When you saw all the issues presented in Stephen’s videos, you might want to look for a master switch for all of the AI in the world and switch it off immediately and completely. But that’s not going to happen. So I go on thinking what is most important for me.
Unlike the issue described in my previous post, my following concern may be clearly subsumed under major categories, such as ‘Opacity and Transparency‘, and under the codes of ‘Honesty’, but among these broad topics it might get buried. It is: Mandatory labeling of artificial agents.
Initially, it might not seem like a big deal, since the first few instances of AI that we encounter, often proudly declare how they use progressive technology, or they affect only trivial occasions. But very soon, artificial agents will be able to appear totally indistinguishable from genuine humans. And then the trust in their owners’ honesty of labeling the AI, will become an extremely critical ingredient for living together with such creatures, when the subjectivity and individuality of the human is the only distinguishing feature. And therefore, I think, we need to be careful even with small subtle indications of fakes and inauthenticity, because they might blur the difference and make the distinction ever more impossible. (Examples below.)
Concealing the artificiality will not only be a matter of dishonesty which some ‘intellectually’ privileged may use in their “war on stupid people” for gaslighting and patronizing them, and will not only grow a cultural climate of ubiquitous suspicion of pervasive fakes.
In education, it simply won’t work, since human learning and imitation sometimes depends on genuine human partners: In a study by Andrew N. Meltzoff (whose office kindly sent it to me), titled “Understanding the Intentions of Others: Re-Enactment of Intended Acts by 18-Month-Old Children“, they experimented with imitation from human vs. inanimate agents and found that “Children showed a completely different reaction to the mechanical device than to the person“. As I wrote in my predictions for this year, people are craving for the real, the genuine and the authentic, and this is, for me, the cue that AI’s role of personal coaching of unique individuals, will be limited.
In many cases, we don’t really care if we are dealing with a real, identifyable, person who addresses us personally. If we get an email from a support center, it’s not too important if the name in the From: field (the boss?) actually is the person who wrote the response, or if he or she crafted the response specifically for me, composed it from reusable fragments, or used a stock template. But to impress the customer, vendors will tend to appear as personal as possible.
I am old enough to remember when advertising letters first used our name within the text rather just on the envelope: wow! When we got used to this sort of computer printout (in equidistant typewriter fonts), our names suddenly appeared in proportial fonts, like real printed advertising brochures! And soon after, it was not only in the header but right within the body … or on pictures … or on a label of the keys of a car we should win from a lottery.
A few more examples that are still innocent but may be not in future: The announcement in our suburban train starts with “Meine Damen und Herren” = My Ladies and Gentlemen. Without the “My”, the phrase would be incorrect and alien (probably similar to French Mesdames et Messieurs). And as long as the automatic recording was at least done by a human, it is still tolerable, as is the “Heartfelt welcome” and the closing phrase “We wish you a nice day”, since the plural ‘we’ might be interpreted as the AI together with its owner organisation. But a 1st person singular should not be abused for faked “feelings” of an AI — or am I too fussy here?
The more we get used to innocent fakes, the more difficult it will become to detect serious ones. For the birthday of a friend, platforms remind me of the day and advise me to “show them that you thought of them”. In the news, we often see that reports about casualties and crimes contain generic stock photos of ambulances, fire trucks, and police cars instead of actual photos of the event. (Maybe my tolerance is too low, and as a non-TV-viewer I am not already sufficiently used to autumn scenes in spring, and family relatives that don’t resemble each other, and windows that always only show some artificial light behind them rather than visible life…) And even cats seem to love artificial ruffling.