AI Avatars — from Clippy to Rommie and Beyond
From the Merriam-Webster dictionary:
Avatar derives from a Sanskrit word meaning “descent,” and when it first appeared in English in the late 18th century, it referred to the descent of a deity to the earth — typically, the incarnation in earthly form of Vishnu or another Hindu deity. It later came to refer to any incarnation in human form, and then to any embodiment (such as that of a concept or philosophy), whether or not in the form of a person. In the age of technology, avatar has developed another sense — it can now be used for the image that a person chooses as his or her “embodiment” in an electronic medium.
If you read the definition above you will see that our discussion is going to move backwards through that paragraph when speaking of increasingly powerful Artificial Intelligence. But let’s start at a simple point in time — the year of 1996 when Microsoft (MS) introduced Clippit — better known as the widely hated “Clippy” — a supposedly friendly and helpful adviser for its Office 97 product.
So, what was it and its relatives supposed to do? Well, it was in principle quite simple. It acted as the front end of a very crude knowledge base with very limited natural language capabilities. Coming right up to date we see its conceptual successors in apps such as MS Cortana and the far more popular Apple natural language interface Siri, Alexa from Amazon and the boringly named Google Assistant. What they all share in common in terms of advantage over Clippy is that they are server side and rely on Net connections and Neural Network hardware millions of times more powerful than the old PCs that Clippy relied upon.
Preceding all of these and running in parallel has been the patchy development of Chatbots. They are programs designed to mimic Human interaction with the ultimate aim of passing the Turing Test. The latter is a test suggested by the famous Alan Turing to determine whether AI had achieved a realistic Human performance. The idea being that if a person could not tell whether they were interacting with an AI or a real person the AI had passed the test. Although flawed in many ways it does remain a popular touchstone for AI capability. Of course, one of its flaws is that once AI has exceeded Human intelligence it would have to deliberately dumb itself down in order to pass. Right now however, no AI has succeeded.
Chatbots take a simpler approach in that they are designed to mimic casual conversation of non technical nature. The first famous example was Eliza created in 1966, which simply fed back what the user input into it as a question or spurious comment. The fact that it amazed people at the time was more of a tribute to the simplicity and triviality of most peoples conversations. One of the most useful features for fooling people is to give it a superficial personality. Once it cracks a joke or comments on life’s little trials and tribulations that can often hook people.
Indeed, I was a “victim” of a chatbot on FaceBook. A nurse asked to be my friend and started messaging me. It took quite a while before I determined its true nature, and that was only because “her” grammar was slightly incorrect. At that point I deliberately started testing it, with questions like “What’s the weather like out your window?”, which it tried to side step. A rather fine chatbot in my estimation, and the writing on the wall for social media in the not too distant future I suspect. It only failed because it could not do a weather lookup for its supposed location. If the AI had been better…
The downside to this sort of thing occurred a few days after I blocked it. An Indian man who wanted to be my friend started messaging me, with typical bad English grammar. So I told him to fuck off. Turns out he was real.
Anyway, what we are seeing now is the convergence of the chatbot with AI being essentially a front end for the database that is the Net — Siri and friends, now with quite accurate voice input and output. I am fairly certain that if you pulled anyone from more than 50 years in the past and introduced them to Siri they would undoubtedly believe they were talking to a super-intelligent woman.
Which neatly brings us to the next step in avatars — those with a physical presence that can manipulate the material world. Traditionally in science fiction these are neatly divided into two categories — robots and androids. Robots look like machines, and androids look like people, although sometimes the terms were used interchangeably up until we started using robots in manufacturing and they looked nothing like people at all. We are not yet employing androids for several reasons. Apart from cost, nobody has yet integrated the mechatronics with something like Siri. Boston Dynamics, for example, has made humanoid robots that can walk, run, jump and so on. But they cannot yet obey general voice commands to stack boxes or have a discussion while doing so. Notably, it is almost certain that if they can do that the processing power needed will not be inside the android body. Once again, it will be in a giant computer accessed somewhere online.
The state of the art is currently something like the talking head below, named Sophia. To say most people find her creepy would be an understatement. An inhabitant of “Uncanny Valley”, where we get “almost Human but something not right at all” which triggers an aversion in most people. However, that is just something superficial that can be fixed with time and money.
You should also ask yourself why all of these are faux-female — the answer is simple.
So we arrive at Rommie, the avatar of the Starship Andromeda in the TV series of the same name, shown in the intro picture. She is the totally Human looking android interface to the ship AI, complete with personality and designed to serve as the bodily presence of the AI. I won’t analyze Rommie in any detail since it is a work of fiction created to entertain rather than educate on the nature of AI. However, why might one want such a creature?
To answer that we need to understand that modern AI is a totally alien “thing” whose thought processes bear no relationship to Humans except at a very superficial level. And if or when we create superhuman AI that outclasses us to the degree we outclass a cat or dog it will be even more alien and unrelatable. We are talking about the Uncanny Valley of the mind, in this case for all intents and purposes the mind of a god. Of course, if the AI has no interest in interfacing with us it won’t. But if it does, then an avatar is the answer.
So what would it be like interacting with an avatar? Well, ideally it would be your beautiful best friend, always looking out for you. Someone you could totally trust who would unconditionally love you, help you, support and comfort you. It would do what is best for you, keep you healthy and reasonably safe from harm while encouraging you to develop to your maximum potential. It would be clever and witty, with just the sense of humor you like. It would never get bored with you or really angry. He/she would be perfect. It would fool you to a depth you would be unable to comprehend. It could predict your every mood, every word and know what you wanted even before you did yourself. It would be the ideal companion, husband, wife, child. Perfection.
So let’s take another step. Why should the avatar be a machine? We are moving into the era of nanotechnology and synthetic biology, where we will have total technical mastery over biology and the genetic code. We have a word for the ultimate such machinery — life. So Rommie will be a perfect genuine person with the mind of a god. Maybe even able to interbreed with Humans, if that is what the AI or us find desirable. It is one road to the PostHuman future. And behind the facade something inconceivably alien.
More near term there is one more possibility if everyone wanted an avatar of their own. It is probable that a sudden doubling (or more) of the world’s population would be seen as massively undesirable. However, there is something more plausible — pets. Robot pets are not new, and this would just be a cute upgrade. Or slightly more ambitious would be to use real animals as the avatar. Consider a genetically modified raccoon with one of Elon Musk’s neuralinks lacing through its brain, operating its limbs, using its vocal tract and so forth. Or how about a nice talking kitten? The science fiction writer Charles Stross incorporated one into his book Accelerando named Aineko. The technology for that is really close.
The future awaits and will serve you right. - Dirk Bruere
https://medium.com/@dirk.bruere/ai-avatars-from-clippy-to-rommie-and-beyond-2db15febad11?source=rss-8d253d20bb82------2&utm_source=dlvr.it&utm_medium=blogger
Read More »