AI and what we don’t know.

United States Secretary of Defense Donald Rumsfeld is generally credited with first saying:

. . . there are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns — the ones we don’t know we don’t know.

When people talk about Artificial Intelligence (AI) it’s generally in the context of knowns and known unknowns. This makes sense for many reasons because smart people don’t want to look stupid about these things, and sometimes prefer to have proofs to support their comments. It’s also because AI research and applications have been a 20-year journey with many unknowns still unknown.

But what of the unknown unknowns? Without going too deep into pure science fiction, what might these unknowns look like, act like, and behave like?

First, let’s establish 2 critical elements: robots and general purpose intelligence.

Hollywood would have us think that AI is a white, human looking robot like Sonny, or a trash-talking illegal alien like Terminator. In reality, your AI is most likely going to be your smart phone, wristwatch or earbud.

The second thing is AI is all around us today doing highly specialized tasks in manufacturing, services operations, computing, weather forecasting, government surveillance and the military cyberwarfare. These are smart programs, applications, and machines that do specific tasks, learn how to process, and are tightly bound with specific parameters; they do not have an escape hatch or exit strategy, they are monitored by security AI.

So what don’t we know? You might say this is a well known unknown, but we can be pretty sure that AI, like the people who train it, will not learn from history. AI will repeat or perpetuate our mistakes. Let’s hope they are the small fixable kind.

But in the context of business, what are some unknowns?

My thesis centers on technology and philosophy since those are the ultimate foundations of Artificial Intelligence.

Technically, one of the first issues will involve data access and interactions. The free and open Internet is one thing, and it may or not prosper, but companies will be able to generate unlimited amounts of data themselves, so what will happen to open source and open access? Will access be controllable? Will we continue to share?

AI’s will argue and disagree and their weapons will be data, which leads to issues of data quality: completeness, conformity, consistency, accuracy, duplication and integrity. Will there be such things as data poisoning?

The AI judgment will yield false positives: the good deemed bad. Who knows how this might perpetuate downstream for learning and input to other systems.

Finally, how do we cope? Where and how do the issues of correlation and causation resolve? What will an AI black swan do or cause?

Philosophically my thesis on the issues of unknowns centers on the Cardinal virtues:

• Will AI have “prudence” or express wisdom, and how will we know?
• Will there be “justice” and how will it learn such?
• Will AI express “temperance” the practice of self-control, abstention, and moderation?
• Will AI contain “courage”, strength, endurance, and the ability to confront fear, uncertainty, and intimidation?

As unknowns, we must leave you hanging without an action plan or to-do list. What we do hope is you considerer, think, and have an opinion.

Be intelligent, not artificial. Think it through and let it be known.