On the one hand, organizations are without question using LLM's well beyond what is actually necessary, and as reality kicks in they're forced to scale back accordingly. However at the same time, on intervals counted in months, we're seeing breakthroughs both in hardware and software that dramatically reduce the cost of inference.
Between corporate FOMO and the rapidly decreasing costs of actually running LLM's I'm interested to see at which side of the spectrum these two meet
The title seems misleading, and reading the article explains the reason more clearly. There's nonsense OKR's and objectives at these companies to burn as many tokens as possible. It turns out that when you make a metric out of token usage, it unsurprisingly ends up becoming extremely expensive.
Inference is affordable, and you don't need a SOTA proprietary model to get a lot of use out of this technology. While you likely will still need a human engineer for quite a while longer, I don't agree that some number of humans + an LLM is going to be (or will ever remain) more expensive than just hiring more humans.
Solving one of the most famous Erdos problems that has remained unsolved for 80 years without using tools like lean but instead a giant reasoning block is quite a lot more than "kinda nothing"
Maybe I'm misreading but that is an absurd ToS in this context. So they're telling us they have a solution to a problem, but don't trust it enough to solve it? I tend to be averse to analogies but this feels like hiring an engineering team to build a bridge, and they tell you they're not liable if the bridge fails and collapses when used to spec.
If you don't actually believe in your product's capabilities, why sell it?
Calling the technology "text auto complete" is not productive to the discussion. Less than a decade ago the idea that a computer could take a fuzzy human-readable description and turn it into executable code was science fiction, but now it's common place. As is the ability to write long form text, and be so hard to distinguish from real that placing an em dash in your text will cause an uproar on this forum. You can describe things by their fundamental functions and make many things sound elementary but I find it counter productive given the capabilities we've seen from this technology
"Pertains" is doing a lot of work in your argument, and you're using it wrong. The data about who viewed your profile pertains to you from the moment the visit happens. That's what that word means, so your first statement is false.
The other important detail is that LinkedIn already has processed this data that definitely pertains to you, whether you paid for it or not, and are trying to sell it to you. In fact, to quote the article, LinkedIn's argument for not giving it to the user is "on the grounds that protecting that data took precedence". LinkedIn isn't withholding viewer data to protect viewer privacy. We know this because they sell it. If the viewer's privacy interest were so compelling that it overrides your Article 15 right (which is what Noyb is referring to), it would also be compelling enough to prevent LinkedIn from selling that same data to Premium subscribers.
The argument being made for this specific feature (not the ones you added) is that you can't simultaneously claim the data is too privacy-sensitive to disclose under GDPR and then sell it as a product feature
Where did you read this? From what I read in the paper it appears to explicitly state that they used NVIDIA GPU's and their MegaMOE code, which is written in CUDA.
Claude's diagramming tool that they have built into their web UI is my goto for this task. It's reliable enough that I often will delegate to it first with what I need written in prose instead of using mermaid/lucid diagram
The strait of hormuz is still closed, and a new government has not been installed.
From a conventional perspective Iran is by all means "losing" the war. However, the United States and the majority of the world desperately want the strait to be opened and have so far been unsuccessful in preventing Iran from blocking it. The US is also greatly interested in regime change, which has also been unsuccessful.
Do you also require computers to grow legs when they "run"?
"Thinking" is just a term to describe a process in generative AI where you generate additional tokens in a manner similar to thinking a problem through. It's kind of a tired point to argue against the verb since it's meaning is well understood at this point
Your work seems more targeted at tracking the real world impact of the bill rather than the changes it makes to the legal code, but a feature on my roadmap is having bill data also be easily linkable to the votes of politicians so you can track the effect politicians have on the legal code per member. Do you plan to build a member tracker on top of this as well? I think it would be super cool to be able to tie news events to a track record of votes by member of congress.
I'm interested in your cubecl-wgpu patches. I've been struggling to get lower than FP32 safetensor models working on burn, did you write the patches to cubecl-wgpu to get around this restriction, to add support for GGUF files, or both?