@TheStalwart It's about size. Models will be the same construction, but use more data input. More data = more gb of file size, but same reasoning. Some of the models are 450gb/1tb which would require that amount of RAM. That's why people are going to specialized models.
Current AI offerings are brutal. Claude is 50% error prone and has gotten incredibly lazy in the past few months. I spend 20% of my time convincing Claude to do work, besides the constant "status updates" every 5 minutes.
@amasad Any chance of us former-Enterprisers to get access to build some low density apps? I'm a UX'er that wants to prototype/build/fine tune. In exchange I give you access to 100% of telemetry and logs/history? I'll also be involved in any UX studies you might do.
@jack The first time someone gets spammed with an advertisement or something unsavory, it will be muted. Thoughts about user dependent filtering? Dare I say a small fine tuned model? or LoRA? be interested to read some of the user stories you have going.
Google released Gemini ... I'm blown away with how seemless it is to let it work with your file system. Actual agents. Now I just need a way to elevate permissions.. being stuck in a project folder is cramped, even for our new AI overlords.
@CodeAverage @Workday Well, if they are anything like this (which I found with a Google search) I would expect some compuserve email addresses to be on the list.
@svpino I've been designing software for over 20 years. I've seen Scrum reign in some self-aggrandizing developers and I've seen it ruin some projects. The key is knowing its a toolbox and you don't need every tool for every job. Scrum might do you some good, ;)
The #AI learned 3D scene representation, but it wasn't taught by people. Surface Statistics: Scene Representations in a Latent Diffusion Model. https://t.co/In8IFlZARU #MachineLearning