Do I need to understand how something works in order to use it? I don’t think so. For example, the physics of flying is still controversial; I can anyhow travel in an airplane. To give a simpler example, why does a bicycle turn to the right or quickly to the left when the front wheel is turned to the right or vice versa. This phenomenon is probably linked to the center of gravity, but how, I do not care. There are black boxes everywhere.
I need to know a lot of car engines to be able to do something if the car does not move. A mental model can be very superficial, and still humans can drive a car. The same applies to many other devices.
Does explainable AI exist, is another question. Deep learning and neural networks may be explainable if the person who asked the question wants to hear a long lecture. Otherwise AI is not explainable.
Ethics and responsibility are more stringent questions. Who is responsible when an automatic car kills a pedestrian and what does responsibility mean? If (and that is a big if) a culprit is found, is it sufficient to sack or fine him/her. Is it possible, even in theory, to find the persons who developed the system?
“The patient has pancreatic cancer with 84.5% likelihood.” is a mystical statement, but are the key evidence and supporting evidence much clearer? The biopsy examination is as mystical, we have to rely on experts. Why not rely on AI?
Christakis’ article “How AI Will Rewire Us” is highly recommended, not only as a background image. Madrigal’s article “Should Children Form Emotional Bonds With Robots?” is equally provoking.
As to intelligent speakers, I find it repulsive that a program somewhere listens to my words, tries to recognize them and give advice. I do not want to live inside a cozy bubble. The new generation wants to and welcomes new technologies. Everything goes round money: men buying cosmetics are fine, progressive, women not wanting to use makeup are pathetic. Why? Because the former consume more.
I had heard about gist translation, but had not the slightest idea how widespread it is in the patent industry. I also had no clue of the current number of patents: over one million patent applications per year in China only. Two questions need to be asked here: how many of these applications are successful and how important the applications are? Patenting the lower left screw in the battery case of a smartphone, for example, is not very significant.
Two things about patents cross my mind: Does patenting accelerate or decelerate development? I do not know. Does patenting cause increase in prices? When talking about medicine, the answer is yes. New methods are invented to treat diseases. In order to go on, a drug has to be more effective than its predecessors. More effective means often higher price. Nobody is interested in patenting a drug which is 10% less effective but costs 90% less.
The central idea of science is that anyone should be able to repeat the experiment (I wonder how many can build their own particle accelerator and compare the results with CERN?) What is the role of obfuscators in this?
I would never ever trust a machine-translated text without seeing and understanding the original. It was possible to pinpoint the mistake in a rule-based MT. A neural network works independently and can produce text that looks perfect but is useless because the it makes mysterious mistakes and the original text pairs can be erroneous. A gist MT my be usable, but nobody can guarantee that. Ethics of translations is one of my favorite topics, and machine-translation is contrary my ideas of trust. At least for the present.
MTPE is principally a good idea. Now it is misused so that a translator is given a MT text for editing, but the “translation” is partly or entirely so bad that it has to be deleted and everything started from scratch. However, editing is paid much less than translating. I understand there are time constraints, but now the editor has to approve horrible texts in order to return them on time.