I'm someone who uses tens of millions of tokens each month, almost exclusively with open weight models that I run on my own hardware. That said you are taking the wrong approach here, this type of mentality is only going to further radicalize those who have decided they're against this technology.
Additionally when the finally bubble bursts and the executives wake up from psychosis and look to distance themselves from this because it's become a dirty word, you'll be one of the first to go. The nail that sticks out gets hammered down and all that.
I do think there are real benefits and productivity gains with this technology, but it does not benefit everyone equally. It's great for the programming parts of my job, but useless in the other 40% of the work. I have coworkers for whom generative AI has no obvious practical application, and yet management is trying to find a way to shoehorn it in anyway. No doubt because they've also drank the kool-aid and are eager to reduce headcount.
This attitude of it making everything more productive and anyone who doesn't follow will be left behind is not just false, it's cruel and myopic. You're talking about people's livelihood being taken away because a handful of executives decided this is how things should work despite the MASSIVE number of shortcomings and poor product market fit.
Edit: I also almost missed where you're seemingly celebrating the devaluation of human labor as a result of this. Please stop and reflect on how your position may read to someone who is just trying to put food on the table.
That's not really how the experts in an MoE work. They activate on token probabilities and are activated on every token. You don't necessarily have a discrete math expert and a discrete physics expert. And if it were you would still need a router that is trained on all of those domains.
MoE models are typically designed for datacenter deployment, where per-token load-balancing is more important, but it's also possible to use a different training objective that encourages domain-specialization of experts: https://allenai.org/blog/emo But yes, this isn't really useful for distributed training as such because of the router.
Something powered by an LLM is going to end up being the tool that makes this accessible in the way it always should have been, and that gives me complex feelings.
This was covered in the issue itself, in fact the issue is pretty well documented with regards to the packaging:
> Publish an official Claude Desktop build for Linux, targeting the two current Ubuntu LTS releases (and Debian) as a signed .deb via an Anthropic-operated apt repository, using the same distribution pipeline Claude Code already uses for Linux.
Also Flatpak or AppImage would make this accessible to every other distro. Alternatively you could run the deb with a Podman Toolbox.
Your point about backwards compatibility with Windows goes both ways, I have old games that I can _only_ run on Linux as they don't work on modern versions of Windows.
You're supporting the developers original work at that price. There's plenty of cheaper devices that take that original work and just throw it on some chips
I want to reduce my dependency on companies like Google, OpenAI, and Anthropic. Aside from the concerns of data sharing I'm also not a fan of how they run their operations, for example Anthropic now using xAI's Colossus data center which is poisoning a marginalized community, or OpenAI getting in bed with the military.
Not everything I want to use an LLM for requires "PhD level intelligence", and increasingly I'm finding more uses that involve sharing my personal data.
Yesterday my local model helped me when looking for a doctor who is in-network for my insurance. I threw it a screenshot from the providers search results and it looked up reviews for all of them.
My local AI is currently upscaling an old british comedy from sub-dvd quality to 1k. (It is not availible other than on DVD.) It looks like it will take about a week for my pair of 5060s to chew through the task.
I own the DVDs so I'm OK upscaling/editing my own copies for my own use. But if I ran the task on an ai service I would no doubt trigger copyright issues.
Additionally when the finally bubble bursts and the executives wake up from psychosis and look to distance themselves from this because it's become a dirty word, you'll be one of the first to go. The nail that sticks out gets hammered down and all that.
I do think there are real benefits and productivity gains with this technology, but it does not benefit everyone equally. It's great for the programming parts of my job, but useless in the other 40% of the work. I have coworkers for whom generative AI has no obvious practical application, and yet management is trying to find a way to shoehorn it in anyway. No doubt because they've also drank the kool-aid and are eager to reduce headcount.
This attitude of it making everything more productive and anyone who doesn't follow will be left behind is not just false, it's cruel and myopic. You're talking about people's livelihood being taken away because a handful of executives decided this is how things should work despite the MASSIVE number of shortcomings and poor product market fit.
Edit: I also almost missed where you're seemingly celebrating the devaluation of human labor as a result of this. Please stop and reflect on how your position may read to someone who is just trying to put food on the table.
reply