AI is Under-Hyped

As the noise level around AI ramps up, as sure as the famous Gartner hype cycle predicts, the AI critics are also ramping up their messaging. Some of it is around the negative impacts AI could have on society, and some of it is good-natured attempts to debunk the reality of what AI can actually do today. The former is an understandable and reasonable conversation for a society to have, the latter has a significant risk for business leaders if they buy into it too deeply.

If we define AI as some terminator like figure but with a west coast progressive sensibility, always saying and doing the politically correct thing, then yeah, we are in for a disappointment. Even real humans fail that test. More importantly, whether we can imbue silicon life forms with all the characteristics of humans, including idealized and desired personality, is a straw dog that sets up any comparison to reality, to smoking failure.

The business of business is not to care whether some new technology exceeds or falls short of some Hollywood vision of what it should or will be. The business of business is to leverage technologies that can make a difference; to have The Important conversations, at the right time, about what will increase productivity, create customer value, and outstrip the competition.

Business exists in a relatively new context: near-ubiquitous broadband connectivity (certainly in “developed” nations), unprecedented elastic compute and storage capacity and a dramatically increasing amount of data. If by 2035 there are one trillion devices, as Arm is forecasting, then the amount of information available for analysis has not yet hit the knee of the hockey stick. Machine learning, specifically deep learning, is a way to analyze all that data in ways humans would never attempt in a spreadsheet, or otherwise on their own. The businesses that get that, are going to leapfrog their competition. Doing something value-generating with all this data, network connectivity, compute, and storage, IS imperative for most companies, whether it is creating better products or providing a better customer experience.

There are those that will say machine learning is really just the same old, same old statistics that people have always done, and that it is prone to error because correlation does not mean causation and any number of other analytical challenges. I’m not sure it is just statistics, especially if you can combine reinforcement learning. For the sake of argument, let’s say it is true, hasn’t statistical analysis done by people also been prone to this same problem? When the statistical analysis was done by people, was there deliberate and explicit practices of reinforcing models, of improving the scope of labeled data used to capture a broader range of scenarios, of tuning models over time? To be really good at something it is not just enough to crank the handle, you have to know what you are doing, and you have to be doing it better than the competition. That is part of what will differentiate companies in the age of machine learning; not just doing it, but doing it well.

I get that the term “AI” is often used to imply, accidentally or nefariously, a greater capability than what exists today. I get that correlation does not equal causation. What I don’t get, is companies that buy into this misdirection, stand back while the competition charges ahead, achieve returns to learning, and moves boldly and deliberately into a new basis of competition and new value propositions/value chains.

Business leaders should leave the distracting conversation of what is, or is not, the definition of AI to pundits, columnists, academics, and scriptwriters. Business leaders should be focused on a bold commitment to what can already be achieved today, what will be achievable in the near term and drive greater productivity, better customer experience, and new sources of customer value.

This site uses Akismet to reduce spam. Learn how your comment data is processed.