Impressive if true. Unfortunately not open source and scarce on exact details on how it works
Edit: not sure why I just sort of expect projects to be open source or at least source available these days.
Edit: not sure why I just sort of expect projects to be open source or at least source available these days.
Why do we think all software should be free, and then think that those that don't give it away are the abnormal ones?
Free software, like open science, clearly has something going for it pragmatically. The developer hours put into it have paid for themselves magnitudes of times over. Megacorps hire people to work on free software. If you can't see the value, that's a you problem.
For now that is fiction, but so is "if all software was free". I do think though that both would lead to a faster rate of innovation in society versus one where critical information is withheld from society to pay someone's rent and food bills.
This is somewhat analogous to music or books/literature. Most composers and performers and authors make no money from people copying and sharing their works. Some pay the bills working professionally for entities who want their product enough to pay for it; some do other things in life. Some indeed give up their work on music because they can't afford to not do more gainful work. And still, neither music nor books go away as copying them gets closer to being free.
It will be interesting to see if this is the case in the long run, assuming "huge" has a positive connotation in your post, of course.
If AGI comes to pass and it winds up being a net negative for humanity, then the ethics of any practice which involves freely distributing information that can be endlessly copied for very little cost must be reevaluated.
Increasingly, I am not putting much weight in any predictions about whether this will happen in the way we think it will, or what it could possibly mean. We might as well be talking about the rapture.
Why do people return Windows laptops when they have to pay for a Windows License Activation? Because every single OEM pays for it; you don't expect to buy Windows because it is a failed B2C business model. Nobody wants it. Same goes for proprietary UNIX, and people wish it was the case for Nvidia drivers. I own CUDA hardware and lament the fact that cross-industry GPGPU died so FAANG could sell licensed AI SDKs. The only thing stopping AI from being "free" is the limitations OEMs impose on their hardware.
> that those that don't give it away are the abnormal ones?
They are. Admit it; the internet is the new normal, if your software isn't as "free" as opening a website, you're weird. If I have to pay to access your little forum, I won't use it. If I have to buy your app to see what it's like, I'll never know what you're offering. Part of what makes Nvidia's business model so successful is that they do "give away" CUDA to anyone that owns their hardware. There is no developer fee or mandatory licensing cost, it is plug-and-play with the hardware. Same goes for OpenAI, they'd have never succeeded if you had to buy "the ChatGPT App" from your App Store.
The internet echo chamber strikes again. Exactly how many people are actually doing this? Not many, and those that are all hangout together. The rest of the world just blindly goes about their day using Windows while surfing the web using Chrome. Sometimes, it's a good thing to get outside your bubble. It's a big world out there, and not everybody sees the world as you do
Paying for Windows? I think you missed my point. If your computer doesn't ship with an OS, paid or otherwise, people think it's a glitch. The average consumer will sooner return their laptop before they buy a license of Windows, create an Install Media from their old device and flash the new hardware with a purchased license. They'll get a Chromebook instead, people don't buy Windows today.
The internet has conditioned the majority of modern technology users to reject and habitually avoid non-free experiences. Ad-enabled free platforms and their pervasive success is all the evidence you need. Commercial software as it existed 20 or 30 years ago is a dead business. Free reigns supreme.
What nonsense. Go into any business and you will find every single piece of software they use is bought and paid for with bells on. The 'Free World' you speak of is only there to get you, an individual, used to using the software so that businesses are made to purchase it. In the old days we called this 'demo' or 'shareware'. Now its 'free' or 'personal' tier subscription.
Go and ask any designer if their copy of Adobe Creative Cloud, 3d studio Max, or AutoCAD is free. Any office worker if Micsrosoft Office(including Teams and Sharedpoint etc) or even google docs for business. Majority of developers are running paid versions of Jetbrains. Running an online shop? Chances are you are paying for shopify software, or something like Zoho to manage your customers and orders.
'Free' as you put it is very much only in the online individual consumer world, a very small part of the software world.
The commercial software market is more alive and expensive than it has ever been.
These all have the property which is that they are scarce physical goods or services. Software is not scarce (though of course the labor to create it is), so this is a really bad comparison.
And again I did not say it should or should not be free, I said there are engineering benefits to open source software and more and more people recognize those benefits and choose to make things free because they see the value and are willing to recognize the tradeoffs. I never said what "should" be done. "Should" is kind of a nonsense term when used in this way as it hides a lot of assumptions, so I generally do not use it, and notably did not use it in my comment. I want to point out the peculiarity in your rather strong response to a word and concept I never used. I think you are having an argument with imagined people, not a discussion with me.
And for what it is worth, I am a robotics engineer and I am designing a completely open source solar powered farming robot designed to be made in a small shop in any city in the world (see my profile), funded by a wealthy robotics entrepreneur who recognizes the value in making this technology available to people all over the world.
So I am one of those engineers making this choice, and not someone just asking for things without doing the same of my work. Everything I produce is open source, including person projects and even my personal writing.
AMD just bought company working with similar things for more than 600m.
Yes, we're not open source, however our license is very permissive. It's both in the software distribution and viewable online at https://docs.scale-lang.com/licensing/
It's open source with a long delay, but paying users get the latest updates.
Make the git repo from "today - N years" open source, where N is something like 1 or 2.
That way, students can learn on old versions, and when they grow into professionals they can pay for access to the cutting Edge builds.
Win win win win
I'm curious, for what reasons are you interested in the source code yourself?
I am the founder/editor of PLDB. So I try to do my best to help people "build the next great programming language".
We clone the git repos of over 1,000 compilers and interpreters and use cloc to determine what languages the people who are building languages are using. The people who build languages obviously are the experts, so how they go so goes the world.
We call this measurement "Foundation Score". A Foundation Score of 100 means 100 other languages uses this language somehow in their primary implementation.
It is utterly dominated by open source languages, and the disparity is only getting more extreme.
You can see for yourself here:
https://pldb.io/lists/explorer.html#columns=rank~name~id~app...
Some that might have become irrelevant have gained a second wind after going open source.
But some keep falling further behind.
I look at Mathematica, a very powerful and amazing language, and it makes me sad to see so few other language designers using it, and the reason is because its closed source. So they are not doing so hot, and that's a language from one of our world's smartest and most prolific thinkers that's been around for decades.
I don't see a way for a new language to catch on nowadays that is not open source.
We do believe in open source software and we do want to move the GPGPU market away from fully closed languages. The future is open for discussion but regardless, the status-quo at the moment is a proprietary and dominant implementation which only supports a single vendor.
> I don't see a way for a new language to catch on nowadays that is not open source.
I do note that CUDA is itself closed source -- while there's an open source implementation in the LLVM project, it is not as bleeding edge as NVIDIA's own.
And this is a good point. However, it also has a 17 year head start, and many of those years were spent developing before people realized what a huge market there was.
All it will take is one committed genius to create an open source alternative to CUDA to dethrone it.
But they would have to have some Mojo (hint hint) to pull that off.
Now should you make business decisions based on that? Probably not. But while I don't claim to be a representative sample, I am pretty sure the number of people who share my beliefs in this regard is substantially "non zero". shrug
Imagine the shift of capital if for example, Intel GPUS suddenly had the same ML software compatibility as Nvidia
On the other hand, if they want better adoption (which would drive sales of their hardware) then Intel / AMD should make a deal to release it as opensource. Closed source will make some profit, but not that much. If this thing really means that everything can run on AMD GPU cards today, then this is a game changer and is worth a lot.
https://www.amd.com/en/products/accelerators/instinct/mi300/...
Another big AMD fuckup in my opinion. Nobody is going to drop millions on these things without being able to test them out first.
First rule of sales: If you have something for sale, take my money.
"Buy now" buttons and online shopping carts are not generally how organizations looking to spend serious money on AI buy their hardware.
They have a long list of server hardware partners, and odds are you'd already have an existing relationship with one or more of them, and they'd provide a quote.
They even go one step further and show off some of their partners' solutions:
https://www.amd.com/en/graphics/servers-instinct-deep-learni...
FWIW I believe Supermicro and Exxact actually do have web-based shopping carts these days, so maybe you could skip the quotation and buy directly if you were so motivated? Seems kind of weird at this price point.
They could break the trend and offer a "buy now" button instead of offering quotes and coffee chats. It's very likely that will kickstart the software snowball with early adopters.
Nobody is going to drop millions on an unproven platform.
> Seems kind of weird at this price point.
Yeah that $234K server is too much for people to do a trial. It has 8xMI300X GPUs along with a bunch of other shit.
Give me a single MI300X GPU in PCIe form factor for $20K and I'd very seriously consider. I'm sure there are many people who would help adapt the ecosystem if they were truly available.
I know this isn't what you're looking for entirely, but my business, Hot Aisle, is working on making MI300x available for rental. Our pricing isn't too crazy given that the GPU has 192GB and one week minimum isn't too bad. We will add on-demand hourly pricing as soon as we technically can.
I'm also pushing hard on Dell and AMD to pre-purchase developer credits on our hardware, that we can then give away to people who want to "kick the tires".
I'm not looking for entry level hardware.
As far as I'm aware you can't simply buy an Nvidia B200 PCIe card over the counter, either.
You can purchase H100 and A100 PCIe cards over the counter. They're great for compiling CUDA code, testing code before you launch a multi-node job into a cluster, and for running evaluations.
AMD has nothing of the sort, and it's hurting them.
I cannot blow 250K on an SMCI server, nor do I have the electricity setup for it. I can blow 20K on a PCIe GPU and start contributing to the ecosystem, or maybe prove out an idea on one GPU before trying to raise millions from a VC to build a more cost-effective datacenter that actually works.
What are you talking about? Have you looked?
https://www.dell.com/en-us/shop/amd-mi210-300w-pcie-64gb-pas...
https://www.bitworks.io/product/amd-instinct-mi210-64gb-hbm2...