Comic Discussion > QUESTIONABLE CONTENT
WCDT Strips 3461-3465 (17-21 April 2017)
TheEvilDog:
--- Quote from: Is it cold in here? on 20 Apr 2017, 13:03 ---
--- Quote from: Bubbles ---You cannot modify individual memories
--- End quote ---
Her memory is known to work differently from one of today's file systems, then.
--- End quote ---
Indeed.
When Emily is sent in to try and recover those memories, its outright stated by AIbino that they created an interface to allow Emily to actually perceive access to Bubbles' memories. We have no point of reference to how memories would be stored in an AI, meaning that the AIbino needed to create a filter to make it easier for Emily (and by extension, us) to navigate.
Would a large cylindrical shape, filled with doors be easier to get through than, say, an infinitely complex web of constructed connections numbering in the billions?
Also, I just can't see AI memories being recorded like:
3/4_4:55pm_repaired_punchbot_again.mem
4/4_1:25pm_faye_gas_station_burrito_lunch.mem
4/4_1:30pm_faye_regretting_gas_station_burrito.mem
Case:
--- Quote from: Is it cold in here? on 20 Apr 2017, 13:03 ---
--- Quote from: Bubbles ---You cannot modify individual memories
--- End quote ---
Her memory is known to work differently from one of today's file systems, then.
--- End quote ---
At least not all of it - part of it may well be capable of storing long strings of data. But humans can do that, too - we even train capabilities like that, e.g. memorizing poems in school, or 'cram learning'.
Remember that files & folders are abstractions of human external memory storage - they complement our memory, they don't replace it. Just as conventional computers are meant to complement symbolic manipulation abilities and logic abilities that we are capable of, but find it hard to execute flawlessly, every time (*).
I'd go farther and ask: Why would AI's find it even useful to be able to recall files from internal memory when they can wirelessly access human-readable external storage? There's no indication that QC-verse-AI can download learned capabilities - i.e. the result of the processing of data - in a Matrix-y way ("I need to be able to fly a Bell 212 helicopter ..."). It seems they have to 'work through' information in order to 'make it their own', just as we have to.
So why would they clutter their internal storage with raw, unprocessed data?
(*) That was actually at least as important a reason for their invention as sheer speed - the most famous incident being the Royal Navy's competition that led to Babagge's difference engine. Turns out that no matter how many humans you throw at the problem of calculating logarithm tables for nautical navigation, no matter how many rows of human 'error-checking computers' you add, there's no guarantee that they'll not make a mistake that'll sink a ship later on. Which actually happened.
Mehre:
Yes i do have CS-background, specializing in neural networks, currently finishing degree and it looks like i will go on to postgrad.
--- Quote from: Case on 20 Apr 2017, 12:35 ---What makes you think that QC-verse AI don't incorporate building blocks that function like neural networks? ...
--- End quote ---
Nothing in the comic i remember suggested it really, although it may be in some form. As im no physicist, it sounded to me as technobabble when i read it and didnt really paid attention to it. I guess it could be some form of implementing neural networks, however as implementation details change so does many other things. Typical nn's in computers and in biology are so vastly different, that some things just doesnt translate (and there are many things we do not know about biological nn's). If this was some yet different form of concept it may not have same characteristics. In short, better safe than sorry.
--- Quote ---Could you elaborate on the meaning of those two terms? And how they relate to memory? From what I've found, the term 'subsymbolic' relates to processing, not necessary storage (though the latter is a part of the former, to some degree):
--- End quote ---
Its just matter of working with information. In past people tried symbolic approach(i.e knowledge systems, math logic and such) and while they made great advances, they found out it has its weaknesses. Since then it always kinda oscillated, currently subsymbolic is on top, but that can change. In neural networks, information is distributed over many neurons and vice versa, thus creating something which resembles chaos even if it is not. I hope it makes things clearer, although i doubt it. I will try to find some good source.
Yeah, we are capable of both. Why? Im not sure, neuroscientist would be better suited for that, but if i would have to guess we kinda emulate it. Or it is matter of abstraction/scope, similar way to how physics theories can be so fundamentally different at different scopes. Turing machines(not our class of computets), RAM machines (our class of computers) and neural networks are proven to be equivalent in types of problems they can solve and they can emulate each other. Thus i would say that parts of our mind emulate symbolic approach. And so we get to your last question, no it is not divide between machine/human divide, because there isnt any divide to speak of. We can do both, so can our computer and any other computational model at par with these three.
--- Quote ---..i can memorize notebook..
--- End quote ---
What i meant by those notebooks in mind is that this hypothetical AI doesnt need to remember it cause she has it on her hard drive. Its not part of her model, but she can read it at will. What you are describing are various types of our human memory. Imagine Siri or some assistant like that, on your phone, hovewer advanced you like. She doesnt need to remember your itinerary because she can access your files at will, at the same time those files are not part of her being/model/software/stuff. Thats not how our memory works, even if it can do same stuff.
Edit: as to why would AI use such file based memory...look at Pintsize :D
Kugai:
Jeph once said that such Human/AI relationships do exist, but he would probably never explore it 'In Comic'. (I'm paraphrasing here, so excuse me if I'm not being entirely accurate). It may or may not be that he's decided to do so.
Still, they do look cute together, even if in the long run they wind up as close as sisters.
Case:
--- Quote from: Mehre on 20 Apr 2017, 14:41 ---Edit: as to why would AI use such file based memory...look at Pintsize :D
--- End quote ---
I find it slightly troubling how easily modern CS-applications are justified with the "Because Pron!"-argument ...
(click to show/hide)Naw, just kiddn' :-D
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version