In this article by Carl Miller, he exposes the hidden reality of ridiculously complicated algorithms and why it can be dangerous to rely on them.
Some Important Points
- All algorithms follow the same steps: input > process > output.
- Though the basic steps stay the same, algorithms have become so much more complicated, with the input stage coming from results of other algorithms.
- If there is something wrong with the output, it is physically impossible for developers to figure out where the problem is coming from. In the same degree, it is hard for a human to understand how a machine makes its decision.
Implications for AI
As technology is becoming widespread, algorithms have a tremendous impact to the way our realities are constructed. They can affect which movie we decide to watch, the kind of information we have access to, as the Facebook-Cambridge Analytica scandal has shown, even leaders that we choose.
And yet, despite their capacity to shape our realities, our knowledge about how algorithms work does not increase. This is partially due to the fact that a number of these algorithms are proprietary and so, is not subjected scrutiny. How then can we be sure that the choices machines and algorithms do for us is the best? How do we know what is true and what is not?
Aside from the lack of transparency to the public, complicated algorithms also have the tendency to be unpredictable. Hence, their very developers tend to have little to no idea on how unintended consequences come about. How then can AI developers troubleshoot concerns effectively, particularly in cases of more advanced machine learning?
‘I’ll lose my job if anyone knows about this.” There was a long silence which I didn’t dare to break. I had begged to make this meeting happen. And now the person I had long been trying to meet leaned towards me. “Someone is going to go through your book line by line,” he said,…