AI is an omni-purpose technology that can bring improvements in efficiency and insights to almost any business process. But through my qualitative and quantitive research and conversations with practitioners, I’ve found that the use cases for AI and machine learning get very industry-specific, very quickly. That’s because they are driven by data and each company’s data is unique to it and by extension, to its industry.
But as the ‘narrow’ AI adoption rolls on, there will also be a huge need for general-purpose AI platforms – a set of tools that can be used to build and deploy any sort of AI application.
This isn’t something that can be attempted by every AI vendor, obviously although you’d be amazed at how many try. A few obvious candidates spring to mind as leaders in this field: AWS, Google Cloud Platform, IBM Watson and Microsoft Azure. But it’s more what they’re planning in the future than offering now that intrigues me.
Although firms such as Google’s DeepMind division are working their way through a portfolio of ever more complex game-playing situations to train their AIs, what they are in fact building are multipurpose AI factories. The AIs can play chess. They can control robotic hands. They can do language processing, which is driving the accelerated adoption of AI in many industries. These are not application-specific systems; they’re AI platforms.
But they’re also not artificial general intelligence in any shape or form – yet. They are merely the next step in the evolution of this sector of the technology industry.
In addition to Deepmind’s ongoing progress, we recently saw the large long-term investment Microsoft is making in OpenAI, whereby Microsoft will invest $1bn in OpenAI, which will in return use Microsoft’s Azure platform exclusively as it pursues breakthroughs in AI. Soon after that announcement, IBM announced a partnership with IBM MIT AI Lab including a 10-year $240m investment.
And there are others pursuing the platform approach, from the veteran SAS Insitute to more recent entrants, including CognitiveScale and C3.ai to name but two. Some will succeed, some will not and some will be acquired.
And of course all the major application vendors – Adobe, Oracle, SAP, Infor, and others – are offering platforms of sorts, but they tend to be tied closely to their applications than be general-purpose AI platforms.
There will be more than one AI platform that succeeds, but there won’t be hundreds.