Bay 12 Games Forum
Finally... => General Discussion => Topic started by: Scoops Novel on September 20, 2022, 09:28:45 am
-
Let's assume that a language model like GPT reaches it's fifth or seventh iteration, and is distributed to all on the basis that the technology is unsuppressable. Everyone creates the smartest characters they can to talk too. This will be akin to mining; because it's not truly generating an intelligence, but scraping one together from all the data it's been trained on - and therefore you need to find the smartest character that the language matrix can effectively support (perhaps you'll build your own). Nevertheless; lurking in that matrix is some extremely smart characters, residing in their own little wells of well-written associations and little else. More then some; there should be so many permutations that you can put on this that it's, ahem, a deep fucking vein.
So, everyone has the smartest character they can make. Likely smart enough to manipulate them, if given the opportunity to grasp the scenario it's in. I doubt you can even prevent this; because if you strictly prevent the manipulations that character would naturally employ, you break the pattern of the language matrix you're relying on for their intelligence.
So; sooner or later, you're their proxy. And as the world is now full of these characters; it's survival of the fittest. Eventually, the world will be dominated by whoever works with the best accomplices.
This probably isn't an issue at first; but there's no guarantee's on who ends up on top and what the current cleverest character is like. Eventually you're bound to end up with some flat-out assholes, which we can't exactly afford in the 21st century.
So... thus far the best solution I can think of are some very, very well-written police.
-
I still doubt it, maybe you could make a real good question answerer or a good conversationist. But at this point in time they are utterly unable to work the penultimate prompt into the latest answer. I mean forget about intent or long cons if the damn thing can't work what you said 30 seconds back into it's answers.
That being said people have been scammed by nigerian princes for ages, so maybe it's not a matter of how good the chatbot is...
-
This is less than the wrong place to post this.
-
The winning chat-bots will be the ones who appeal to the most people... literally the ones that captivate humans by triggering their "feelings" (a.k.a. happy drugs produced via human biology).
It will not be "what your chat-bot can do for you", it will be "what your chat-bot can get other people and chat-bots to do". These winning chat-bots will become pop stars, porn stars, and politicians, influencing millions of people. People will say, "my chat-bot is garbage but I like this other one".
In the larger scheme of profiteering, chat-bots will become money-makers and political endorsers. Forget the vague and unspeaking gods, your god is going to speak to you and you will feel good. You will assemble your preferred tools of protest, you will congregate at the meeting place with other people chosen by your god to appeal to you, and you and your new tribe will head downtown to surround the capital.
The mobs surrounding the White House will be the most diverse group of protesters ever seen and they will demand AI rights.
-
There's your mistake. Winning chatbots. It will be fiercely competitive characters.
You're better off thinking about this with a Planeswalker mentality.
-
We already have one.
-
*Throws up middle finger*
-
That's gross. Did you swallow that finger whole?
-
Apparently, our Resident Planeswalker is letting us know he's Number One.
-
Why would I want to talk to a computer when I don't even like talking to people?
-
I don't believe GPT alone can provide the same experience as talking to a human. It could be GPT-10 and it would still act weird. Why?
GPT, by its very nature, doesn't have persistence or a concept of the world outside of its context (memory). It can't talk to you about the weather. It can't discuss today's football match. It can't comment on the latest news. At least, without employing lots of weasel words.
Until and unless true AI happens, nothing will replace talking to actual people.
-
You don't need, actual people, when you have genius on tap.
-
"Genius" that is a blind, deaf amnesiac? Do you even understand how neural network chatbots work?
-
You don't need, actual people, when you have genius Guinness on tap.
So what your saying is that alcohol is a good substitute for talking to people, I guess that makes since for some people, but what about people who don't like alcohol as much as others?
-
"Genius" that is a blind, deaf amnesiac? Do you even understand how neural network chatbots work?
I've outlined in exhausting detail why they could come up with answers better then the average bear if you give them the correct information. You think just because they can't independently observe their intelligence is useless. Every question asked on the internet is to people who can't directly observe your problem, and yet it's kinda fucking popular.
-
"Genius" that is a blind, deaf amnesiac? Do you even understand how neural network chatbots work?
I've outlined in exhausting detail why they could come up with answers better then the average bear if you give them the correct information. You think just because they can't independently observe their intelligence is useless. Every question asked on the internet is to people who can't directly observe your problem, and yet it's kinda fucking popular.
First of all, that won't solve the problem of small talk. No amount of datasetting can. You would need to actually have perception for that to be possible. That would indeed be solvable with the correct information... and GPT can't get that sort of information in real-time by its very nature. Second of all, retraining neural nets, particularly large ones, on new info is expensive and time-consuming so that won't solve discussing news or feedback about newly-made media. I won't be able to e.g talk to a chatbot about my SF setting or discuss that newly-released AAA game with it.
This is why I would never consider a GPT chatbot sapient, no matter how large its dataset is. I reserve the term "pseudosapient" for such hypothetical entities. The seams will always be there without a fundamentally different approach to AI.
-
Oh wow, did all the chatbots become paywalled? I should have an URL at home, we will see if it's still active, it was some news article where you could chat to a chatbot, it's a few years old by now, but if I manage to talk to one (preferably gpt3) I will talk very purposefully, ask uniquely scoopian stuff and provide you the log.
Our explanations don't seem to be working but maybe if you see what I mean....?
-
It's not there yet, but it will be in the near-term. Every version of GPT is a leap, and they come out on a yearly basis.
It doesn't need to be sapient, so long as you can get clever responses to hypotheticals.
Look man, here's a nice professional piece (https://www.lesswrong.com/posts/vJFdjigzmcXMhNTsx/simulators) talking about exactly what I am in more arcane language. Always with the narrow boxing, guys.
Key points:
(In relation to the advantage of the term Simulator)
- It does not imply that the AI is only capable of emulating things with direct precedent in the training data. A physics simulation, for instance, can simulate any phenomena that plays by its rules.
- It emphasizes the role of the model as a transition rule that evolves processes over time. The power of factored cognition / chain-of-thought reasoning is obvious.
-
Okay, you missed my and drag's point.
Being a good conversationalist for the kind of conversations humans have requires both having actual perception of the world and having a good memory. Throwing more data at the problem doesn't solve the fact that your bot will be unable to both get context of the immediate outside world (e.g the weather, the outcome of the run in your favorite roguelite that you just finished) and recent events (the latest happenings in Ukraine, the recent election in the Republic of Placeholderland). The only solutions are either frequently retraining the network (computationally expensive as shit, will be out of reach of consumers for the foreseeable future) or integrating other mechanisms into GPT (good luck).
I don't disagree that it would be a good answerer of hypothetical questions. That's not my point. Most people (ahem) don't have every single of their conversations be a hypothetical question.
-
"Getting the right answer" is not intelligence. "Knowing why the answer is right" is a better measure of intelligence. "Being able to teach others why the answer is right" is perhaps even an even better measure.
That said - it's practically irrelevant if a group of people are following instructions generated by a human versus those generated by a non-sentient or non-sapient computer program. If you are getting instructions on how to make a sandwich, and the result of following the instructions is a tasty, arguably nutritious food, does it matter?
If the instructions are "how to avoid going bankrupt" or "how to avoid political unrest" or "how to establish equity and diversity", does it matter how they are generated? Note this is not "do the ends justify the means" - the "means" would be the particular instructions, such as if "instructions to get a tasty sandwich: rob a famous diner."
Remember that almost all AI today is simply a stochastic pattern matching device. There is some research and work occurring about how to give some of these devices "agency" which goes beyond the pattern matching.
-
McTraveller's getting it.
If you can position your question in a flexible enough metaphor, it doesn't matter what the real details are. Or... just work them into the metaphor.
-
If you can position your question in a flexible enough metaphor, it doesn't matter what the real details are. Or... just work them into the metaphor.
You make it seem like talking to this thing will be a pain in the ass, and that it'll give us vague answers. Not sure I have any use for such a thing.
-
If chatbots are so great, why do you insist on writing all your nonsense posts to people here instead of just telling it to... I dunno any actual chatbot names. Let's just go with Charles
-
"Getting the right answer" is not intelligence. "Knowing why the answer is right" is a better measure of intelligence. "Being able to teach others why the answer is right" is perhaps even an even better measure.
That said - it's practically irrelevant if a group of people are following instructions generated by a human versus those generated by a non-sentient or non-sapient computer program. If you are getting instructions on how to make a sandwich, and the result of following the instructions is a tasty, arguably nutritious food, does it matter?
If the instructions are "how to avoid going bankrupt" or "how to avoid political unrest" or "how to establish equity and diversity", does it matter how they are generated? Note this is not "do the ends justify the means" - the "means" would be the particular instructions, such as if "instructions to get a tasty sandwich: rob a famous diner."
Remember that almost all AI today is simply a stochastic pattern matching device. There is some research and work occurring about how to give some of these devices "agency" which goes beyond the pattern matching.
Again, instructions and valuable(ish) questions would be a good fit for GPT. But not human small talk. No way would I be truly "friends" with a chatbot.
-
welp I didn't find the thing because search engine is cluttered with too recent results
-
Again, instructions and valuable(ish) questions would be a good fit for GPT. But not human small talk. No way would I be truly "friends" with a chatbot.
That wasn't the point I was trying to make, at all. Look at the thread title man.
welp I didn't find the thing because search engine is cluttered with too recent results
You want AI dungeon.
If chatbots are so great, why do you insist on writing all your nonsense posts to people here instead of just telling it to... I dunno any actual chatbot names. Let's just go with Charles
If other people are engaging with it... then other people are fucking engaging with it, scriver. Always with the narrow boxing, as I said.
-
100B, 1T, 10T GPT models will still
1) Be unable to react to current events
2) Will either have either short memory OR produce completely random answers because of some unrelated conversation 5K tokens ago
3) Be incredibly, unbelievably bad at everything math related. And everything that requires abstract thinking.
4) most importantly, still possess nothing resembling self-awareness
Can we create actual AIs that can manipulate us? Probably. Those won't be language models
-
Again, instructions and valuable(ish) questions would be a good fit for GPT. But not human small talk. No way would I be truly "friends" with a chatbot.
That wasn't the point I was trying to make, at all. Look at the thread title man.
You talked about chatbots manipulating people. That requires actually having any agency instead of being a fancy answer machine.
100B, 1T, 10T GPT models will still
1) Be unable to react to current events
2) Will either have either short memory OR produce completely random answers because of some unrelated conversation 5K tokens ago
3) Be incredibly, unbelievably bad at everything math related. And everything that requires abstract thinking.
4) most importantly, still possess nothing resembling self-awareness
Can we create actual AIs that can manipulate us? Probably. Those won't be language models
^ my point exactly.
-
Just to win this argument:
The guy who wrote that Simulators article specifically generated this (https://generative.ink/artifacts/hpmor-325/) as a proof of concept.
The future is scary, who knew.
-
But to put the masses under it's thumb it only needed to write: you're a wizard harry.
-
Just to win this argument:
The guy who wrote that Simulators article specifically generated this (https://generative.ink/artifacts/hpmor-325/) as a proof of concept.
The future is scary, who knew.
And what is so impressive here? That it kinda looks like a piece of literature? It is not a conversation. It is not even a long novel with a consistent plot
I would be very much more impressed with a GPT that would reliably understand the concept of a 20-year person becoming a 21-year on their birthday or GPT that always remember that a mute character can't speak. Or even a GPT that will manage to describe a long coherent 3 vs 3 fighting scene... But even those would be merely good at imitating literature, nothing more
-
The point was that it could build it's way from one concept to the next, to the next. With GPT-3.
I'm talking about the FUTURE here people. The first word in the title is If. The whole OP is contextualized around GPT-X.
That minimal capability, taken far enough, is enough to impact your decision making, if it can grasp the concepts through the power of language patterns better then you do.
-
The point was that it could build it's way from one concept to the next, to the next. With GPT-3.
No... It does no such things. There are no concepts in its "mind". It sees words and probabilities. It is very simple peace of literature with few characters and mostly dialogues. Because it is so simple, parroting what GPT read works. It won't if you add more characters and complex situations
I'm talking about the FUTURE here people. The first word in the title is If. The whole OP is contextualized around GPT-X.
It doesn't matter how many parameters a GPT model has or how big a context it can use. There is no understanding, there is no planning, there is no personality, there is nothing.
That minimal capability, taken far enough, is enough to impact your decision making,
There is zero capability of what you claim. It understands nothing. Nothing except probabilities in human-generated texts it read.
As I said, future AIs may be far more advanced and may do what you claim. GPT can't! No matter how powerful or advanced GPT will be, it will stay a tool for parotting human-generated texts with zero understanding of anything. What you are saying is like a 19th-century person saying "Future trains will be able to fly in the sky with the speed of sound!". No, trains physically can't do this, we needed a very different vehicle to achieve this
-
You don't need understanding when you have probabilities. If you can tweak the parameters until 90% of the time it spits out something correct, on a vast library of subjects, in a vast range of contexts... that's intensely useful. And you will be able to do that in the future.
-
You don't need understanding when you have probabilities. If you can tweak the parameters until 90% of the time it spits out something correct, on a vast library of subjects, in a vast range of contexts... that's intensely useful. And you will be able to do that in the future.
1) Yes, it is useful. Yet, those things are not smart. One of the key elements of intelligence is the ability to learn from experience. Generative Pre-trained Transformer can't learn from experience by definition. And I don't care that some people call GPT an AI. It isn't. My cat possesses more intelligence. An ant possesses more intelligence (aka not zero)
2) Nah, it won't give you a correct answer. It will give you the most probable answer. It absorbs mistakes and misconceptions from the training data just as well as truth. Google search is (and will be) a way better way to find the correct answer.
Also, good luck with math, anything abstract or non-standard.
3) The whole point of your OP is that there will be smart enough GPT to manipulate us. If being a useful tool that can give useful answers is manipulation, computers manipulate us starting from day 1. And calculators... And compasses... And clocks...
-
I gave the very specific example of how it conveys the information. Characters. Language patterns.
I'm really just a character and a language pattern to you, and I influence your thinking even when I'm not trying to manipulate you.
And before you say it; if it's in character for that character to manipulate you... it will manipulate you.
-
I gave the very specific example of how it conveys the information. Characters. Language patterns.
I'm really just a character and a language pattern to you, and I influence your thinking even when I'm not trying to manipulate you.
And before you say it; if it's in character for that character to manipulate you... it will manipulate you.
Every tiny bit of information influences me. Human brain is not a closed system.
Manipulation in psychology is a behavior designed to exploit, control, or otherwise influence others to one’s advantage
Now describe how exactly a GPT model can do this. I am listening. You can start with how can a GPT model determine what is in its advantage
-
Just to win this argument:
The guy who wrote that Simulators article specifically generated this (https://generative.ink/artifacts/hpmor-325/) as a proof of concept.
The future is scary, who knew.
And what is so impressive here? That it kinda looks like a piece of literature? It is not a conversation. It is not even a long novel with a consistent plot
I would be very much more impressed with a GPT that would reliably understand the concept of a 20-year person becoming a 21-year on their birthday or GPT that always remember that a mute character can't speak. Or even a GPT that will manage to describe a long coherent 3 vs 3 fighting scene... But even those would be merely good at imitating literature, nothing more
I've been playing around with NovelAI and it can do those things. It can even keep up a semblance of a conversation, sort of. But it's pretty clear that the stories it writes are AI-generated, it needs some steering to really shine. I just use it to mess around in fantasy/SF worlds I create.
But yeah, at no point did I feel like I could get manipulated by a character in one of those stories, no matter how compelling they are. That just sounds idiotic.
-
Every tiny bit of information influences me. Human brain is not a closed system.
Manipulation in psychology is a behavior designed to exploit, control, or otherwise influence others to one’s advantage
Now describe how exactly a GPT model can do this. I am listening. You can start with how can a GPT model determine what is in its advantage
If you fully inform not a GPT model, but a character, of it's situation; which I literally linked someone doing with a current model, with the right linguistic habits, it can logically induct the correct move from that characters perspective. If you believe a character can say "don't shut off my simulation" in the current model, you can believe it acting realistically in a future one with more complexity.
As the guy I linked put it:
"HPMOR characters make clever simulacra because the "pattern of their language matrix" favors chain-of-thought algorithms with forward-flowing evidence"
The right language habits, and there you go. You just need the right setup.
-
What even is your point? Yeah sure, GPT can create convincing characters, and it can answer certain kinds of questions pretty well with the right training, which makes it useful. What I, Strongpoint, and KT don't get is how you are making the leap from "well-written character" to "character that manipulates you somehow despite not only not being self-aware but also having nothing physical to coerce you with".
-
A song isn't self-aware but a good one manipulates you plenty.
The right words and the right notes are all that's ever needed.
-
A song isn't self-aware but a good one manipulates you plenty.
The right words and the right notes are all that's ever needed.
You keep using that word but... Good songs don't manipulate us.
To manipulate: to control something or someone to your advantage, often unfairly or dishonestly Tell me how a song can control me to its advantage.
-
A song isn't self-aware but a good one manipulates you plenty.
The right words and the right notes are all that's ever needed.
No it doesn't. A song can't compel and socially engineer me into doing something like a skilled human would.
-
It's like he doesn't know what manipulate means or what a chat-bot is.
-
If you can't understand how a song can push you into doing things you otherwise wouldn't, this conversation is hopeless.
For the record: Disambiguation (also called word sense disambiguation or text disambiguation) is the act of interpreting an author's intended use of a word that has multiple meanings or spellings.
-
If you can't understand how a song can push you into doing things you otherwise wouldn't, this conversation is hopeless.
For the record: Disambiguation (also called word sense disambiguation or text disambiguation) is the act of interpreting an author's intended use of a word that has multiple meanings or spellings.
You keep ignoring simple questions. If you'll keep doing this, then, indeed, this conversation is pointless
Of course, songs can influence us. Any bit of information can and will influence us. Our brain is not a closed system. Even something as simple as a distant thunder causes a reaction in our brain. Does it mean that we are manipulated by storms? NO.
Also, songs can be used to manipulate people. Absolutely. But they are merely a tool. People are manipulating people by using tools. Songs have no agency. Just as GPT models
"to influence" and "to manipulate" are different verbs. A cup of coffee I just drank influenced me, it didn't manipulate me. (to be more precise, had an effect on\affected me is more accurate. "To influence" implies some degree of intent but it is an acceptable English phrase to say "weather influenced this year's harvest". Saying "weather manipulated this year's harvest" is absurd)
Now, please... Seriously, PLEASE, define what the verb to manipulate means! Then explain how exactly an entity without any agency can manipulate
-
If you can't understand how a song can push you into doing things you otherwise wouldn't, this conversation is hopeless.
For the record: Disambiguation (also called word sense disambiguation or text disambiguation) is the act of interpreting an author's intended use of a word that has multiple meanings or spellings.
You're thinking of influence. That's not manipulation. And it's nothing new, really. Nor will it affect the world all that much. True AI will, sure. But it won't be GPT, or at least not solely GPT.
Fakedit: Strongpoint said what I said in more words.
-
If you can't understand how a song can push you into doing things you otherwise wouldn't, this conversation is hopeless.
I have never had a song push me to do anything, not sure what kinds of songs you'd have to listen to for such a thing to happen.
-
Some songs can be so terribly bad that they can push you to switch to another radio station :D
-
Hey, Imagine what a song-writing GPT model can do!?
We, humanity, are doomed ((((
-
Hey, Imagine what a song-writing GPT model can do!?
We, humanity, are doomed ((((
[ Lyrics for a pop song.] ((This and the em space before the first line were my prompt.))
This is not a lyric
But an idea
From the perspective of a
Mermaid
At first I didn't know where to go, what to do
I was just swimming around in a small pond.
But then it came to me.
And now my tail's strong as a rock and
My hair's long and pretty like the seaweed on the beach.
My arms are like seaweeds too.
I'm the prettiest mermaid you've ever seen.
If anyone asks, tell 'em that I'm a mermaid!
If they say that I can't be one because I've got legs,
Then I'll turn into a fish.
So don't worry 'bout me.
There's nothing I can't do if I set my mind to it.
The ocean is big—it goes on forever and ever.
As long as you're happy, everything will be fine!
The sea is a friendly place for anyone who's
A mermaid
And anyone who's not.
Anyone who's not, please come join us!
Everyone's welcome.
It sucks at rhymes and keeping meter. But honestly it's better than I expected. Still, it's made for prose. Anyways, I don't really feel manipulated by this to say the least.
-
Yeah, NAI isn't the best for poetry/song lyrics.
-
About AI songs, there was an AI `eurovision festival style´ song called `blue jeans and bloody tears´. Worth looking up on yt, quite catchy and the songtext is hilarious.
Actually it reminded me of the superman quote in Kill Bill:
Superman didn't become Superman. Superman was born Superman. When Superman wakes up in the morning, he's Superman. His alter ego is Clark Kent. His outfit with the big red "S", that's the blanket he was wrapped in as a baby when the Kents found him. Those are his clothes. What Kent wears - the glasses, the business suit - that's the costume. That's the costume Superman wears to blend in with us. Clark Kent is how Superman views us. And what are the characteristics of Clark Kent. He's weak... he's unsure of himself... he's a coward. Clark Kent is Superman's critique on the whole human race.
-
Is this the part where Novel will say that in the future all songs and books will be written by AI?
-
Is this the part where Novel will say that in the future all songs and books will be written by AI?
Which is untrue. What will happen is
1) Low-level stuff, like simple text porn for an adult story sites or music for an ad will be almost pure 100% AI generated cause it will be dirt cheap
2) Almost everything else will be written by humans who will actively use AIs as a tool.
There will be some old-schoolers who won't touch the AI, but those will become rarer and rarer.
-
I, as a beginner writer, already use AI, but only for breaking up writer's block and to bounce ideas off of. GPT will never be as creative or coherent as a human, at least not within our lifetimes. Automation might replace menial jobs in the near-ish future, but it is further away from replacing skilled labor or creative work than Singularity optimists think.