@n_dimension@pmdj@Mustardfacial@GossiTheDog Fax had legit usecases that were quickly realized, AI is still groping in the dark to throw itself at ANY use case and claim it as an AI success.
I had IDENTICAL conversations at the dawn of the internet. Before Spotify, before Netflix, before google, before Facebook, before online gaming. Useless, all of that Internet is useless, what can you use it for that you can't use a phone or a teletype for?
You folks seem to expect ready accessible, infrastructure complete, end user offerings, two years into pop-culure release. We are at the "Wow really? You can make international connections without incurring international phone call charges?" Stage.
AI is here to stay (right before it facilitates human extinction). You can either throws rocks at it and be the "there is nothing of value in this newfangled internet thing" person or you can try to learn how to use the tech. I am puzzled by the allegedly tech literate folks not seeing the emergent potential. The Luddites I can understand.
The only way to stop a bad guy with AI Is a good guy with #AI
@CommieGIR@Mustardfacial@GossiTheDog Cynics might say it’s worthless even if you do use AI because there’s virtually no software that uses it. 🤪 Supposedly the background blur in MS Teams will use it, and so will the video speech recognition/live translation feature that’s apparently in the latest version of Windows 11.
@CommieGIR@Mustardfacial@GossiTheDog More or less! I don’t know if it can be used for anything other than neural networks, e.g. general matrix operations. To be fair I’ve also not looked into what APIs you use to program these things… it’s certainly nothing as general purpose as OpenCL though. On the plus side, the specialisation means that it’s fairly compact and doesn’t wast all that much die area.
@CommieGIR@Mustardfacial@GossiTheDog They aren’t GPU cores, there are chunky but (supposedly) efficient purpose-built neural processor subsystems on the Ryzen AI and Intel Core Ultra 200V SoCs. Impetus seems to mostly be Microsoft’s “Copilot+” branding which requires a 40TOPS+ NPU. Plus, keeping up with Qualcomm. There is very little software that actually uses it though. (The GPU is probably faster but consumes more energy for the same job.)
@GossiTheDog I do find it interesting that both AMD and Intel haven't announced that they're going to start making AI-focused GPU's. Yes, nVidia has been the leader for raw GPU compute for a decade or more, but both AMD and Intel compete (somewhat) in the GPU market.
Or even that no China-based chip manufacturer hasn't stepped into the ring either.
@goatsarah@GossiTheDog does a local deepseek phone home? (Have you watched network traffic while it runs?) Will a local copy tell you about Tank Man or is it censored like the official chat?