Would be a good idea, finally we would learn what those baby sounds and chat bubbles are supposed to mean. Although I suspect they are really just talking about how to gef laid.
I'm fairly certain simglish was a deliberate choice. There was nothing stopping them from having text.
Really, these sorts of ideas feel like we're getting to the "put everything on the blockchain!" phase. "Let's spend more GPU power for creating speech for the Sims than it takes to run the Sims itself!"
Yes, it was a deliberate choice, from early on. So was the storytelling aspect that the Family Album supported by letting you take screen snapshots and write anything you want about them, then publish and share them online along with the save file. Will Wright actually mentioned both of those features in his 1996 talk to Terry Winnograd's user interface class at Stanford, which I attended (then later I went to work with him at Maxis on Dollhouse, later renamed The Sims).
Will Wright on Designing User Interfaces to Simulation Games (1996) (2023 Video Update)
What about from person to person, you talking about the information that's contained within the objects, so there can be information in another person, that you want to interact with in the environment.
Have you looked at any reasons why you would want to do that?
Will Wright:
Oh yeah, I mean, that's that's the hard problem.
I mean, simulating ants is hard enough, when you get to people there's really no hope.
There are two issues here.
You can look at this as a technology.
It's not a product right now.
And their are a few directions this could go.
I could see this becoming, let's say, a multiplayer network MUD kind of a thing.
You might have a thousand people playing SimCity from the bottom up, each person building their own house in a big multi-user space.
In which case that issue is a little less important, because most of the people are real people, and you're dealing with puppets.
As a standalone game, which is probably our our closer target, we have to deal with the problem you're bringing up, which is how do we deal with people to people?
And it's hard, I mean there's just -- I'm sure Terry can elaborate on that more than I can.
But the best thing we can do is prop up a convincing illusion.
We don't have to be doing a valid simulation of human personality.
What we have to do is we have to put up something that's ambiguous enough to where somebody can read in what they want.
Actually in this thing what I have right now are people come up and they converse, but you don't hear what they're saying, they just gesture, and sometimes they look mad, sometimes they kind of look contemplative.
It's kind of interesting how much people will read into that.
This is kind of dynamic that we've seen again and again where something happens in SimCity and they said "oh I was running my nuclear reactor near the red line, and then there was so much smoke coming out of it, this plane crashed, and because of that, this and that happened", and they'll describe this long causal chain of events that I know does not exist.
I designed the simulation, I know that there's no linkage between the power output of the power plant, the planes crashing, but they're convinced it exists.
Don Hopkins:
They're using it as a medium to tell stories about.
Will Wright:
Yeah!
Don Hopkins:
Where they're using it as a piece of paper, to write.
Will Wright:
Yeah, that's exactly right.
There's a parallel simulation going on here in the game.
Everybody's taking a linear path through this, and they're basically, most people will attempt to understand things like this with a story.
They'll think about "I did this, then that happened, because of that", and so the story becomes...
There's a new technology available and people are exploring uses for it, experimenting with dropping it into older problems and seeing what the results are.
Isn't that kind of tech exploration central to the hacker ethos? (not trying to No True Scotsman anyone who disagrees, just how I've always understood it)
Your right, I jumped the gun slightly on this, exploration is good.
The inevitable four thousand startups trying to monetize shoving a LLM into everything when the technology is excessively wasteful and unwarranted, that's going to be the bad bit.
Prompt: Write an a cappella cover of a Philip Glass song in Simlish.
ChatGPT-4:
Creating an a cappella cover of a Philip Glass song in Simlish presents an interesting challenge. Philip Glass is known for his minimalist compositions that often consist of repetitive structures, and Simlish is a nonsensical language. Since Glass's compositions are largely instrumental, I will create an a cappella adaptation inspired by his musical style while using Simlish-like sounds.
Verse 1:
Doo doo zib, doo doo zib, doo doo zib, doo doo zib,
Bada bada zoob, bada bada zoob, bada bada zoob, bada bada zoob,
Glarbni glarbni glarb, glarbni glarbni glarb, glarbni glarbni glarb, glarbni glarbni glarb,
Sul sul sul, sul sul sul, sul sul sul, sul sul sul.
Chorus:
Froodle froodle, zibba zibba,
Dag dag, drib drib, yibbs yibbs,
Slooble slooble, wib wib, dor dor,
Lerpna lerpna, gort gort, zor zor.
Verse 2:
Doo doo zib, doo doo zib, doo doo zib, doo doo zib,
Bada bada zoob, bada bada zoob, bada bada zoob, bada bada zoob,
Glarbni glarbni glarb, glarbni glarbni glarb, glarbni glarbni glarb, glarbni glarbni glarb,
Sul sul sul, sul sul sul, sul sul sul, sul sul sul.
Chorus:
Froodle froodle, zibba zibba,
Dag dag, drib drib, yibbs yibbs,
Slooble slooble, wib wib, dor dor,
Lerpna lerpna, gort gort, zor zor.
Remember, this adaptation is meant to be playful and fun, as it doesn't have any real meaning or convey any specific message. The focus here is on capturing the essence of Philip Glass's minimalist compositions using Simlish-like sounds in an a cappella context.
I asked ChatGPT-4 to write a poem in Simlish, then translate it to English.
Surprisingly the English translation rhymes!
The fact that Simlish lacks any precise meaning probably makes it a bit more tractable.
Michel Kandel's translations of Stanislaw Lem's robot poetry from Polish to English also rhymed, but he's an world class translator and also an accomplished science fiction writer and editor himself.
Here's a previous discussion of Marcin Wichary's translation of one of Lem's stories from Polish to English. He created the Lem Google Doodle, and he stalked and met Stanislaw Lem when he was a boy. Plus a discussion of Michael Kandel's translation of the poetry of the Electric Bard from The First Sally of Cyberiad:
>Translations of [Lem’s] works are difficult due to passages with elaborate word formation, alien or robotic poetry, and puns. — Wikipedia [...]
>Memoirs of a train traveller: The day I stalked Stanisław Lem [...]
>I have always wanted to ask Michael Kandel how he translated the poetry of the Electric Bard from The First Sally of Cyberiad.
>Especially the one that is "A love poem, lyrical, pastoral, and expressed in the language of pure mathematics. Tensor algebra mainly, with a little topology and higher calculus, if need be. But with feeling, you understand, and in the cybernetic spirit."
Prompt: Write a poem in Simlish.
ChatGPT-4:
While it's important to remember that Simlish is a nonsensical language without a proper grammar or vocabulary, I can still attempt to create a playful poem using some known Simlish phrases and sounds. Keep in mind that this poem will not have any real meaning:
Enjoy the playful sounds and rhythm of the poem, even though it lacks any real meaning!
Prompt: Translate that poem to English, as best you can.
ChatGPT-4:
As a language model AI, I must inform you that Simlish is a fictional language created for The Sims video game series. It does not have a direct translation to English or any real-world languages, as it is designed to mimic the rhythm and flow of speech without conveying any actual meaning. However, I can attempt to provide a creative interpretation of the poem, loosely inspired by its structure and sounds:
I'm sorry but this is about as accurate as a layman would be. Philip Glass doesn't use repetition just for repetition sake, he does it to play with the rhythm. Most of these are pointless repetitions and it's not how Glass does it at all.
(1) For example here: "Doo doo zib, doo doo zib, doo doo zib, doo doo zib"
(2) Here: "Sul sul sul, sul sul sul, sul sul sul, sul sul sul."
He'd likely go for something like: "Sul sul sul, sul sul sul, sul sul sul, sulna sulna sulna, sul sul sul, sulna sulna sulna, sulna sulna sulna, na na na, na na na, sul sul sul, sulna sulna sulna, na na na".
Source: Am a composer who composes music in minimalist style with strong influences from Philip Glass.
Thank you! I copied your constructive criticism into the GPT chat that wrote the Simlish Phillip Glass song, and asked it to have another go at it, considering what you said. Does anybody want to run this through Vocaloid? ;) Doo doo la!
ChatGPT-4:
Thank you for your valuable feedback and insights into Philip Glass's minimalist style. I appreciate your expertise and will attempt to create a more accurate a cappella cover of a Philip Glass song in Simlish, keeping your suggestions in mind:
Sul sul la, sul sul la, sul sul la, sul la sul la,
Sul la, sul sul la, sul la, sul la sul la,
La la, la la, la la, la sul sul la,
La la, la la, la sul la, la sul sul la,
Zib zib doo, zib zib doo, zib doo zib doo,
Doo zib, doo zib, doo doo zib doo, zib doo zib doo,
Zib doo, zib doo, doo doo zib, doo doo zib,
Zib doo zib, zib doo zib, zib doo zib, doo doo zib,
Sul sul na, sul sul na, sul na, sul na sul na,
Sul na, sul sul na, sul na, sul na sul na,
Na na, na na, na na, na sul sul na,
Na na, na na, na sul na, na sul sul na,
Doo doo la, doo doo la, la la doo, la la doo,
Doo la doo, doo la doo, la la doo, la la doo,
Doo la la, doo la la, la la doo, la la doo,
La doo, la doo, doo la doo, doo la doo.
Please note that this is a creative interpretation of a Philip Glass song in Simlish, and it may not fully capture the nuances of his minimalist style. However, I hope this version reflects a better understanding of the rhythm and repetition that characterizes Philip Glass's compositions.
Ok well yeah this looks more realistic as far as I can tell. Still looks like it added the kind of variety I suggested (e.g. "zib zib doo" -> "zib doo zib doo") whereas an actual artist would come up with other kinds of variety (e.g. Glass uses other kinds of "additive" techniques like "sul na, sul na doo, sul na doo re, na doo re, doo re, doo re..." etc). But that's probably nitpicking since you're free to do whatever.
EDIT: "doo doo la" -> "la la doo" looks a bit more "creative" since I don't think this is a suggestion I made (unless you added more examples to your response). So, yeah not bad.
EDIT2: Nevermind, I did use that kind of transformation in my example (1) above: "doo zib zib" -> "doo doo zib". In short, it seems to have vaguely understood that the point of repetition here is to emphasize transformations. But it should be able to come up with novel transformations like Glass does.
How do you know it’d take more compute to produce a Sims text output once every few seconds, compared with actually rendering a real time 3D game? It seems extremely likely to me that this stuff will get incorporated into games, and it’ll be pretty fun too. If you have some optimized LLMs running on the client I don’t think it’ll be more computationally expensive than rendering a game.
> How do you know it’d take more compute to produce a Sims text output once every few seconds, compared with actually rendering a real time 3D game?
I don't. I assume it would need to be constantly running to know when it wants to speak and there will be multiple actors on the screen all the time. Do we have actual estimates for how much a response costs in ChatGPT? All I know is it takes a lot of video cards to power that system.
> If you have some optimized LLMs running on the client
Do these currently exist? I was under the impression that tech to date is compute intensive if you're looking for near real time interaction.
For true Sims and LLM you would need an open source unrestricted model, Sims games simulate life so you have mean,evil,criminals sims and you also have adult stuff. ChatGPT is defective since I can give it same prompt 10 times and 2 of 10 will refuse to show it's output because it thinks it is too racist and created bad content while 8 of 10 times he did it.
So with a safe prompt there is always a chance the AI will go on a bad direction and then refuse to work, and make you pay for the tokens of his "I am sorry ....long speech"
There’s a huge middle ground between ChatGPT and an “open source unrestricted model”. OpenAI’s playground lets you specify your own system prompts for GPT4, which allows for a lot more latitude in user prompts and responses.
But if you’re looking to generate truly racist and vile stuff, yes, you probably will need a model with no training or inference filters. I’m not sure anyone’s investing in building that though.
The number of people that want to create racist content is small, the issue I had is when the API is itself racist, as I mentioned in other places, you give it safe prompt and the API generates unsafe stuff. There is already unsafe stuff in ChatGPT they only added a pathetic filter on top and made us p[ay for it\s failures.
tbh, for characters interacting with a game world, a script and triggers would suffice (provided you don't mind the conversations being banal and repetitive, but perhaps a bit less banal and repetitive than a model trained on a restricted vocabulary). Maxis could have done it, but chose not to, wisely IMO
In theory, a full blown LLM gives you a lot more variety and ability to handle novel situations, but it also gives you a lot more potential for conversational gambits that don't affect the game mechanics in the way you want them to and general weirdness (I love the article's anecdote about the Sim who thinks his neighbour Adam Smith wrote Wealth of Nations!). I'm sure someone will ultimately end up designing great LLM-driven game experiences, but I don't imagine they'll look much like The Sims.
Problem with scripts is that at the end of the day you feel like you're just exploring a state machine.
It's what I always find with open world games for example. No matter what kind of character I build or how I behave I'm only ever going to get predetermined dialogue which I could have looked up and saved myself the time.
Yes in the sense of the Sims it's probably overkill (they sold millions of copies even without a script) but it's only being used as a playground to test the ideas and see what's possible, there's no suggestion that this is particularly what the Sims itself should be.
Ez, just stick a GPT-powered "reverse-moderation" layer in front of the request. Rate the response of ChatGPT for "did it do what the user asked, or did it provide a cop-out", and if it was rated as disobedient, regenerate the response until you get something acceptable.
And why should I do this instead of Open AI, is user input is sane then they should retry a few times if their AI is racist or their filter is stupid until they give me what I asked.
imagine this issue when you are just the devloper and not the user, the user complains about this but you try and works for you, but then it fails again for user, in my case the word "monkey" might trigger ChatGPT to either create soem racist shit or it's moderation code to false flag itself.
Whatever the ChatGPT API returns is “poisoned” by OpenAI themselves. The point of the reverse moderator is to ensure that the LLM produces the kind of output you want as a developer, including things like JSON schema conformance (OpenAI might throw a human-readable “As an AI blah” message in the place where you are parsing machine-readable JSON) - the reverse moderator takes care of detecting that and retrying, with the hope that a subsequent response will be “better”.
If you want a layer to moderate what the year is seeing, you can add that as well. The point of the reverse moderator is to get GPT to do what it’s told without lying about itself, more or less.
But it makes no sense that I the OpenAI customer have to pay for the fact their product is racist.
Again:
1 I give them safe/clean prompt
2 AI returns 2 of 10 times unsafe crap that is filtered by them
3 I have to pay for my prompt, then have to catch they non deterministic response and retry again on my money
What should happen
1 customer give safe/clean prompt
2 AI response in racist/bad way
3 filter catches this , then it retries again, a few times, if the AI is still racist/bad then OpenAI automatically adds to the prompt "do not be a racist"
4 customer gets the answer
For shooters having neural network powered opponents has been done. If found this IEEE paper:
Bots trained to play like a human are more fun, Bhuman Soni; Philip Hingston
Apparently, human like opponents were considered to be more fun. This kind of thing could be integrated easily in games, as the neural network is positively minuscule compared to an LLM.
Also there is nog chance of accidents if your npc's cant talk (well, they might shoot you, but they won't heckle you).
So I'd expect this to be integrated in real games much sooner than LLM based characters.
If I was Ubisoft or EA, I would invest in trash-talking bots and then allow, out of the goodness of my heart, players to pay-to-mute. If I was Ubisoft or EA.
35 comments
[ 3.7 ms ] story [ 85.4 ms ] threadReally, these sorts of ideas feel like we're getting to the "put everything on the blockchain!" phase. "Let's spend more GPU power for creating speech for the Sims than it takes to run the Sims itself!"
Will Wright on Designing User Interfaces to Simulation Games (1996) (2023 Video Update)
https://donhopkins.medium.com/designing-user-interfaces-to-s...
Will Wright - Maxis - Interfacing to Microworlds - 1996-4-26
https://www.youtube.com/watch?v=nsxoZXaYJSk
Here's the transcript, when a student asked him about that:
https://youtu.be/nsxoZXaYJSk?t=4466
Student:
What about from person to person, you talking about the information that's contained within the objects, so there can be information in another person, that you want to interact with in the environment.
Have you looked at any reasons why you would want to do that?
Will Wright:
Oh yeah, I mean, that's that's the hard problem.
I mean, simulating ants is hard enough, when you get to people there's really no hope.
There are two issues here.
You can look at this as a technology.
It's not a product right now.
And their are a few directions this could go.
I could see this becoming, let's say, a multiplayer network MUD kind of a thing.
You might have a thousand people playing SimCity from the bottom up, each person building their own house in a big multi-user space.
In which case that issue is a little less important, because most of the people are real people, and you're dealing with puppets.
As a standalone game, which is probably our our closer target, we have to deal with the problem you're bringing up, which is how do we deal with people to people?
And it's hard, I mean there's just -- I'm sure Terry can elaborate on that more than I can.
But the best thing we can do is prop up a convincing illusion.
We don't have to be doing a valid simulation of human personality.
What we have to do is we have to put up something that's ambiguous enough to where somebody can read in what they want.
Actually in this thing what I have right now are people come up and they converse, but you don't hear what they're saying, they just gesture, and sometimes they look mad, sometimes they kind of look contemplative.
It's kind of interesting how much people will read into that.
This is kind of dynamic that we've seen again and again where something happens in SimCity and they said "oh I was running my nuclear reactor near the red line, and then there was so much smoke coming out of it, this plane crashed, and because of that, this and that happened", and they'll describe this long causal chain of events that I know does not exist.
I designed the simulation, I know that there's no linkage between the power output of the power plant, the planes crashing, but they're convinced it exists.
Don Hopkins:
They're using it as a medium to tell stories about.
Will Wright:
Yeah!
Don Hopkins:
Where they're using it as a piece of paper, to write.
Will Wright:
Yeah, that's exactly right.
There's a parallel simulation going on here in the game.
Everybody's taking a linear path through this, and they're basically, most people will attempt to understand things like this with a story.
They'll think about "I did this, then that happened, because of that", and so the story becomes...
Isn't that kind of tech exploration central to the hacker ethos? (not trying to No True Scotsman anyone who disagrees, just how I've always understood it)
The inevitable four thousand startups trying to monetize shoving a LLM into everything when the technology is excessively wasteful and unwarranted, that's going to be the bad bit.
ChatGPT-4:
Creating an a cappella cover of a Philip Glass song in Simlish presents an interesting challenge. Philip Glass is known for his minimalist compositions that often consist of repetitive structures, and Simlish is a nonsensical language. Since Glass's compositions are largely instrumental, I will create an a cappella adaptation inspired by his musical style while using Simlish-like sounds.
Remember, this adaptation is meant to be playful and fun, as it doesn't have any real meaning or convey any specific message. The focus here is on capturing the essence of Philip Glass's minimalist compositions using Simlish-like sounds in an a cappella context.Surprisingly the English translation rhymes!
The fact that Simlish lacks any precise meaning probably makes it a bit more tractable.
Michel Kandel's translations of Stanislaw Lem's robot poetry from Polish to English also rhymed, but he's an world class translator and also an accomplished science fiction writer and editor himself.
https://en.wikipedia.org/wiki/Michael_Kandel
Here's a previous discussion of Marcin Wichary's translation of one of Lem's stories from Polish to English. He created the Lem Google Doodle, and he stalked and met Stanislaw Lem when he was a boy. Plus a discussion of Michael Kandel's translation of the poetry of the Electric Bard from The First Sally of Cyberiad:
https://news.ycombinator.com/item?id=28600200
>Translations of [Lem’s] works are difficult due to passages with elaborate word formation, alien or robotic poetry, and puns. — Wikipedia [...]
>Memoirs of a train traveller: The day I stalked Stanisław Lem [...]
>I have always wanted to ask Michael Kandel how he translated the poetry of the Electric Bard from The First Sally of Cyberiad.
>Especially the one that is "A love poem, lyrical, pastoral, and expressed in the language of pure mathematics. Tensor algebra mainly, with a little topology and higher calculus, if need be. But with feeling, you understand, and in the cybernetic spirit."
Prompt: Write a poem in Simlish.
ChatGPT-4:
While it's important to remember that Simlish is a nonsensical language without a proper grammar or vocabulary, I can still attempt to create a playful poem using some known Simlish phrases and sounds. Keep in mind that this poem will not have any real meaning:
Enjoy the playful sounds and rhythm of the poem, even though it lacks any real meaning!Prompt: Translate that poem to English, as best you can.
ChatGPT-4:
As a language model AI, I must inform you that Simlish is a fictional language created for The Sims video game series. It does not have a direct translation to English or any real-world languages, as it is designed to mimic the rhythm and flow of speech without conveying any actual meaning. However, I can attempt to provide a creative interpretation of the poem, loosely inspired by its structure and sounds:
So is GPT-4 actually creative, or just terrifying?(1) For example here: "Doo doo zib, doo doo zib, doo doo zib, doo doo zib"
You'd actually do something like: "Doo doo zib, doo doo zib, doo zib, doo zib, doo doo doo zib, doo zib, doo zib, doo doo doo zib, doo zib zib, doo zib zib, doo doo zib, doo doo zib"
(2) Here: "Sul sul sul, sul sul sul, sul sul sul, sul sul sul."
He'd likely go for something like: "Sul sul sul, sul sul sul, sul sul sul, sulna sulna sulna, sul sul sul, sulna sulna sulna, sulna sulna sulna, na na na, na na na, sul sul sul, sulna sulna sulna, na na na".
Source: Am a composer who composes music in minimalist style with strong influences from Philip Glass.
ChatGPT-4:
Thank you for your valuable feedback and insights into Philip Glass's minimalist style. I appreciate your expertise and will attempt to create a more accurate a cappella cover of a Philip Glass song in Simlish, keeping your suggestions in mind:
Please note that this is a creative interpretation of a Philip Glass song in Simlish, and it may not fully capture the nuances of his minimalist style. However, I hope this version reflects a better understanding of the rhythm and repetition that characterizes Philip Glass's compositions.EDIT: "doo doo la" -> "la la doo" looks a bit more "creative" since I don't think this is a suggestion I made (unless you added more examples to your response). So, yeah not bad.
EDIT2: Nevermind, I did use that kind of transformation in my example (1) above: "doo zib zib" -> "doo doo zib". In short, it seems to have vaguely understood that the point of repetition here is to emphasize transformations. But it should be able to come up with novel transformations like Glass does.
I don't. I assume it would need to be constantly running to know when it wants to speak and there will be multiple actors on the screen all the time. Do we have actual estimates for how much a response costs in ChatGPT? All I know is it takes a lot of video cards to power that system.
> If you have some optimized LLMs running on the client
Do these currently exist? I was under the impression that tech to date is compute intensive if you're looking for near real time interaction.
So with a safe prompt there is always a chance the AI will go on a bad direction and then refuse to work, and make you pay for the tokens of his "I am sorry ....long speech"
But if you’re looking to generate truly racist and vile stuff, yes, you probably will need a model with no training or inference filters. I’m not sure anyone’s investing in building that though.
LLaMA already exists.
No it's not. Elon Musk thinks there's a huge demand for racist AI, and he's probably right, or at least alt-right.
Elon Musk yearns for AI devs to build 'anti-woke' rival ChatGPT bot:
https://www.theregister.com/2023/03/06/ai_roundup/
In theory, a full blown LLM gives you a lot more variety and ability to handle novel situations, but it also gives you a lot more potential for conversational gambits that don't affect the game mechanics in the way you want them to and general weirdness (I love the article's anecdote about the Sim who thinks his neighbour Adam Smith wrote Wealth of Nations!). I'm sure someone will ultimately end up designing great LLM-driven game experiences, but I don't imagine they'll look much like The Sims.
It's what I always find with open world games for example. No matter what kind of character I build or how I behave I'm only ever going to get predetermined dialogue which I could have looked up and saved myself the time.
Yes in the sense of the Sims it's probably overkill (they sold millions of copies even without a script) but it's only being used as a playground to test the ideas and see what's possible, there's no suggestion that this is particularly what the Sims itself should be.
imagine this issue when you are just the devloper and not the user, the user complains about this but you try and works for you, but then it fails again for user, in my case the word "monkey" might trigger ChatGPT to either create soem racist shit or it's moderation code to false flag itself.
If you want a layer to moderate what the year is seeing, you can add that as well. The point of the reverse moderator is to get GPT to do what it’s told without lying about itself, more or less.
Again: 1 I give them safe/clean prompt 2 AI returns 2 of 10 times unsafe crap that is filtered by them 3 I have to pay for my prompt, then have to catch they non deterministic response and retry again on my money
What should happen
1 customer give safe/clean prompt 2 AI response in racist/bad way 3 filter catches this , then it retries again, a few times, if the AI is still racist/bad then OpenAI automatically adds to the prompt "do not be a racist" 4 customer gets the answer
Bots trained to play like a human are more fun, Bhuman Soni; Philip Hingston
Apparently, human like opponents were considered to be more fun. This kind of thing could be integrated easily in games, as the neural network is positively minuscule compared to an LLM.
Also there is nog chance of accidents if your npc's cant talk (well, they might shoot you, but they won't heckle you).
So I'd expect this to be integrated in real games much sooner than LLM based characters.