* AI without CUDA & Nvidia: Someone to help me get Llama 2 inference running on stacks without CUDA, and without Nvidia GPUs. See also >>36955392
* AI evals/benchmarking: Someone to get oai-evals (OpenAI's evals repo) running on a few different open and closed LLMs and also to have different LLMs play the evaluator (for the model-graded evals), as a python script that'll make a nice markdown formatted table or similar of results
* GPU cloud prices and availability: Someone to extend ComputeWatch (https://gpus.llm-utils.org/tracking-h100-and-a100-gpu-cloud-...) so that it covers all gpu clouds, and all gpus. But for interface, instead of the website, just a github repo with hourly automatic updates for now
* Comparing "ChatGPT for your data" tools: Someone to use all these different "ChatGPT for your data" things (https://llm-utils.org/List+of+tools+for+making+a+%22ChatGPT+...) and write up a comparison of them
* Reproducing and evaluating what's "state of the art" in different AI image/video fields: Someone to make a curation of "state of the art" in AI for images and video - e.g. for lip syncing what's best, for content aware fill what's best, for text to singing what's best, etc. For now just for things that are research based, maybe in future covering products too. Emphasizing papers that also have workable and reproducible code - you'd reproduce their results as part of the eval. Would be a github repo
* LLM powered trading/investing: Someone to explore using LLMs for stock and crypto trading and write up interesting blog posts about the process and the results. Not expecting much, and it can be naive implementations, that's fine. For fun, not profit.
* AI photo + voice + text clone: Someone to make a "clone me" demo - take or upload some photos, chat with it, read out an audio script, then we use existing APIs and it makes a clone of you that you can chat with. For now, a python script that I can run locally, then once that works, a demo website
* Customer service / chat automation demo: Using Llama 2 70B specifically. Python script. Feed it a few pages of example conversations, and then it does a few-shot (25-shot perhaps, or more, but not fine-tuning) prompt and simulates both sides of the chat.
* Startup studios: Someone to go through my giant giant list of startup studios, organize it, and help me do some qualitative (not quantitative) analysis
* Written inspiring future visions: Writer to help make visions of what the future could look like with AI - emphasis on inspiring and beautiful and optimistic, but also with a focus on challenges but from the perspective of what solutions end up emerging to those challenges.
* Evaluate LLM tools for devs: Someone to evaluate all the different LLM-powered tools aimed at software engineers and compare them all of what works best for daily usage
If you've already posted in this thread as a freelancer, but you see something above you're interested in, please email me. I've not reviewed most of the comments in this thread.
Have built some fun other tools with the help of freelancers, such as >>36710370 and https://computewatch.llm-utils.org/. I also wrote this post that was submitted and on the homepage today: >>36951872 .
Please email me clay@studionumberzero.com, please make your email 5 sentences/bullets or fewer. Please include links to 3 live things you've built/written, preferably that I can play with, public github repos if not possible.