Aspirational vs. Artificial Intelligence
I strongly advocate disposing of this term. Renaming the entire field Aspirational vs. Artificial would be a step in the right direction and force us to rethink the assumptions we built this industry on in the hope of a better future. If you must, please use Cognitive Intelligence as described here.
Artificial intelligence is just that, Artificial. We aspire to create intelligence, and yet we foster ignorance. Marketing has far exceeded any capacity we have to fulfill the promise. The data shows we don’t understand its practical application in most use cases. From simple mistakes in food processing control systems, creating and testing drug models, to advanced flight controls that fail, we can’t afford to trust flawed models. AI as we know it today is killing us all, literally.
The problem starts with mundane aggregation and has very little to do with the sexy end product. The entire system is flawed, and we blindly accept what we don’t assume to be correct. It’s wrong from the start — all of it. Want to test that thesis in your business? Ask your data scientists, how good is their data, and where do they get it? That is just one factor in the 85% project failure rates.
When building a data set to “teach” the computing models, we rely on whatever information we can accumulate. Sometimes that is highly targeted, hygienic, and robust in a controlled market. Financial, automotive, and aerospace are good examples of acquisition, controlled and managed by the consumer. Even when the resource is rich, collected, and vetted, we often can’t take in more than a sample due to computing constraints. So we create abstractions and projection models to synthesize more data we believe is statistically representative of the base data. How can you acquire something you don’t know is missing? What happens when you discover it was missing from the start? How many companies resort to risk and reward analysis for refactoring?
Bias problems. Bias starts with the acquisition. Check out this post for more detail. We discuss the issues in terms of sex, race, relationships, culture, status, and a myriad of other perimeters. But the actual crime is at acquisition — models created from “available” data are not complete in almost all cases. Worse, if the information were robust, with 100% coverage, it would likely be too much too quickly for systems to ingest and store. Formats and acquisition bias further limit utility. This reality creates a terrible cycle of miscalculation that we magnify the statistical values to create teaching sets.
Then comes data integrity. How often do we go back and revisit the accuracy and consistency? Flight control systems are an example where an entire industry has focused on solving these issues. Even in this highly controlled, closed-loop model, we don’t have AI, and we continue to have severe edge case failures. Humans, all be it less, maintain control of these systems. Autopilot car programs are streaming engines that can process vast amounts of information from models created over a decade. Impressive as they are, they have predictable limits. Acquisition resolution, model identification, and utilization,
We employ armies of humans who work behind the scenes to clean data and process information. They are the silent force behind artificial sentience. They want a voice now. That voice will precisely highlight how flawed this process is from who is “hired” to who is “fired” and under what rules they operate — another more insidious and inherent bias.
I could ramble on for hours, but my target audience is already saying, so what?!
Stop using Artificial Intelligence as a term and business process that does not exist in reality.
Call it what it is, machine learning. It’s not sexy, but we need to lower the tone and tenor in the marketplace. Stop using it to promise reduced costs, better response times, and magical insights. Most of all, we need to stop scaring employees that this technology will replace them anytime soon. Toyota sees the truth in this position. Check out this link on robotics on assembly lines.
If there is any promise, it’s in solving our world’s graying and its impact on younger people to remain productive. Toyota has your back.