Flow of things brought me to developing a game on Unity 3D. It has in-game particles/ribbons editor, so no max needed to create those. Though, tanks made in max. And shader is jsut the same as sc2 uses.
For the last creature I didn't use its natural texture, just applied heat wave normal and white specular. To save space and time. Tried to convince myself that it looks good, but looks like it doesn't. Thanks for clarifying it, dudes.
Actually I have a little hope. If we link (in max) attachment point to the physics object, and it will work, then UnitGetAttachmentPoint() would return its coordinates in game, and it would be used for gameplay. But I will only be able to test it if anyone tell me hoe that UnitGetAttachmentPoint() function work (it didn't work for me).
Currently all the computing power in the world has less processing power than a single puny human brain. Also current electronics are starting to hit physical limitations.
If we create a model of neuron which would include all its molecules, it will indeed consume world's computational power.
But if we exclude biological stuff, and only create a model of its informational beheving, namely how it reacts on signals and how it generates signals to other neurons, it won't be a complex model - a network of 1000 of which could work in real time on an average desktop computer. So, we would need only about 90 millons desktop computers or a good supercomputer to imitate human brain. After all, it depends on the model of neuron we use.
Also, brain's way of thinking may be not the only one, and not optimal way of thinking. He had not much choice, due to notorious cellular heritage.
Was referring more to the practical side of most people have only one computer and if it dies due to a "lethal" mutation they will be pissed while nature can have multiple progeny and the weak will die and not procreate so it is an issue of available sample sizes. Actually a lot of learning programs do take weeks to months to evolve as it is purely an iterative process and rate of progress is limited by generation time as seen in organisms.
Well, for artificial evolution we curtainly wouldn't destroy computers. And yes, evolving may take months. For example, author of and vids said it took a month to evolve, even with only a few seconds required to figure out organism's quality. But in this case it's obviously too much. Depends on how well evolving model optimized. So, a person that good with both, biology and programming, can find interesting areas where artificial evolution of something is more effective than the traditional brain-driven engineering.
Anyway, this conversation looks a bit out of topic, I just stated that natural-science knowledge can be a source of programming ideas.
For example, I have saw a map somwere on this forum, where a guy used "genetic algorighms" approach to find the best counter to a group of units within a limited resources amount to spend.
Not only biological related models can help, just simple physics based models are useful.
For example, if we want to evenly seed something on a map and using random generator, objects will be spread too chaotically, but if we let them act like a set of same charges and magnetically repulse each other for some time, they will spread more even. Or if we give them larger scale repulsion and lower scale attraction, they will be gathered in a groups, for example, single trees will be organised into a set of a few forests.
In other words, self organisation and imitating modelling ftw.
We can only talk about software evolution. Problems with transistors are not problems of the idea that thinking as a process requires neurons that consume like 99% of energy jsut to support itself as a biological object, and the remaining energy is spent on calculations, and even that 1% isn't used with a great efficiency, while computers don't need to support its biological state, they were created to calculate, adn thus will do it better and better, while brains are saturated and can't be boosted. So, as far AI emerges, humanity is doomed, it will be just another step in the chain: initial bullshit -> inanimate matter; matter -> life, life -> organisms with minds. And finally AI.
And yes, nature has seeds, while computation of artifical evolution has a limit of cpu perfomance. We can have lots of "seed" of "programs", but have to run the imitation from hours to weeks to see a progress of the organisms.
And maybe this idea to evolve something not so good for a map, but sc2 may be an appropriate environment to run an experiment within it, and visualise the results. Using the result as a map AI may be fun if it won't be just super hard, but rather inventive, smart and full of fresh ideas. The only question is the experiment itself, like what should evolve and what should it do to check if it good or bad.
...Meanwhile, made a rock to try an improved texture generating approach.
Yea, and that's why AI will overrun humans: their transistors won't have the burden of biological origin. They just eat electricity and fight, no problems.
It tits fits my map's needs, and I don't understand, why should it hide its origin, there's nothing to be ashamed of. If anyone wants to change anything, i can post a max file of it.
And your and Evil's areas can be used for AI evolving, for example. AI control system would be encoded into a set of genes, which could be be replied with mutations and then selected based on game results. Though, this would require server side bank and high popularity of the map, because only thousands of generations could produce something strong.
Or AI can fight each other locally for days until a perfect one will evolve, and it will be hardcoded in the map.
I'm just not sure, what map genre would be more appropriate for interesting AI competing.
You see, even 100M years wasn't enough to modify or repace all shitty mouse features in us. No wonder that model still has carrier mouthes, I barely spent an hour on it.
Yes, further modifications would sooner or later completely remove its carrierness. Like human bodies have no signs of that shitty mouse we evolved from during last 100M years.
Rollback Post to RevisionRollBack
To post a comment, please login or register a new account.
Flow of things brought me to developing a game on Unity 3D. It has in-game particles/ribbons editor, so no max needed to create those. Though, tanks made in max. And shader is jsut the same as sc2 uses.
@Charysmatic: Go
Thanks for mentioning his name. I checked some images and added particles to make it look closer to the original look.
Applied that 4 headed dude's original textures
For the last creature I didn't use its natural texture, just applied heat wave normal and white specular. To save space and time. Tried to convince myself that it looks good, but looks like it doesn't. Thanks for clarifying it, dudes.
Recently sc2-ized some more wow models for my map:
@JacktheArcher: Go
Actually I have a little hope. If we link (in max) attachment point to the physics object, and it will work, then UnitGetAttachmentPoint() would return its coordinates in game, and it would be used for gameplay. But I will only be able to test it if anyone tell me hoe that UnitGetAttachmentPoint() function work (it didn't work for me).
Played a bit with physics.
Recently starcraftized a couple of wow models for fellow mapmakers.
btw, the game i've been making models for these months is released on EU server under Mines and Magic name
If we create a model of neuron which would include all its molecules, it will indeed consume world's computational power.
But if we exclude biological stuff, and only create a model of its informational beheving, namely how it reacts on signals and how it generates signals to other neurons, it won't be a complex model - a network of 1000 of which could work in real time on an average desktop computer. So, we would need only about 90 millons desktop computers or a good supercomputer to imitate human brain. After all, it depends on the model of neuron we use.
Also, brain's way of thinking may be not the only one, and not optimal way of thinking. He had not much choice, due to notorious cellular heritage.
Well, for artificial evolution we curtainly wouldn't destroy computers. And yes, evolving may take months. For example, author of and vids said it took a month to evolve, even with only a few seconds required to figure out organism's quality. But in this case it's obviously too much. Depends on how well evolving model optimized. So, a person that good with both, biology and programming, can find interesting areas where artificial evolution of something is more effective than the traditional brain-driven engineering.
Anyway, this conversation looks a bit out of topic, I just stated that natural-science knowledge can be a source of programming ideas.
For example, I have saw a map somwere on this forum, where a guy used "genetic algorighms" approach to find the best counter to a group of units within a limited resources amount to spend.
Not only biological related models can help, just simple physics based models are useful.
For example, if we want to evenly seed something on a map and using random generator, objects will be spread too chaotically, but if we let them act like a set of same charges and magnetically repulse each other for some time, they will spread more even. Or if we give them larger scale repulsion and lower scale attraction, they will be gathered in a groups, for example, single trees will be organised into a set of a few forests.
In other words, self organisation and imitating modelling ftw.
@DrSuperEvil: Go
We can only talk about software evolution. Problems with transistors are not problems of the idea that thinking as a process requires neurons that consume like 99% of energy jsut to support itself as a biological object, and the remaining energy is spent on calculations, and even that 1% isn't used with a great efficiency, while computers don't need to support its biological state, they were created to calculate, adn thus will do it better and better, while brains are saturated and can't be boosted. So, as far AI emerges, humanity is doomed, it will be just another step in the chain: initial bullshit -> inanimate matter; matter -> life, life -> organisms with minds. And finally AI.
And yes, nature has seeds, while computation of artifical evolution has a limit of cpu perfomance. We can have lots of "seed" of "programs", but have to run the imitation from hours to weeks to see a progress of the organisms.
And maybe this idea to evolve something not so good for a map, but sc2 may be an appropriate environment to run an experiment within it, and visualise the results. Using the result as a map AI may be fun if it won't be just super hard, but rather inventive, smart and full of fresh ideas. The only question is the experiment itself, like what should evolve and what should it do to check if it good or bad.
...Meanwhile, made a rock to try an improved texture generating approach.
@DrSuperEvil: Go
Yea, and that's why AI will overrun humans: their transistors won't have the burden of biological origin. They just eat electricity and fight, no problems.
@SoulFilcher: Go
It
titsfits my map's needs, and I don't understand, why should it hide its origin, there's nothing to be ashamed of. If anyone wants to change anything, i can post a max file of it.And your and Evil's areas can be used for AI evolving, for example. AI control system would be encoded into a set of genes, which could be be replied with mutations and then selected based on game results. Though, this would require server side bank and high popularity of the map, because only thousands of generations could produce something strong.
Or AI can fight each other locally for days until a perfect one will evolve, and it will be hardcoded in the map.
I'm just not sure, what map genre would be more appropriate for interesting AI competing.
@SoulFilcher: Go
Do I look like a force of nature?
You see, even 100M years wasn't enough to modify or repace all shitty mouse features in us. No wonder that model still has carrier mouthes, I barely spent an hour on it.
@fate1990: Go
Yes, in my map these guys are teamless creeps. But adding a team color is a question of two clicks.
@AlexO6: Go
Yes, further modifications would sooner or later completely remove its carrierness. Like human bodies have no signs of that shitty mouse we evolved from during last 100M years.