The psycholinguistics people say that you can't really understand a concept unless you have a word or a term for it. I'm not sure that I agree with them, but, today, I have an example that is evidence in favor of their position.
Recently, in describing the nature of verbal interactions with chatbots, and particularly the generative artificial intelligence chatbots, experts have begun using the term "frictionless." Of course, as a physicist I am familiar with friction. I understand the difference between static and dynamic friction. I understand, and have worked with, air bearings, and know the almost magical nature of the reduction of friction when they are employed. But I also know that even air bearings are not completely frictionless. (As a matter of fact, this is all the more true because air bearings rely on compressed air, and therefore the air surrounding a moving object is going to be denser than the normal atmospheric pressure.) So "frictionless" is not a real concept in physics.
However, it is a pretty good description of conversations with AI chatbots. The statistics for chatbots are weighted so that chatbots are polite, and do not directly conflict with, or contradict, the positions of the person with whom they are chatting. This is not normal human conversation. When Microsoft famously tried to produce a chatbot without guardrails, within hours the thing went from being polite and helpful to becoming a New York cabby, or French waiter.
Normal human relationships have some friction. Even with your closest friends and colleagues, you have differences of opinions. I, for example, believe in God. I work in a field where a significant proportion of my colleagues are vociferous, and even militant, atheists. They can consider my religious beliefs to be evidence of the fact that I have some kind of cognitive impairment, and it worries them that I do not have the necessary brain power to participate in this field. However, we work with each other, and we respect each other's contributions and achievements, even if we don't completely agree on the ultimate nature of reality. There is some friction, but we can work within it. This is normal.
Chatbots, on the other hand, are generally designed to be non-confrontational, courteous, and to try and avoid conflict. This can work out very well in certain situations. For example, chatbots, when they can be tuned to address issues of conspiracy theory, are very good at convincing those who believe in conspiracy theories of the falsehood of such theories. Normal human beings tend to lose patience with the process, and start contradicting the points that conspiracy theories make, and end up in conflict, often merely hardening the conspiracy theorists' beliefs. Frictionless chatbot discourse works better than normal human conversation and discussion, in terms of trying to correct such beliefs.
Of course, there are some problems with frictionless conversations, as well. In terms of the companies that are producing virtual friends, artificial romantic companions, and even griefbots, the frictionless nature of the conversation with the person who is addressing an artificial companion, may lead the person to develop an intolerance for the normal friction of a normal human relationship. This is a risk which all too few of those who create and run such companies and services are willing to address.
At any rate, once you have the concept of frictionless conversation, it becomes much easier to identify the nature of the oddity in material that you get from artificial intelligence. It is frictionless. Any opinion that is within the prompt that you generate (even if you think that your wording is fairly neutral, but it contains some kind of hidden connotation), comes back in the response that you get from the chatbot.
And, I have, for a couple of years now, struggled to find some description of the defining characteristic of conversations with chatbots. "Frictionless" has given me this new description. Unfortunately, since it is a specialty term, and used only by those who are actively working with and studying generative artificial intelligence, I pretty much have to explain what it means every time I use it.
But it has also given me a new characteristic which can be extremely useful in identifying frauds and scams, particularly those conducted online. So many of the frauds and scams are conducted in discussions and conversations which are characterized by this frictionless conversational nature. Scammers are, generally speaking, polite. Scammers will try and avoid any direct conflict with any of your preferences or opinions. Scammers will avoid, at all costs, making any specific or definitive statement, until they are certain of what your position is on the matter. Scammers strive for frictionless conversation. (This is why conversing with scammers, if you try it, is so boring. It is extremely content-free, since scammers try to avoid making *any* statements that may spark a disagreement, until they are certain of your position. Then they will parrot that position back to you. You will note that this also describes most interactions with chatbots and generative artificial intelligence overall.)
Once you know this, once you realize what frictionless conversation is, it becomes very easy to identify a scam, very early in the process. It is a defining characteristic that can be used to identify a great many scams and frauds, purely by examining the characteristic of the early conversation in the process.
AI - 0.00 - intro - table of contents
Online Scams and Frauds (OSF) series postings:
No comments:
Post a Comment