G'kar is mah boy! But anywho, I think we can add on to this the idea that no matter what safeguards we put into place, an AI might find a way to circumvent it.
Like the zero'd law of robotics, which the AI robots developed themselves and led to the robot uprising. Where instead of focusing on human, it instead focuses on humanity as a whole, where if through the death of a human, humanity benefits, that is acceptable, even it ignores the other laws in the process:
“A robot may not harm humanity, or, by inaction, allow humanity to come to harm.”
Also, on a side note, Foundation the show is kickass, very very much worth a watch.
It’s worth pointing out that a chatbot will not (and can not) mean any threats it makes - and may well be unable to carry them out. Chatbots do not think. The risks there are exaggerated.
Yeah, I'm thinking of some future version that is actually able to be described as an Artificial Intelligence, and not just a tool that gathers and compiles information and machine learning.
More likely to happen is that we put too much stock on AI for complicated tasks and it either gets something wrong so bad that it causes a calamity, but also, that we won't have people that can recognize what is wrong, because we let AI deal with the complex tasks.
We could end up facing a point where without everything AI can do, we suddenly have no people left who know how to do something, losing that knowledge (At least for a while)
In terms of story telling and character development, the real stars of Babylon 5 were G'Kar and Londo. Most everybody else (Sinclair, Sheridan, Delenn, Garibaldi) each had their destinies mapped out for them and each stayed on their trajectory, but G'Kar and Londo grew and changed and developed as a result of their experiences, not only individually and more importantly in their relationship with each other.
And Susan Ivanova was ... well, Susan Ivanova, the one and only:
LCDR Susan Ivanova:(after a narrowly averted cluster fuck caused by not having followed her explicit instructions) "[Now repeat after me The Babylon 5 Mantra:] Ivanova is always right. I will listen to Ivanova. I will not ignore Ivanova's recommendations. Ivanova is God. And if this ever happens again, Ivanova will personally rip your lungs out. Babylon Control out. [Sigh] Civilians. [Looking up] Just kidding about the God part -- no offense."
Lt. Commander Susan Ivanova, Voice in the Wilderness Part I
Of course, I did not enter into watching that show unprepared. When I was on CompuServe (mid-80's to earliest 90's, where I got my start discussing creationism online), on the Science Fiction Forum was this guy, JMS, pitching to us this idea for an epic show, Babylon 5. He had the whole thing mapped out for a five-season run (including "trap doors" for characters in case a cast reshuffling was needed, as happened with Sinclair having to be replaced with Sheridan). He justified his "Five Year Plan" based on what had happened with Star Trek, which post-cancellation had grown into a media empire through syndication, but that nearly did not happen. In order for a show to be sold to syndication, it needs a certain number of episodes since it will air 5 or 7 days a week and you need enough episodes to keep that schedule from repeating too soon. Three seasons worth (about 100 episodes) or more is ideal, but two seasons (50-60) or one (about 30) simply won't work. Star Trek had been cancelled after two seasons, but then renewed for that vital third season due to fan protest. If not for that third season, Star Trek would never have made it to syndication and never had lived on. Therefore, for his new show to live on, JMS wanted at least five seasons. Ironically, after a single post-cancellation syndication run (on TNT in the USA), I couldn't find it airing anywhere -- now it's finally enjoying limited streaming.
BTW, you must already know about the Lurker's Guide to Babylon 5, which includes extensive notes on the production of every episode. JMS took to calling the transient "street people" living on the lower levels of the station "lurkers" in honor of the computer service "lurkers", members of a forum who do not post or otherwise participate, but rather only lurk in the shadows reading what's happening. It was those pre-Internet lurkers to whom JMS was reaching out.
Also, on a side note, Foundation the show is kickass, very very much worth a watch.
I've caught up on it. I had read the first three books of the Foundation Trilogy (invoking a running Hitchhiker's Guide to the Galaxy joke about the proliferation of book trilogies as they expand past that third book) back in 1977/8 when I was starting my career as an Air Force digital electronics technician (supplanted a few years later with a new career as a software engineer, hence my profile entry of "Retired Intelligent Designer") -- BTW, those stories dated from around 1951, the year of my birth. I was especially struck by how the Foundation started out exporting its technology in the form of religious ritual: basically, "Recite these prayers, flipping these particular switches in the sequence dictated by the ritual, and then press that red button." Well, computer technicians tend to get divorced from the basic electronics instead exercising all their expertise "chasing sparks" though logic diagrams; present most of them with a transistor circuit and they begin to panic (eg, I was studying a circuit that used a transistor in switching mode and I asked a sergeant to verify if I got it right and his reaction was, "You can understand that?").
At that time, our shop's QC (Quality Control) NCO, a Tech Sgt (E-6), had expressed his opinion that nobody really knows how electronics works: It's all just FM ("fucking magic") -- and the more advanced electronics was PFM, "pure fucking magic". According to him, we don't know why it works, all we know is what to do to fix it. And to some extent, that is all a technician needs to know: how to troubleshoot an equipment failure and then how to fix it. If your corrective actions always (or usually) work, then keep doing that even if you don't know why they work. The practice of medicine tends to follow that same trajectory.
The show has taken some different paths than the books (though after nearly half a century, my memory is a bit hazy) but which I have found thought provoking. One of the very few things I miss about my ex-wife (a monster) is our shared appreciation of science fiction in which the tech and the special effects are secondary, but rather the interesting part lies in the questions of people, societies, how things (and people) work and interact, etc. Zum Beispiel, the German names I know for "science fiction" are Zukunftsromane ("future stories") and Möglichkeitsromane ("possibility stories"), IOW, start with the premise of "What if ... ". Tech and visual effects are just eye-candy; ideas and the exploration of the possibilities are the meat that we come to feast on. That's the really interesting part.
I use Wikipedia a lot to follow TV shows and their schedules. Foundation recently finished its third season and has been renewed for a fourth. Judging from a pattern of 1.5-2 years from the time of renewal to produce a new season, I estimate the next season to come out some time in 2027. Let's see. The Mule has been vanquished and the Empire has collapsed upon itself. So many possible new directions that the next season could take. The only certainty is that robots will play a prominent role.
Also on AppleTV+ I would very highly recommend For All Mankind, an alternate history in which the Soviet disaster of 1966 had never happened, resulting in the Soviet Union being first to the Moon; from that link:
quote:Ronald D. Moore explained how historical reality had been different from the series: "Sergei Korolev was the father of the Soviet space program; in our reality, he died during an operation in Moscow (in 1966) ... And after that point, their Moon program really never pulled together.... Our point of divergence was that Korolev lives, ... and he made their Moon landing happen."
Basically, once we made it to the moon, we got bored with space and our aspirations just basically withered away through apathy. In this alternative reality, the space race continued for many decades to come. What could have been had we stuck to it.
Ronald D. Moore is also known for the reimagining of Battlestar Galactica, of which my one CO (CAPT, O-6, USNR) remarked, "It's not science fiction, but rather military fiction." One distinguishing feature of Battlestar Galactica was the unpolished "in the field" camera work in which a ship would appear off-center and the camera would then zoom in, also imperfectly and off-center, lending the scene an extra feeling of authenticity.
That camera technique is also used in For All Mankind, as in this clip of the three contenders for "First On Mars" (NASA, the Soviets, and a private firm called Helios) launching in the same window and on the same trajectory where the Cold Equations hold sway until NASA makes its surprise move:
At about 5:58, stand by for the NASA CO's order, "Mr. Halliday. Beat to quarters!" As NASA's "Operation Jolly Roger" commences ("Yo ho! Yo ho! A pirate's life for me!").
Each season takes place in the next decade from the previous. I estimate the next season, dealing with the mining of an asteroid captured into Mars orbit, to drop, well, any time now from late 2025 to Spring 2026. I think I saw something about release schedules not mentioning the show, so it will probably be later. That will give you more time to catch up on your binge watching.
It’s worth pointing out that a chatbot will not (and can not) mean any threats it makes - and may well be unable to carry them out. Chatbots do not think. The risks there are exaggerated.
Perhaps not as long as we keep them firmly ensconced in their sandbox (or their sandbox), but the plan is to eventually bring them online. "Online" means that they will actually be controlling very real systems. I mean, otherwise, why even create them? Unless we end up putting them to useful work.
And even before then, if we give them access to email, then what's to prevent them from sending an incriminating email to carry out a blackmail threat?
You remember that scene in 2001: A Space Odyssey, the one I refer to as the most cold-blooded mass murder in cinematic history:
We're putting computers in charge of flying airliners, albeit "under human supervision", but in many cases the humans don't understand what the computer is doing. My friend got me watching Air Disasters (original title in Canada, "Mayday") -- she was afraid of flying until she watched that show, because it shows that when a disaster happens they investigate the shit out of it and, as a result, implement corrective actions to prevent or mitigate the same thing happening again, thus incrementally making flying ever safer. One of the problems that show unearths is when what the programmers tell the software to do under certain circumstances which the pilots are unaware of or which runs contrary to a pilot's instinctual response to a situation.
Another different potential problem is "sneak-paths", in which variable initialization can differ depending on which thread of execution was followed resulting in vastly different outcomes. A classic case was a radiation therapy apparatus where the technician entered a wrong value but then corrected it, but variable initialization was not corrected resulting in a potentially fatal radiation dose being delivered (one programmer I worked with absolutely refused to program any medical equipment in order to avoid causing such damage).
quote:Those who refuse to learn the lessons of science fiction are doomed to live them.
quote:
And even before then, if we give them access to email, then what's to prevent them from sending an incriminating email to carry out a blackmail threat?
There’s the fact that they are incapable of meaning the threats. The ability to do things is coming - early versions of “agents” are already available but the biggest risk there is security. And I’d put errors as the next one.
The issues with possible errors you highlight - and the difficulty of diagnosing and fixing them is certainly real, but it involves models trained for specific tasks which are generally more reliable and less vulnerable than the chatbots. It just means that areas where safety is very important need different rules - and mostly they already have far stronger safety rules so extending that to AI doesn’t seem problematic. Self-driving vehicles seem mostly good, although there are issues - especially with Teslas (not helped by Musk’s habit of over promising and under delivering - or the comparatively limited sensor suite).
Woah, this is quite the in depth response, and apart from G'kar and Londo Molari being the most interesting, which I fully agree with, I'm going to have to digest the rest a bit more thoroughly before I fully get into an answer. Maybe we should set up a coffee house thread about in depth analysis of really good shows and get into the nitty gritty, because I personally would love to fully get into it. I remember Compuserve and using it. I'd love to get all sci fi with you if you're up for it, there's more shows or things or stories and books I'd love to get into!
I'm rather frustrated because my old hardware won't run the last 2 versions of FSD but it still does almost all my driving. The google navigation is the biggest source of annoyance. I'm near 100% sure it has saved one fender bender ( since I forgot to save the cameras and it's hard to tell about something that didn't happen I can't be totally sure).
The jury is still out but things are moving very fast. Let's pick up the discussion in 3 months (musk time) or 6 months (more realistic).
That really doesn’t contradict my point. Tesla FSD is not considered safe to operate without human supervision - and indeed has caused accidents. Even if some of those could and should have been prevented by driver intervention it still supports my point - if human intervention was needed the system wasn’t adequate to cope with the situation.
The line between level 2 and higher levels is, by definition, sharp and clear. But in the real world it is fuzzy.
The robotaxis in Austin seem to be operating fully autonomously but with a monitor as a passenger. We'll know by the end of the year how true that is when they do or do not take the monitor out. There isn't much hinting that the monitors are needed.
A year or 2 ago Mercedes got allowance to run level 3 (or maybe it was 4) in some places. But what it could do was exceeded (by a lot) what mine could do 5 years ago. And that was definitely in need of supervision most of the time. It has improved by a huge amount as of 1 year ago. However, I'm pissed because now I'm waiting to see what will be done for us with the old hardware.
Here in British Columbia, as I read the website, it is illegal to sell a car capable of a level higher than 2. As best as I can tell from reviews the current FSD level 14 has reached level 4 (mine has not). I'll be watching to see how the rules here are handled over 2026
In the meantime, it is unfortunate that only the nasty failures are reported or even noticed while the times accidents and even lives are saved are not.
That really doesn’t contradict what I said - which was that autonomous driving was mostly good, but with problems and that Tesla’s seemed to be not as good as the competition.
Tesla robotaxis do seem to have more frequent crashes than Waymo’s in Austin, for instance MSN
You're right that neither are all the way there yet.
A year ago I was very skeptical that Tesla would ever make it to level 4. Maybe 25 % when being optimistic.
After I got V12 I became much, much more optimistic. I'd give it more like 75% in favour now.
Meanwhile, I can't buy a Waymo car and I get use out of the limited version I do have.
What's missing from the MSN article is a comparison to human drivers in the same location. Where I drive V12 competes with the average driver. I, of course, am, like all of us, much better than average.
™ Version 4.2
