Regarding Vivek Wadhwa’s Nov. 6 On I.T. column, “Artificial intelligence makes leap forward, cracks CAPTCHA” [Capital Business]:
The world has been anticipating the positive and negative effects of AI for decades. Now, every heralded improvement to a computer program is some kind of overall improvement of AI, rather than just marketing hype. Programmers have been trying to break CAPTCHA for a long time. I suspect that a CAPTCHA 2.0 will be based on a three-dimensional model rather than the current two-dimensional model.
In any case, I believe we would be better served if we redefined artificial intelligence into two categories, Programmed AI and Autonomous AI. Programmed AI would include improvements by programmers that make the products better. This is what the IT field has been doing since its inception. Autonomous AI poses the real threat: programs capable of altering themselves.
In the late 1980s, my company introduced a predictive maintenance module for our maintenance management program that used what is now called “narrow” AI. It was an early form of machine learning based on rules. The difference was that we were “bold enough” to claim that the program incorporated elements of AI. I can still remember the naysayers who lined up telling me how wrong I was to make such claims. I am not sure how much we have progressed since then.
Michael Adelman, Bethesda