So I’ve just finished binge-reading every word of every blog post, comment, programming blog, essay, book and keynote speech which @steve has created, at least, all of which I’ve been able to access.
There are a few notable exceptions, number one is the power point presentation from around 2022
Big thanks to @danielmewes and @nephilim. You two have really helped me to get up to speed.
I feel like I’ve been right there alongside everyone for the last fourteen years and I’ve come to know so much about the project as well as all the members of the community.
I feel like it’s only fair to put this out there.
I’m going to write an article for a popular science/engineering website called hackaday.com , I think this is a great community of like-minded individuals. I’ve got way too many ideas. I could write a book with all that I’ve learned.
So my question to the community in general is; what is it about Steve’s work that you think needs to be communicated in order for newbies to get excited? I want to make sure that my article has the most impact possible, ideally enough to pay Steve’s rent and get him a sandwich.
I’ve broken it down to three broad questions, 1. Who is Steve? 2. What is Phantasia to you? And 3. Why should we care? ( I.e. What about the project makes you excited?)
So my question to the community in general is; what is it about Steve’s work that you think needs to be communicated in order for newbies to get excited?
I would say that it’s very important, especially now that LLM’s are rampant, that your readers understand the differences between Steve’s work and what people tend to call ‘AI’. I think this passage from ‘Creation’ summarizes it nicely:
“A computer simulation of an atom is not really an atom. It does not really have mass or charge, it only behaves as if it does (just as the virtual ball above behaves rather like a ball, yet clearly is not one). But if you make some molecules by combining those simulated atoms, it is not the molecules’ fault that their substrate is a sham; I think they will, in a very real sense, actually be molecules. If we then go on to make materials out of these molecules, it is not unreasonable to call these materials real. In fact, it is even fair to describe them as material things.
In terms of the hierarchy of organization that we have already postulated to exist in the natural world, we can restate this principle at any level we like. For example, if we simulate nerve cells using computer code, then they are not really nerve cells. But if we use these simulated nerve cells to build a brain and the brain thinks, it is not the brain’s fault that its constituent neurones are a sham; it will still be a brain and its thoughts will be real thoughts. If it then goes on to proclaim itself to be conscious, who are we to deny it?”
Also: Happy to be of help. I assume you went through my old forum posts of how I reconfigured Gloops in all kinds of new and interesting ways to see what will happen? I enjoyed doing that and I’m looking forward to doing the same to Phantasians 🙂
Actually I’m giving myself a bit of a break from reading and going back to studying C#. I’m going to tackle the forums and all of Steve’s quora answers in the nearish future.
That is the one thread which I know for sure that I’d like to read but the forums are a large and meandering stream of consciousness and I’m still integrating all that I’ve read thus far.
-I would say that it’s very important, especially now that LLM’s are rampant, that your readers understand the differences between Steve’s work and what people tend to call ‘AI’.
You make an excellent point, perhaps I should start with what do we mean when we say AI? Or perhaps even just when we say intelligence. I honestly feel like the artificial part cheapens the legitimacy of what Steve’s been creating.
Artificial intelligence is a good fit for an LLM because it does the job of intelligence in some way without ever being actually intelligent. Kind of like how an artificial heart does the job of an actual heart but it’s not likely to be confused with the real thing.
Maybe ABI Artificial Bioplausible Intelligence, or something similar would be a good term to use.
I’m personally most excited to do what I’ve already been doing, dissecting it and coming to a better understanding. I’ve spent the last two months analyzing code and comments in my free time while I’ve only actually run the software itself for less than 2 hours.
So my question to the community in general is; what is it about Steve’s work that you think needs to be communicated in order for newbies to get excited?
From my recruiting experience, it depends:
– creatures fans: meantioning steve
– those who want sentient AI: comparison between LLM and what steve creates (normaly LLM are a like reek staues, lookin human like, with every detail – and steves AI is the little mouse at the floor)
– those whowant human like AI (and waifus) – the biochemestry that will enable realemotions
– the LLM fanboys – instand learning abillity of steves project
– the blackmirror crowd – hat this episode is based on a real event but the main difference steve is not a quitter unlike collin
– neuro sama fans: meantioning that steve grand is the first pritish guy with a AI daufghter and vedal987 just the second.
– general AI entusiasts: a compleatly different AI aproach based on a mamalian brain that could spark a new AI wave
– general gamers, meanting how spouiled i was with the Norn NPC brain that i was unsatisifed with all other NPCs (incluing current AI ones) and hoping steves new game will solve that
But since you are making an aticle, go after the programmer and engenering crowd, there i could use that article most efficient. Alwayschoose one single target demorgraphic and aim properly, if you hit well, other audiences will follow. And always have a positive attitude, after all you are doin this to help the other person to discoversomething amazing.
Oh and i never post a link but meanting to search “frapton gurney” because then it doesn’t look like an add and i ask for initial investment, lowering the bar for those who gone here to truly join. That way hey can ‘discover’ it on their own. Besides we don’t want those who can’t even use google in here!
Oh and if you manage to explain what steve means by “cheating” in AI and why he hates it, i would be interested, since i don’t really grasp why LLM cheat and steve doesn’t. (normalyi say “steve says he hates cheating in AI, but don’t ask me what cheating in AI means…)
I think that he is referring to the fact that it’s called “intelligent” when it’s really just explicitly programmed (I count using massive labeled data sets as programming rather than learning) to give intelligent sounding answers.
LLMs fall under Steve’s category of ” The sausage machine theory of mind” where comprehension and understanding of the subject matter isn’t really necessary for them to work. You simply feed data into one end and get a result of the other end. It completely ignores the holistic nature of the mind.
If asked whether ChatGPT likes carrots, it’s not outputting an opinion based on its own experience or understanding of the world, ‘carrot’ is just a data point it uses to search for what other people are likely to say about their experiences with carrots in what amounts to a huge lookup table.
Now Phantasians will never know what carrots are like either but they could form their own opinion of something decidedly carrot-like within their simulated universe, who knows they could even be repeatedly traumatized in the presence of carrots and run in terror anytime they see one. ChatGPT will never form these kind of associations which stem from living in a real dynamic and messy world and ignoring this push and pull of external stimulus has allowed AI researchers to focus on the much simpler problem of word or image predictions.
Granted, there’s been phenomenal progress in recent years and a lot of really intelligent people trying their damndest to make it work, but the oversimplification is what I believe Steve means when he says they cheat.
As for “cheating in AI” or more appropriately “cheating in AL” (artificial life) perhaps I can nudge you in the right direction.
If you know anything about Steve and the history of “Creatures” you will undoubtably learn about the fanbase and the repositories. These websites dedicated to modding Creatures are truly what kept the IDEA of what Steve was trying to create alive. The fanbase has expanded and shrunken due to age over the decades…but thankfully has not died. Part of the fanbase was creating Norn/Grendel/Ettin/Geat genealogies that were not discovered through actual breeding…but through using various programs to “hack the gene-code”.
I had a brief chat with Steve about “natural” and “selective” breeding and “hacked” breeding. With this current iteration the genes are really complicated as compared to “Creatures” but of course the possibility always exists…because this is digital and not organic.
Basically I would say that “cheating in AL” refers to creating genes that would normally takes many many generations to occur naturally…but injecting them into the genealogy artificially. There are ‘natural’ camps that want to have the RNG calculations play out through normal progression…and other camps that have no qualms about forcing the calculations or over-riding them.
I myself am a ‘naturalist’…and Steve has told me that he leans into the ‘natural’ camp himself. He understands how many people enjoy using ‘trainers’ and other hacking tools in games. Of course the appeal to be in ‘god-mode’ could be fun…but for many people it takes away from the enjoyment of the game because there was no accomplishment…there was no risk…so the reward is just as bland as the playthrough.
I both see what you are getting at, and disagree. Well sort of disagree presuming I understand you correctly (there’s always room for interpretation.)
I see two main assumptions. Number one, gene manipulation is always cheating and therefore cheapens the experience.
Number two, this is only possible due to the digital nature of our creatures.
Certainly there are ways to use gene editing to “cheat”, for example in creatures @genesis, myself and many others were able to create genes which would produce oxygen, glucose and other vital metabolic chemicals without using realistic chemistry (e.g. nothing + nothing = glucose + oxygen).
Steve attempted to fix this with an amazing simulated chemistry based around reduction and oxidation (redox), but the permutations of the system proved to be too many and complex to resolve when he still had bigger worries about memory, navigation, and executive functioning so we ended up with a more creatures style chemistry which unfortunately will still allow this.
I’m not certain that genuine attempts to simulate biology could be viewed as cheating (creating a new ‘gills’ organ for example, with chemoreceptors and emitters which facilitate the conversion of water into oxygen.). Steve himself has often spoken of evolutionary time scales and the fact that we can’t rely on a simulated primordial soup to evolve into a simulated intelligent creature within our lifetimes and he had to stand in for evolution.
As for the possibility of gene manipulation in natural animals and plants, we have many tools at our disposal cas-9 CRISPR, gamma radiation, and embryo manipulation to name a few. The barriers to entry may seem intimidating, but for under 10000 dollars Canadian an independent scientist can purchase a usb gene sequencer and all the chemicals, enzymes and materials to play God in real life. The CRISPR kits used to come in at about 300 dollars CAD on Amazon, but then, what is the point in manipulating genes if you can’t sequence the results.
From what I’ve read, Steve views the 1:1 genotype: phenotype relationship as cheating when it is used in genetic programming (GP). This is generally where a 2d simulated world is populated by critters with simple genomes, say about 15 genes which are represented by simple letters from the alphabet each encoding a different state that an individual aspect of the physiology can be in (locomotion speed, metabolism, size, etc.). The program starts out with a completely random gene pool and evaluates through some kind of ‘fitness test’ where it evaluates the most successful, then it runs again, taking the winners genomes and making small changes. This process takes place many times and leads to often rapid improvement and specialization of the organisms.
A really, really neat implementation of such a system is the A-Life sim Bibbites. If this level of pared down cellular automata is interesting to you in the slightest, I’d highly recommend that you check it out. The creatures involved tend to evolve fairly interesting neural networks that are reminiscent of multi-layer perceptrons.
Bibbites is available for free from the developers website or for a small fee through steam if you prefer automated updates. I actually only learned of it due to my research into reverse engineering the code of Phantasia.
I fear I may have written too much… I also don’t presume to know Steve’s innermost musings, it’s just my interpretation.
Part of the fanbase was creating Norn/Grendel/Ettin/Geat genealogies that were not discovered through actual breeding…but through using various programs to “hack the gene-code”.
yeah i myself made a geat breed in c3, the shee ones using the c2 shee statue as sprite base. And the genetic editing i done gone far beyond good and evil… but natural breeding was always a part.
I think the worst i did was moddifying the c3 gene splicer to work with just one norn and run an endless loop running overnight (with a nearly empty genom to prevent common genom doubling to break my game) until i reached generation 1234 or 12345, not sure anymore?
But with natural worlds, i had to resort to ‘fix’ the imortal and fastager mutations that always poped up. And i had a line of twin creating c3 norns i could only get with natural breeding but never figured out how they managed that?
But i tought the cheating was something more fundamental inside the brain itself? Guess one does not exclude the other?
were able to create genes which would produce oxygen, glucose and other vital metabolic chemicals without using realistic chemistry (e.g. nothing + nothing = glucose + oxygen).
yeah done that, got borring really fast, i called mine lifing statues. They walked around as babys and after 10-30 minutes they stoped and never moved again. Why should they?
Even the doll norns (compleatly hollowed out genom, only left the bare minimum, the brain and positive as well as negative reactions with the environment. Even disabled the existence of most chemicals in their body) moved more than those. They where quiet active.
more creatures style chemistry which unfortunately will still allow this.
a hand from god aproach might fix that, experimemted in c3 with a auto death detection if the 3 death causing chemicals where to high/low for an external death trigger. But somehow never made it? Forgot why. Imortals and similar always have similar traits, we can never compleatly eliminate them. But we can minimize them. But that’s a problem way down the line …
the fact that we can’t rely on a simulated primordial soup to evolve into a simulated intelligent creature within our lifetimes and he had to stand in for evolution.
True, evolution is slow, but i guess selective breeding is cheating in evolution? Have a normaly extremly (!!!) agressive rodent as pet that is currently in the process of domestication. They lost their agression, next step is making them nosy and friendly. With my ducks i use a different cheat, hybridisation, this gives a lot of desired traits and i hope the increased mutation rate of hybrids gives me a pleasant suprise, but can’t count on that one. Same with the citrus plants plant for next year, hybridisation with a extremly coldhardy relative should give me ‘soon’ coldhardy kunquat, oranges, mandarins and similar.
for under 10000 dollars Canadian an independent scientist can purchase a usb gene sequencer and all the chemicals, enzymes and materials to play God in real life.
if you don’t want to work precise, i got a few ideas how you could cut the costs waaaay more. It would cost nearly nothing but your ethical compas and sense of morals…
Same with the citrus plants plant for next year, hybridisation with a extremly coldhardy relative should give me ‘soon’ coldhardy kunquat, oranges, mandarins and similar.
How cold does it get where you are in the winter? I could only ever dream of growing citrus, but I’ve had couple of peach trees growing in my yard for the last 3 years and it regularly gets to -45 degrees C or colder here in the winter. conventional wisdom says such fruit won’t grow here (although I still haven’t gotten any peaches off of them yet.)
i had a line of twin creating c3 norns i could only get with natural breeding but never figured out how they managed that?
C3 was weird! In docking station I ended up with a line of norns who were all born without legs (again, through natural breeding, never could figure out how to engineer parapalegics myself.) I built them a telepathic wheelchair that used a brain interface to allow them to move left or right at will. I even ran several wolfing runs with them and once they had the wheelchairs, they survived as well as any other norns.
How cold does it get where you are in the winter?
the citrus i use for hybridisation is resistant up to -25°C and crosses hopefully as well. Others do that as well (if you are in the US, just buy a tree by this guy https://youtube.com/@themulberries?si=vKXQTDfAdF1T9RJX )
I could buy seeds by him as well and cut down years of time, however i am here for the jurney. Besides i got acces to a great non bitter trifoliate (the coldhardy one, normaly bitter but rarely not bitter at all) and i want a specific cross between cunquat for sweet peal, australian finger lime for the texture and a comercial orange line (onix™) for color and taste.
not sure if pure trifoliate might survive -45°C? The tropocal fruit forum should know that, but in your case i would just test it, one year inside to get stronger or two. Lot’s of fertiliter and then either it survives or dies. If you want i can send you pure trifoliate seeds that taste like sour citrons, no bitternes. And the coldest temperatures they can survive is often limited by the coldest temperatures recorded in that area. I assume if a plant survives -25 °C it can handle much colder. After all it is already frozen
.
.
Oh and about the c3 norns, yeah the magic of real evolution, figuring out how it does things might take a lifetime *looks at steve and a brain*
i was thinking a bit about the background of most current AI developers (mathematics, statistics and programming) and then i figured out with they think a chinese room like an LLM would work for an AI
It is because they don’t understand the brain and therefore the mistery box LLM works like the mistery box Brain. But once someone introduces people understanding brains in itself, stuff changes and “cheating” (flaws) become vissible.
this gives me hope, because that means once the problem os understood, a shift in AI aproach might start with steves philosophy as starting point. And people like you Mr @FroggyGoofball proofe that it is in the realm of posibility, after all dping something the first time is a herculian task. But doing it a second time is “just” repeating an instruction.
Well, I’m certainly not yet able to recreate all that Steve has accomplished. I think that I’ve made it to the point that I can explain most of it however.
To be completely honest, I have not even looked into the executive decision making process yet, but I am currently working on a document to help people understand how the visual system works.
Everything about these creatures is incredibly complex, it’s difficult to grasp how interconnected everything is.
Steve said I’m obsessed, and that’s probably true, but I feel like I’m chipping away at understanding. My previous obsession was a cryptography problem, but it was too hard, after throwing everything I had at it for 18 months I was seemingly no closer to solving it than when I first started and my mental health was beginning to suffer.
I try not to think about it too often because it drives me crazy! I have a bunch of notebooks full of scribbles and algorithms which I can’t bear to get rid of.
This circle still haunts my dreams.
Anyway, I guess my point is that I may be obsessed, but this seems like a healthier obsession than some other options I have available and hopefully I can help to demystify Phantasia for others.
hopefully I can help to demystify Phantasia for others.
not really the more inunderstand, the more phascinating and mystical it gets
And that’s what i love about this
But luckily i never get obsessed like that *he said in a community of a game he played as child*
