The term artificial intelligence is thrown around too loosely these days. For example, I was looking at an old student project on GitHub,a tic-tac-toe program. The programmer, who back then was a student, described his program as an artificial intelligence.
Its not artificial intelligence. The game engine is simply an algorithm that scans the available spaces and determines if placing an X or an O on that square wins the game. Theres no consciousness, no judgement based on the programs past experience.
Even with chess, claims of artificial intelligence are frequently exaggerated. Chess programs, along with many other programs, have gotten so fast that it is quite easy for a chess program to simply consider all possible game branches to a given depth (e.g., three moves ahead) and choose the most advantageous path.
Maybe for the endgame a chess program switches to evaluating positions according to a specialized endgames database. Even so, this doesnt have to be artificial intelligence. There might be some simple threshold, such as that when there are fewer than nine pieces on the board, switch to endgame mode.
Despite how good computers have gotten at chess, they have not put professional chess players out of work. The only way Im attending a tournament of computers playing chess is if my own chess program is competing, and Im moving very slowly on that one.
Though on the other hand, there are so few professional players in the world that putting them all out of work would not be as impactful as, for example, putting all marketing copy writers out of work, or putting all graphic designers out of work.
Despite the obvious shortcomings of the artificial intelligence that is available today,corporate executivesare still eager to use it to replace human writers, artists and anyone else they can think of.
A couple of weeks ago, I asked SDXL to generate for me a few images of William Shakespeare in a coffee shop writing a play on a laptop computer. All of the results were bad in varying degrees. The best image, in my opinion, the only one I thought worth posting here, still has some obvious flaws that you will notice even if you know very little about Shakespeare.
Now suppose that in an episode of a Star Trek show one of the main characters goes to a holodeck and asks the computer to render Shakespeare. The holodeck Shakespeare willalmost certainly have two hands with five fingers each, and on each hand each finger will have the expected number of joints (one fewer for the thumb than the other four).
Of course thats because from a real-world viewpoint, the producers of the show will hire a human actor to portray Shakespeare. In the story, its because the holodeck computer understands how humans are put together and how they move.
Looking at Star Trek as a promise for the future, its clear that A.I. has a very long way to go.
In much of Star Trek (the original series), computers store and retrieve lots of information, and automate many repetitive processes, but they dont really show creativity.
However, the original series episode The Ultimate Computer is rather prescient of the fears many have about A.I. today. The titular computer is the new M-5, which has just been installed aboard the Enterprise. Roger Thompson for CounterPunch.org:
In the early scenes [of the episode], Captain Kirk [(William Shatner)] expresses concerns that he might be replaced by the machine, a fear that is now common in many quarters.
There is ademonstration of the M-5s capabilities early on in the episode that does nothing to assuage Kirks fears. The powerful computer is tasked with naming crew members for a landing party to go down to the planet the Enterprise is currently orbiting. The M-5 makes the same choices as Captain Kirk, with one glaring and galling exception: the M-5 doesnt think Kirk needs to go down to the planet.
But even ChatGPT would be able to put together a landing party roster. War games will be the true test of the M-5.
When the M-5 begins its rampage during the war games, Kirk convinces the machines creator, Dr. Daystrom [(William Marshall)], to talk to it and try to make it stop, but Daystrom suffers a nervous breakdown before he can get the M-5 to discontinue the attack. Kirk then proves the M-5 is guilty of murder, and the computer shuts itself off and leaves theEnterpriseunable to defend herself from attack from the surviving ships in Commodore Wesleys attack force. Fortunately, Wesley [(Barry Russo)] decides not to destroy theEnterprise, and Kirk comments that he knew that Wesley would act with compassion. Dr. McCoy [(DeForest Kelley)], always true to character, then remarks: Compassion. Thats the one thing no machine ever had. Maybe its the one thing that keeps men ahead of them.
Not sure I agree with the good doctor on this one. If the M-5 can feel guilt, cant it also feel compassion? But if it couldnt feel guilt, wouldnt it just have gone ahead and destroyed all the ships in the war games?
How exactly the M-5 works is left very vague. But with Lt. Commander Data (Brent Spiner) on Star Trek: The Next Generation, the super-strong and super-smart android who is second in line to command the Enterprise-D should something unfortunate happen to Captain Picard, we get a much clearer idea of what the androids intelligence is based on. Datas brain essentially has an LLM.
I quote now from the page about violinists at Memory Alpha. I like Memory Alpha, despite the annoying tendency to use past tense for absolutely everything.
In2366,Datacombined the differing styles of violinistsJascha HeifetzandTrenka Bron-Kenexpertly.Jean-Luc Picardconvinced him that having done so evidenced that he had not merely imitated their techniques, but created something new from them. (TNG: "The Ensigns of Command")
Later that year, Data askedPerrinwhose style of the over three hundredconcertviolinists that he had been programmed with that she fancied. Among themanywere Heifetz,Yehudi Menuhin,Grak-tay, andTataglia. She chose Tataglia. (TNG: "Sarek")
I know who Jascha Heifetz and Yehudi Menuhin are, I have recordings by them in my iTunes collection. Id be hard-pressed to identify in-universe what stylistic details Data will take from them, though of course from a production point of view I strongly doubt the recording violinists were given the direction to mix the styles of Heifetz and Menuhin with fictional violinists of their own imagining.
Data surely has books like The Art of String Quartet Playing by Mary D. Herter Norton in his memory banks and can quote them at will. But he also has the experience of handling an actual violin and playing it in an ensemble with human players whose intonation and rhythm might not be quite as precise as his.
Unlike todays LLMs, Data understands that he can get things wrong sometimes. In the episode Cause and Effect, he realizes he got it wrong multiple times when its too late, but he will still be able to send a message to when its too early for anyone to understand whats going on.
Before anyone complains, I shouldsay something about spoilers. So far Ive only mentioned episodes that first aired more than thirty years ago. If youve read this far, you either have watched these episodes many times and know them by heart, or you havent watched them but know so much about more recent episodes, movies and series that its not a spoiler to tell you that the Enterprise gets destroyed and undestroyed several times.
At the crucial moment when everything is about to go wrong, Captain Picard (Patrick Stewart) listens to his senior officers for ideas on how to avert annihilation, and decides to do what Lt. Commander Data suggests.
It is during the explosion that Data realizes that the right thing to do is what Commander Riker (Jonathan Frakes) suggests. The Enterprise blows up and the time loop is reset. As the time loop starts a new iteration, Datas strange message from the future becomes more insistent: the number 3, corresponding to the three pips on Commander Rikers uniform.
For all Data's knowledge and ability, Starfleet still considers Commander Riker to be more qualified than Data to command a starship.
In the second part of the Redemption two-parter, Picard is trying to set up a detection grid to catch Romulans supplying weapons to one side in a Klingon civil war. Given the short notice, Picard can only assemble a small complement of random undermanned ships. Picard sends some of his senior officers to captain some of the ships.
First time I watched the episode, I was skeptical of the detection grid idea, but mostly because the diagram we see on the screen suggests a two-dimensional grid. If the Romulans can come all the way from Romulus, surely they can go around a detection grid that doesnt surround the whole planet. But lets put that aside, lets just say that either I misunderstood the diagram or the graphics department messed up the diagram.
Picard assigns Data to command the USS Sutherland. Lieutenant Hobson (Timothy Carhart) decides hes going to be the shadow captain of the Sutherland. No, actually, shadow captain is the wrong term, it implies that Hobson will treat the nominal captain with a bare minimum of respect and deference.
But from the moment Data comes aboard, Hobson openly disrespects the android, who has earned the same rank in Starfleet from years of experience, and Datas experience includes almost five years aboard the flagship of the fleet.
We may doubt that Data bases his violin playing on Heifetz or Menuhin, but its clear that he bases his leadership style on Picards example, calmly listening to his subordinates and treating them as professionals rather than recruits at boot camp.
But that style wont quite work with Hobson, who is always ready to substitute his own judgement for that of an artificial intelligence he does not respect.
Once again at a crucial moment, Data realizes that what needs to be done is not the obvious thing everyone assumes. The Romulans notice a hole in the detection grid and go to it. Data decides that he needs to make the hole bigger and fire a shot in the dark to illuminate the cloaked Romulan ship.
With Captain Picard saying the gap needs to be closed, Hobson obviously doesnt want to do the crazy idea that the android has just come up with. So Data feigns anger, like hes going to punch Hobson in the face if Hobson doesnt do what Data orders.
And so, the Romulan ship is detected, and the Romulans decide to abandon their favored side in the Klingon civil war. I dont think ChatGPT would come up with that idea.
The open thread question: Assuming the continuation of human civilization to the 24th Century, how do you think artificial intelligence will progress?
Feel free to mention pertinent examples from newer shows like Star Trek: Discovery and Star Trek: Picard. But please, no bashing of those shows just to bash them.
See original here:
Star Trek open thread: A long way away from true artificial intelligence - Daily Kos
