Getting our AI straight
Artificial intelligence is everywhere and so are countless misunderstandings, beginning with what it is and is not. AI broadly describes analytical methods and models that extract and recognize meaning from data. AI is not akin to human intelligence despite its references to neurological terms like neurons.
Unlike humans, who learn through life-long experience, AI’s memory is limited to its training, validation, testing, and operational data sets, and selected algorithmic strategies. AI is data-hungry, yet volume is not enough. The data used to train AI systems must be representative of the problem(s) it is assigned to solve. Finding, structuring, and selecting good data can be as daunting as the math and coding behind AI.
The promise, benefits, and perils of data science are magnified by AI. These systems can stereotype and discriminate for and against groups, including people. The implications for companies that want to exploit data and AI are that they involve more than data science and programming. Gaining compelling, competitive value while minimizing risks is a multidisciplinary endeavor. I will touch on these issues in upcoming posts.