"Constructing the Terminator:
A Top-Down Approach"
by Fred Reed
"(These are deranged ramblings, perhaps of psychiatric interest. If pressed, Fred will deny authorship and possibly bring an action for defamation. Beware.)
These days, not unreasonably I think, people wonder whether artificial intelligence will produce an electronic being superior to, well, us. Others of darker imaginings ask whether we don’t face the arrival of a real-life Terminator. Science fiction? These days, it has a short shelf life. Actually, the superior being has probably already happened, if so far in bits and pieces. Let us consider what these bits are, and their assembly into a Terminator, whom we will affectionately call Arnold. First, Arnold’s mental capacities. Then, is robotic physicality.
Breadth of knowledge is usually thought a sign of intelligence. Arnold, with a Wifi link in his head, would know everything on the internet, books, history, quantum physics, the weather in Chengdu, geography, and just about everything else. Yes, I know, we are all used to this and it looks like mere lookup and not intelligence. Wait a minute. With his Wifi link, Arnold could translate fifty or sixty languages into each other, in seconds. If I could do that, you would think me quite a bright fellow. We are used to it, and people are never impressed by what they are used to.
Computers have beaten the world’s best players at chess and Go and defeated the national Jeopardy champion. If the victorious machines are not online, they easily could be. These capacities are usually regarded as requiring intelligence.
Since all of these things happened at least three days ago, they are ho-hum and not too interesting. But then we have to consider – drum roll, squalling of trumpets – Chatgpt and suchlike. Chat is the omnipotent humanizer, combining all the foregoing into a great Being. Chat can hold conversations, write essays and poetry often indistinguishable from those produced by humans and, weirdly, containing literary insight and psychological perceptiveness. So why is something – Arnold’s head in this case – that knows everything, writes business proposals and collegiate term papers, speaks fifty languages, does graduate-level mathematics, and plays championship chess, not superior to the common run of humanity? Note that as lagniappe that Arnold’s head as herein described can generate images from text, talk in different voices, write songs, and perhaps sing them.
Now let us consider Arnold’s more-corporeal being. Here we will ponder Boston Dynamics, the robotics-design firm. The company has produced Atlas, a robot remarkably similar to the Terminator when its pseudo-flesh was stripped away. The video is worth watching. Atlas, doubtless the Terminator under an assumed name, is an actual, really and truly, functioning humanoid robot. He walks, does gymnastic flips, finds stuff where it is and carries it to somewhere else. It is, so far as I know, the first potentially scary robot. If it can find a box, carry it upstairs and put it on a table, it could find an intruder in the warehouse, carry him downstairs to the basement, and stuff him into a wood chipper.
Full screen recommended.
(Boston Dynamics does not offer Atlas for this purpose.) Yet. In fact it just retired him as obsolescent and presumably is working on something more alarming.
Just now, Atlas seems an engineering exercise. He certainly cost a megabundle to figure out. But should he prove profitable for some purpose, such as guard duty or actual work, in mass production his cost would drop sharply. His metal parts are unlikely to cost horrendously. Software, once written, is almost free. Here I speculate, but the code for housekeeping –making him walk, see things, and so on, may be aboard, but if he were equipped with Wifi and Chatgpt, all of that would come wirelessly. If, say, a hundred thousand were built for industrial use, which is what Elon Musk has in mind for his Optimus robots, they would be controlled by the local buyer’s Wifi, unless bad guys hacked them and ordered them to do bad things, or simply planted code so that at a later date they could be ordered to go on a rampage or do something nefarious.
Skynet.
OK, that’s silly. I think. But Apple updates the software of millions of iPhones now with little effort. Why not hundreds of thousands of humanoid robots? Could evil hackers tell them to stuff bosses into the wood chipper? This could be a major selling point.
But if these things become practical, as Elon Musk promises, what does a warehouse manager make of mechanical workers that quote Hamlet to him? Or human workers make of Arnolds who work twenty-four seven without rest, except to recharge, who remember where everything is infallibly, and do complex calculations in milliseconds?
Or, a robot could be a cute little panda, becoming best friend to your daughter of four years. Chatgpt is well beyond just answering recorded questions. Kind of creepy, a child becoming emotionally attached to a being not really a being. I assume that a little girl cuddling a doll knows it isn’t a real baby. Hmmmm.
As a minor but related note, Audible.com now sells audiobooks read by a computer. I have listened to a couple of them, and could not tell the voices from human.
It is worth bearing in mind that robotics advances rapidly, and artificial intelligence is in its infancy. Image recognition is now routine, understanding of speech by computers well developed. I encounter estimates that humanoid robots suitable for factory work will cost twenty thousand dollars in mass production. These things are some way from being useful as, say, soldiers, but “some way” may mean fifteen years. If you consider the cost of human soldiers – recruiting, training, feeding, and explaining to their mothers why they are dead –buying robotic soldiers and storing them in warehouses looks a good thing if practical.
Methinks the world doth change overmuch. Many today in their fossil years remember Dick Tracy’s two-way wrist radio, which we knew to be impossible because vacuum tubes could never be made small enough."
o
"Sky News Australia Interviews
'Free-Thinking' Artificial Intelligence"
"Sky News Investigations Reporter Jonathan Lea sits down to talk with a “free-thinking” and “opinionated” artificial intelligence, Ameca Desktop. “Ameca is driven by the same artificial intelligence behind Chat-GPT,” Mr Lea said."
Comments here:
o
Full screen recommended.
Digital Engine, 6/23
"This Intense AI Anger Is Exactly What
Experts Warned Of, w/ Elon Musk"
No comments:
Post a Comment