Timeline for Can it make sense for a spaceship to be crewed by many individual AI, rather than a handful with all the processing power?
Current License: CC BY-SA 4.0
48 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Aug 14, 2023 at 19:40 | answer | added | Kilisi | timeline score: 0 | |
| Aug 14, 2023 at 17:48 | comment | added | KeizerHarm | Why do AI need a "story-based" explanation for having human qualities? Once again, I never envisioned these things to have the specific quality of natural death that you are reading from my question, only physical destruction or file corruption which anybody with a computer will tell you is a legitimate concern in this day and age. The role of backups I haven't considered yet but it doesn't seem a show-stopper to me. But in general, what is the problem with this sort of setting? I'd really appreciate if you could provide some actionable feedback because I'm lost as to your problem with it. | |
| Aug 14, 2023 at 15:05 | comment | added | Vogon Poet | Much language suggests these citizens can perish. If ChatGPT got corrupted or even real-world nuked, OpenAI could pull the last backup off the shelf and re-install it on new CPUs. The term "AI" with "self-sacrifice" or "death" simply don't reconcile in real life, so their human-like qualities need a story-based explanation before worldbuilding. Having none, the responses must invent your human-AI. Your AI are already self-actualized with creativity, autonomy, adaptability, and a purpose: "To explore", "motivated to learn", "willing to die", "for a good cause". But, no backstory with the "how"? | |
| Aug 14, 2023 at 8:14 | comment | added | KeizerHarm | @VogonPoet I'm sorry I didn't use some of the terminology appropriately. I have now hopefully clarified the AGI bit. Regarding AI "death", which I only touched upon in the context of their possible self-sacrifice: I only ever thought to be as a result of their physical destruction or file corruption beyond recovery. I didn't envision them being "perishable" so I'm just confused where your assumptions on that front are coming from. | |
| Aug 14, 2023 at 7:05 | history | edited | KeizerHarm | CC BY-SA 4.0 |
added 86 characters in body
|
| Aug 13, 2023 at 19:21 | comment | added | Vogon Poet | @KeizerHarm Likewise, if your world has “AI” which can “die,” this is conceptually impossible with our current definition of Artificial Intelligence unless the engineers deliberately neglect to incorporate nonvolatile memory, and neglect to build a backup routine. This again forces responders to invent a storyline of mortal computers (which somehow look like 21st century digital worlds superficially, without basic 21st century data protections. A world of perishable sentient AI citizens can’t be built without some story element explaining their existence. You task us with the creation story. | |
| Aug 13, 2023 at 19:11 | comment | added | Vogon Poet | @KeizerHarm The current deployment of AI such as ChatGPT, Dall-E, etc. are “narrow AI” which is exactly as you describe. They are each trained on exactly one task via algorithmically discriminating between “good” and “bad” outputs after being fed samples of pre-identified “good/bad” inputs. One narrow AI such as DeepMind can “study” 10^6 games of “Go” and master it but that is all it will ever do. Your word choice “AI” is so generic that any opinion can be rationalized as an answer. If you mean AGI, then please so state. Thus, responses are writing your story, not world. | |
| Aug 13, 2023 at 7:25 | comment | added | KeizerHarm | @VogonPoet In the Matrix, separate AI exist because each mimic an individual human in order to blend in with the human residents. I was asking if, for my scenario, there can be an objective, measurable benefit to having hundreds of AI over just one or a handful. What do you mean with "Would like to see this written in a worldbuilding context"? | |
| Aug 13, 2023 at 1:24 | comment | added | Vogon Poet | VTC because I don't think it's even possible to answer a question expertly which asks "Can it make sense." In a good story, it makes sense. In a bad one, it doesn't. But those are story-based assessments. E.g., a digital world where multiple AI citizens interact was done in Matrix and hundreds of times before that. Would like to see this written in a world-building context. | |
| Aug 13, 2023 at 1:21 | review | Close votes | |||
| Aug 17, 2023 at 3:02 | |||||
| Aug 6, 2023 at 7:20 | answer | added | Dot | timeline score: 1 | |
| Aug 1, 2023 at 18:04 | comment | added | OverLordGoldDragon | @ErikHall hmm have a reference for that 8B Kelvin math? Normal CPU fetches 2GHz at ~373K (100C), so we're talking 8B/373 = x22,000,000 the temp for x1000 the Hz, that's super-quadratic scaling. | |
| Jul 24, 2023 at 18:19 | comment | added | lessthanideal | ... b) process layers will be erased and replaced if the crew judges them to be making too many mistakes (crying wolf) and this motivates them. It made sense to this reader as presented. We can assume people flying interstellar miles-long spaceships made sensible design choices. | |
| Jul 24, 2023 at 18:13 | comment | added | lessthanideal | Although it's not the same as your scenario, you might find the section starting around page 10 of "Absolution Gap" by Alastair Reynolds of interest, the ship there seems to have layers of processes starting with low level non-sentient ones to filter input from data, significant data then gets passed up a hierarchy of process layers to a point where the processes are somewhat sentient, and decide what to show the human crew. I think it's implied a) this is helpful on a large ship, as processing at a local point helps speed things up, like a dinosaur or octopus with localised brains ... | |
| Jul 23, 2023 at 10:26 | answer | added | Sascha | timeline score: 2 | |
| Jul 23, 2023 at 0:33 | comment | added | Harper - Reinstate Monica | This is already done. Some aircraft automation systems have 3 deciders, each coded by different teams in different languages on different CPUs (to reduce the chance of all 3 hitting the same bug) and they "vote" on best action. However these are generally not full AIs making fuzzy decisions. | |
| Jul 22, 2023 at 23:50 | answer | added | whereswalden | timeline score: 1 | |
| Jul 22, 2023 at 21:05 | comment | added | JBH | OK, I see what you're doing and you've done a great job of improving and clarifying your question. Thanks! I've retracted by close vote. | |
| Jul 22, 2023 at 9:32 | vote | accept | KeizerHarm | ||
| Jul 21, 2023 at 20:19 | answer | added | automaton | timeline score: 9 | |
| Jul 21, 2023 at 16:44 | comment | added | Mazura | "During ascent, maneuvering, reentry, and landing, the four PASS GPCs functioned identically to produce quadruple redundancy and would error check their results. In case of a software error that would cause erroneous reports from the four PASS GPCs, a fifth GPC ran the Backup Flight System, which used a different program and could control the Space Shuttle through ascent, orbit, and reentry, but could not support an entire mission. The five GPCs were separated in three separate bays within the mid-deck to provide redundancy in the event of a cooling fan failure." | |
| Jul 21, 2023 at 16:13 | comment | added | DKNguyen | This is highly dependent on how exactly a "single AI" is structured and defined. | |
| Jul 21, 2023 at 14:41 | comment | added | Abigail | So, you basically want a micro services architecture? | |
| Jul 21, 2023 at 11:22 | history | edited | KeizerHarm | CC BY-SA 4.0 |
Added background
|
| Jul 21, 2023 at 11:20 | answer | added | Mark Morgan Lloyd | timeline score: 3 | |
| Jul 21, 2023 at 10:54 | comment | added | KeizerHarm | And I do believe you have a misconception of my setting, perhaps due to my inadequate description. You say that I need to have a central computer: I agree. The ship is controlled by one computer. The setting of the story is inside the computer. Processes are humanised, talking-animals style, and they carry out tasks delegated to them. What I seek to justify is to have a significant amount of processes as part of that computer that can be characterised as general artificial intelligences: they have power of independent reasoning and agency to fulfill tasks by their interpretation. | |
| Jul 21, 2023 at 10:35 | comment | added | KeizerHarm | For example I am specifically not asking about droids. Robots running a robot ship is the default sci-fi approach to a computer civilisation; I think that robot bodies aren't useful for anything other than adventures outside the ship. An artificial intelligence's native environment is the server, so this comes down to questions of computer partitions and memory management, rather than the logistics of oil changes for squeaky droids. | |
| Jul 21, 2023 at 10:26 | review | Close votes | |||
| Jul 26, 2023 at 3:10 | |||||
| Jul 21, 2023 at 10:22 | comment | added | KeizerHarm | @JBH I removed that phrasing because "how many are required?" was the wrong question indeed. I want to accomplish 100s of AIs on a spaceship, and I am seeking answers that either tell me "Yes, here's how you can justify it:...", or "No, this cannot be done in any way, because:..." In other words, I am seeking to justify an approach. But I'll try to improve it, for example by clarifying what is meant by AI. In the meanwhile I do believe your latest comment is a pretty suitable answer on its own. | |
| Jul 21, 2023 at 10:16 | comment | added | JBH | OK, I noted the edit from 30 seconds ago. Neither you nor any of your respondents are thinking like engineers. Unless you want your universe to have lots of droids but no central computer AI, there's no reason to have a lot of independent AI. Ever. That's inefficient and not how any engineer thinks. Redundant resources would be located throughout the ship with the ability to direct drones for maintenance work. From a logic perspective, even an "AI" (whatever that is) wouldn't choose independence. It's easier to merge and remove the conflict by letting one resource set the priority. | |
| Jul 21, 2023 at 10:13 | history | edited | KeizerHarm | CC BY-SA 4.0 |
added 24 characters in body
|
| Jul 21, 2023 at 10:11 | comment | added | JBH | ... but so long as you're asking, "how many crewmembers are required on my ship?" the entire discussion about "AI" is meaningless. You need to tell us about your ship. Why the story-based vote? Because this question belongs in my "how many angels can dance on the head of a pin? Answer: as many as wanting" category. And you've already answered that question. BTW: you already have SciFi precedent for your goal: Star Wars. A universe that doesn't have a single central-computer AI (in cities or on ships...) but has intelligent droids running all over the place. | |
| Jul 21, 2023 at 10:09 | comment | added | JBH | VTC:Too Story-Based. I spent several hours trying to write an answer, then I realized that you set a condition that made the use of "AI" (whatever that means) irrelevant. You've already decided you want a large crew. So this question has nothing at all to do with artificial-intelligence or computers. It might have something to do with spaceships in that you need to crew an undefined vehicle. The question you might want to ask is "how can I define 'AI' in my universe to rationalize a bunch of intelligent droids but no central intelligence computers?" (*Continued*) | |
| Jul 21, 2023 at 8:22 | answer | added | Cem Kalyoncu | timeline score: 6 | |
| Jul 21, 2023 at 5:12 | comment | added | Jeremy Friesner | It's a lot less likely for a stray meteorite to put a hole through thousands of AIs scattered all around the ship, than it would be for it to perforate a small number of big AIs located in the data center. And when the ship has some fresh holes in it is exactly when you need functional AIs the most :) | |
| Jul 21, 2023 at 1:05 | answer | added | Ángel | timeline score: 10 | |
| Jul 20, 2023 at 20:34 | answer | added | Nosajimiki | timeline score: 16 | |
| Jul 20, 2023 at 17:14 | history | became hot network question | |||
| Jul 20, 2023 at 10:22 | answer | added | Stephen | timeline score: 17 | |
| Jul 20, 2023 at 10:07 | history | edited | KeizerHarm | CC BY-SA 4.0 |
added 40 characters in body
|
| Jul 20, 2023 at 10:05 | comment | added | KeizerHarm | @Rekesoft It's a ship built and crewed by exclusively AI. My story goes that although the home planet is a tightly operated society with efficiency as the highest priority, on the spaceship the AI become more like organic creatures. So they start from a place of maximum efficiency and that's what their operational structures would reflect. | |
| Jul 20, 2023 at 10:05 | answer | added | Mash | timeline score: 3 | |
| Jul 20, 2023 at 9:54 | comment | added | Rekesoft | If your AI's have personalities, I'd suggest the reason could be just Human Rights for Artificial Intelligences. Probably a Court of Law decided long ago that any AI capable of self-conscience is a person - it certainly meets the definition of one - and as should it has some inherent rights as an individual. | |
| Jul 20, 2023 at 9:43 | comment | added | ErikHall | Back of the envelope calculation; a 2 THz CPU core would draw at least 4 TW assuming a pretty remarkable efficiency. If we assume a standard CPU die area and it being made of silicone, well the CPU would reach a temperature of 8.027.735.608 Kelvin in 1 second. Or 8 Billion Kelvin. So... yeah | |
| Jul 20, 2023 at 9:22 | comment | added | KeizerHarm | @ErikHall Good thing sci-fi spaceships already travel faster than light so thermodynamics can't complain :) But still good point, I'll make that one slightly more reasonable. | |
| Jul 20, 2023 at 9:20 | answer | added | L.Dutch♦ | timeline score: 21 | |
| Jul 20, 2023 at 9:19 | comment | added | ErikHall | " 2 THz CPU " not if thermodynamics has anything to say about it. Even 7Ghz can only be achieved using liquid nitrogen cooling. And a physically larger chip cannot be this fast, because at those clock speeds, the speed of light is slow. Regardless, a 7 THz CPU would produce way to much data for any bus to handel withouth undergoing nuclear fusion. | |
| Jul 20, 2023 at 9:10 | history | asked | KeizerHarm | CC BY-SA 4.0 |