree that the future of AI will be massive. A recent report by ICD predicts compound annual growth (CAGR) of 54.4% with total spending on AI solutions surpassing $46 billion by 2020. The promise is huge, but so are the challenges. Enterprises find it difficult to enter the world of AI, mainly because of three, interrelated, reasons.
The first is finding the best point of entry, in other words a business case that makes good sense. There is a long list of potential solutions out there, ranging from help desk automation, customer support, recommendation engines, fraud detection, chatbots, fault detection - you name it. But selecting a solution and proceeding to adapting the solution to your business model and environment are two separate things. The highest ROI will probably come by automating a core process that is intensely data-driven and which, presently, suffers from too much manual work, inefficiency, low quality, and high cost. But soon as IT and business start analysing those high ROI areas they hit upon the two other big challenges in AI adoption: talent shortage and data.
Attracting and retaining highly skilled talent in data science, AI and machine learning is hard. There is global scarcity of such talent, as well as wide supply fluctuations across geographies. And many businesses are simply not “attractive” enough; either because their brand does not represent cutting-edge technological innovation, or because they do not have the culture or the types of interesting problems that would appeal to high-calibre AI talent. And, of course, there is also the much-discussed problem of gathering and curating the necessary data to train and test the machine learning algorithms: enterprise data are either too many and difficult to access, or too few. As a result of these two challenges - of talent and data - most businesses are either stuck in “pilot mode” or in “we are still thinking what to do” mode when it comes to AI.
But what if businesses could have shared access to AI talent and data?
Sharing talent can come in various forms. You could pool talent, or co-invest and share their creative outcome; which is what start-up accelerators and incubators do. The obvious problem with that approach is intellectual property. Businesses need the competitive advantage of owning their own way of solving a business problem. Acquiring an interesting start-up can confer ownership; but how many startups can you, or should you, acquire? Given the accelerated pace by which innovation in AI is progressing, when is the right time for an acquisition? The risk of acquiring something too early looks as bad as the risk of getting in the game too late and being left behind by the competition; which is another way of stating the “innovator’s dilemma”. But what if those startups, or the many talented individuals that work on AI tools and algorithms, the academic researchers, or other expert teams, could let you embed their AI innovations in your microservices environment? What if there was a marketplace that matched buyers and sellers of technological solutions?
A marketplace for AI tools and applications is not such a novel idea, of course. The big vendors are creating such marketplaces in order to boost their developer ecosystems and lock customers in their cloud. Where those big vendor marketplaces fail is in their centralization: tools and applications must be approved by the power that be, i.e. the vendor, who selects on the basis of their own self-interest. Such “soviet-style” centralization stifles innovation. An alternative scenario would therefore be to replace an authoritarian model of innovation selection with democratization: a truly open marketplace for AI tools and applications where ownership is protected and innovation is encouraged to flow freely. Having an open, secure, and democratic marketplace for data would also solve for the second big challenge of companies: finding the right data to train their algorithms. Imagine being able to quickly find the most innovative tools and all the data you need to spin up an AI application in your environment, and scale it. Or being able to turn your data into a monetizable asset. Or to further monetize your AI innovations by trading them globally.
The future of business in the 4th Industrial Revolution is leveraging AI as a cognitive multiplier to exponentially scale your business. To turn this future into reality we need to democratize AI and data. Distributed ledger applications provide the framework to build such democratic marketplaces, and thus share the bounty of the AI revolution more equitably among its creators as well as among the owners of data.