While most everyone agrees that machine learning and artificial intelligence (AI) demonstrate a lot of potential as smart analytic and presumptive tools within the enterprise, the technology still has a way to go before it can be considered a fully trusted alternative to human decision-making in all applications.
Clearly, machine learning/AI has proven to be a highly capable tool in sifting through massive amounts of data. Some examples would be the Human Genome Project and other biomedical research from population genetics to cancer research and schizophrenia. These tools function from an inference standpoint to pinpoint correlations that might otherwise be missed. “Finding something that is too obscure and too multivariate for a pretty bright person to see through and figure it out,” noted one CIO Executive Council (CEC) member.
Challenges include the lack of intuitive intelligence and judgement capabilities. Although the technologies can be smart, they need to be able to include a variety of factors in order to make a correct determination. These systems are also designed to make mistakes in order to learn and become smarter, which is not be a good thing when you are dealing with mission critical or time sensitive transactions.
These are just a couple of issues raised by tech leaders in a roundtable discussion hosted by the CIO Executive Council. The meeting was convened at the request of a CEC member wanting to hear how other executives viewed the topic and if these technologies were being used or piloted at their respective companies.
Smart Systems Lack Intelligence Savvy
Issues noted during the roundtable include:
- While the delivery mechanism has a lot of potential, the piece that is missing is intuitive intelligence and judgement, which is critical in healthcare. The technologies can be smart, but it has to have the ability to factor in a variety of things to make a determination. Right now, it seems we are far from that point.
- The component technologies are things we live with every day for case-based reasoning – things like natural language processing, speech recognition. But now, people are looking at it as more than the sum of these individual parts. Specifically, there are two application angles on this: How does machine learning and AI affect our own internal environment? And, how do we use it to simplify our world?
- Security is a key application area for machine learning/AI, although many experts think a lot of work has to be done before turning over the keys to intelligent systems. In the financial industry, for example, there may be miscalculations and safeguard missteps that can have a domino effect when it comes to automated high-frequency trading. These intelligent systems are by and large ‘blunt instruments’ and have to make mistakes to learn and become more proficient and reliable.
- One of the biggest hurdles may be social acceptance as machine intelligence evolves beyond being a tool used to clarify statistics into an intuitive and decisive technology. We are still debating the statistics argument in presidential elections and a long way from letting a machine learning algorithm settle an election or being comfortable with a non-human decision.
IT executives are also concerned that businesses may be driving AI development more from a revenue and profit standpoint than general best-use of the applications. “I’m not at all surprised to hear that the vast majority of conversations you have are business driven versus technology drivers for adoption,” says a roundtable participant. “You need to see more IT people bringing forward those use cases and pain points with the business because they’re not really bringing it to the table.”
CIO Executive Council round tables are produced by the CEC at various locations across the country, based on member requests to discuss specific topics and engage with other members who are involved in or interested in the same topic areas. The meetings are candid and offered exclusively as part of the CEC’s membership programs and services.