Aside from the refreshing fact of this not being yet-another-perlin-noise video, the author makes some good point about feasibility of corrosion algorithms and how those can be heavy and don't work well on a "just in time" generated map.
The second solution in particular is fascinating, although it doesn't offer a good solution to generate a seamless heightmap in the current state. I guess that combining that with some Perlin noise to determine which areas get to have starting seed points for the ridges would work?
Thoughts?
Special mention to the remarkable work of graphics and rendering that is behind the video. Some of those 1-second transitions imply a considerable amount of custom code, from the corrosion examples to the various overlaying heightmaps.
I'm very much torn between Plasticity and MoI for my next sculpting project. I've always used Blender but for some stuff I'm finding myself in need of some proper NURBS tool. I've spent a few hours on Plasticity, I'll do the same with MOI and see how it goes! :)
Currently the only thing that is stopping me from an insta-buy is that the export quality seems to be pretty low poly, even when the detail slider is all the way to 1.
Of course my trial version is nowhere near there, but watching several videos it appears that even the highest settings still are far from a proper high-poly export.
Since I work with 3D prints, that is a blocking issue for me, BUT the videos are all several months old.
Anyone with a more recent experience with Plasticity has something fresher information about that?
This is a good question - not only for the actual ethics of the training, but for the future of AI use for art. It's both gonna damage the livelyhood of many artists (me included, probably) but also make it accessibly to many more people. As long as the training dataset is ethical, I think fighting it is hard and pointless.
"Using open source code is a skill: knowing how to navigate repos and someone else’s code, understanding how to troubleshoot and navigate communities to get help, discerning between quality projects and junk… this experience is a hard-won component of being a modern software explorer. It can take you further than you might realize, past mere bits and into the land of electrons and atoms."
Very wise words. Coming from sw/hw industries I probably could work around heat pump microcontrollers without too much hassle and I well know the pain of physical components messing up your debug process.
But such industries rarely rely on open source, and all the OSS I used was for personal projects. That is definitely a big limit for my future work opportunities! :/
I wonder what kind of interesting applications could be done on e-ink harnessing the advantages of the long battery life and not suffering by the slow refresh rate. Sure, porting doom or implementing a terminal is an interesting and probably challenging feat, but still it's applications that don't shine on an e-ink device.
Maybe some "slow" strategy game, that updates upon certain events but might remain unmodified for hours at a time?
Or - more in general - an application that is required to be on for a long time but really doesnt' change often.
I think eInk price labels are now getting to be quite a common thing in some stores - while it's probably not fair/legal to change prices while the store is open, it means prices can be updated very easily overnight.
A traditional roguelike, in the line of TGGW/Cogmind/Nethack/Brogue/DCSS, would probably be nice. Not really a "slow strategy game", granted, but the fact that animations/colors aren't necessary makes it a good fit IMO
Well put.
Big fan of the "Commercial illustrators will keep their jobs, but will mostly need to learn to use AI as a part of their workflow to maintain a higher pace of work" part.
I'm a sometimes-illustrator (but my style is pretty far from what Generative AI is doing), and I recently published a 1.1 of a game manual which uses Midjourney images. I'm currently investing in a "proper" illustrator because the MDJ images lack character, but it's also true that in a few months from now this might change: I'll stick with the illustrator to have more consistency in the images, but probably the AI could do a fancier job there.
Besides, the "things will change in 2 months" point is a good one, but it's been used since a year and a half and things haven't changed yet. Sure, the quality of the produced images improved, but not in a qualitative scale.
> I'm a sometimes-illustrator (but my style is pretty far from what Generative AI is doing)
Why not train your own personal AI on your artwork? Corridor Digital did this in the latest attempt to automatise animation, they hired an illustrator to create an animation style for them, then trained the AI on their drawings.
I've actually done it [0], I'd like to have an AI assistant that I could directly use the results from, and the results were really terrible, mostly laughably terrible. I think it was too far from what the models handled correctly at the time, and given that issue it was not enough training images. Although I had also tried with a model that was better at handling stylised 2D. I'd like it to work, but I don't think it's viable for most people.
Ethics of the use of generative AI in the first place aside, I'm pretty sure the illustrator was aware of what they were intending to do with their work (they even were interviewed about it in the behind the scenes video)
I view this in the same way I view the use of an actor's voice for ai generations. Even if the person knows what you're doing with their data, it still feels really scummy and unethical. The idea that we can sample someone else's labor and be able to own that and generate shit from it in perpetuity (probably without paying them) feels very alienating.
This could have been all with consent and adjusted payments. AI does not just replace an artist, it can also speed up the work tremendously. It gives new possibilities using volume.
I'm not in illustration, but isn't it already common to hire someone to create a "style book" of what it should look like, and then have other illustrators follow that? eg, I recall animated shows working that way.
That's an interesting take!
Currently I see two reasons why I wouldn't do that:
1 - Since I'm either working for game companies or for my own project (https://fsd-wargame.com/) using AI-generated things is kinda damaging in terms of marketing. You never know when some uproar could arise against a project/game solely based on more or less petty outcries against AI. I generally sympathize with artists, but sometimes it's just whiny.
2 - My illustrations are line-art and cartography (https://www.artstation.com/thelazyone) , which are not the easiest to handle with AI. I'm sure that with enough effort there's gonna be a good model, but I haven't seen any so far.
The question is, since commercial illustrators can be more efficient using AI, will the total number of jobs in the space lower, or will the expectation for commercial illustration increase, thus increasing the workload and keeping the number of jobs the same.
In all of human history, work has always increased. This is akin to Parkinson's Law, where work expands to fill the time (and now resources) available.
I don't disagree, but concerning particular trades this is not true. In the mid-19th century there were more than seven thousand blacksmith shops in the US, which employed over fifteen thousand people, but today there are fewer than one thousand professional blacksmiths. Many of the products they produced either have lower demand or are produced by other means. If you consider the entire metalworking industry, we have many more total workers, but very few have the skills of a blacksmith.
The number of people who do the current work of an illustrator might go down eventually due to AI, but there will likely be more total people employed in the process of producing illustrations. It is just likely that fewer of them will have the skills that today's illustrators need, and also likely that fewer of them will command extraordinary wages. Many of the jobs that replace it will likely be closer to the median wage than today.
Also we will eventually turn the corner and start having population decline. For the US this might be just a few decades away. And some time after that, work would eventually decrease.
Work has always increased, but work in a specific profession doesn't necessarily increase. There are certainly fewer phone switchboard operators today than there were 100 years ago.
Indeed, but that just means that humans will have to find new jobs, not that jobs will become obsolete. How well they will find new jobs, though, is another story, based on socio-politico-economic conditions of the country they reside in.
In most of human history, the type of jobs available were relatively stable century to century; today, the types of jobs aren't even stable decade to decade.
The automation of physical labor let us turn to intellectual labor and creative labor. The coming automation of intellectual and creative labor is not like the previous automations of physical labor, because it leaves human jobs no where else to turn to.
CGP Grey's "Humans Need Not Apply" video[1,2] covered this almost a decade ago:
> Imagine a pair of horses in the early 1900s talking about technology. One worries all these new mechanical muscles will make horses unnecessary.
> The other reminds him that everything so far has made their lives easier -- remember all that farm work? Remember running coast-to-coast delivering mail? Remember riding into battle? All terrible. These city jobs are pretty cushy -- and with so many humans in the cities there are more jobs for horses than ever.
> Even if this car thingy takes off you might say, there will be new jobs for horses we can't imagine.
> But you, dear viewer, from beyond 2000 know what happened -- there are still working horses, but nothing like before. The horse population peaked in 1915 -- from that point on it was nothing but down.
> There isn’t a rule of economics that says better technology makes more, better jobs for horses. It sounds shockingly dumb to even say that out loud, but swap horses for humans and suddenly people think it sounds about right.
That does not say anything about how much work exists in aggregate. The human population has gone up, so it can be simultaneously be true that the amount of work being done increases even as each worker works fewer hours. As well, this also says nothing about the quality of work, as GDP is going up, so it can also be simultaneously true that the quality of work increases even as each worker works fewer hours.
I think the relevant metric is the amount of work per people, not the agregate amount of work. If tomorrow there's twice as many people but only one more job because of AI, then sure! the agregate amount of work has increased.
> In all of human history, work has always increased.
Production has increased. It's not clear that work has increased.
Mills and factories used to employ people by the hundreds of thousands and maintain people in a blue-collar standard of living. Now, no manufacturer even exists in the top 25 employers in the US--it's all service industry.
The vast majority of the decendants of the people working those manufacturing jobs are not working in better jobs than those were.
I think it's a fair question. I guess the answer is "because we have superconductors, and no material with 'very small resistance' instead".
Also note that while there is zero resistance you still have parasitic currents and general interaction with the rest of the environment. So no perpetual motion for us today!
Even a 'very small resistance' would likely be prohibitive. Pure silver for instance is an excellent conductor, but when you're talking about the last little bits it's orders of magnitude difference between that and a superconductor, and those orders of magnitude difference in resistance translate into orders of magnitude more current. So even a small resistance would cause your MRI machine to have a resolution so low as to be unusable.
MR imaging is possible with permanent magnets and resistive (copper wire wound) magnets at low field strengths, below 0.35T or so. Above that the heating of the magnet windings becomes excessive and it would be very difficult to maintain a stable enough field strength.
Superconducting magnets are very nice as long as there is no quenching. The material used for conductors must be mechanically stable and perform consistently from one production batch to another. One reason why current high tc superconductors are not popular...
The material used for conductors must be mechanically stable and perform consistently from one production batch to another. One reason why current high tc superconductors are not popular
I was going to ask: Why don't we use the current high temperature superconductors, and start building grid interconnects? I guess part of the answer lies in the cost that would be incurred because of the mechanical properties of the existing superconductors.
Because it is easier to keep a localized thing cool than something that is 100's of km long. That's why iceboxes are boxes and not ice ribbons and why most applications of superconductors right now are using them as coils for magnetic field and volume economy. For instance there is one grid component that uses superconducting coils as inertia free stabilizers, they can be used to source and sink current very rapidly to absorb transients in the load. This allows older and less stable grids to be used to transfer wind power because the power over time is a bit lower. Allowing those peaks unfiltered onto the grid would cause parts of it to go down.
Because it is easier to keep a localized thing cool than something that is 100's of km long.
If we simply decided to make this kind of thing a priority, we could probably manufacture suspension components at scale. (Or create small tunnel boring machines and bury them?) We wouldn't need to replace all of the lines. We'd just need enough interconnects to make transferring more power economical.
The current best "High temperature", standard pressure superconductor only works at -150ish degrees Celsius. That's not as bad as cryogenic superconductors, but it is still a blocker for large scale use.
I think that the initial question was "why a 0 resistance compared to a ridicously low resistance". And my point is that it's easier to get a superconductor than some material with "ridicously low" resistance. As you said, silver is unusable for potent magnets, and such is any other non-superconducting magnets.
Probably if we had materials with a billionth of the resistance of silver they would work, but we haven't. And we have superconductors, luckly. :)
Sounds like a logical step forward from current debuggers!
I had some colleagues who bragged of not needing debuggers for their workflow... Well, sometimes I really do.
It does look very nice, but as with all fancy debugger features I wonder if I'd spend more of my life debugging the debugger when it doesn't debug than it'd save me.
I also think that while a debugger is a great and useful tool, being able to debug without one is a requirement, because sometimes you don't have one (or it stopped working properly - see recent Android Studio releases) and if one can't do it just using logs one is really stuck. So it's worth getting practice at that.
The second solution in particular is fascinating, although it doesn't offer a good solution to generate a seamless heightmap in the current state. I guess that combining that with some Perlin noise to determine which areas get to have starting seed points for the ridges would work?
Thoughts?
Special mention to the remarkable work of graphics and rendering that is behind the video. Some of those 1-second transitions imply a considerable amount of custom code, from the corrosion examples to the various overlaying heightmaps.