Oh yeah I was thinking more along the lines of video games or movies where there are too many people creating it. For books, etc you can definitely donate.
There should be a way to pay only the workers when you buy something. In that case, you could pay them but only after pirating and making sure you enjoy it. Since there is nothing like that, I think you should pay only content from small creators. Big creators already have plenty, and paying for anything else just gives money to greedy executives who then lower the quality of the content to make more money. Of course, if you have the means and don’t pay anything, you are just making sure there will be less of that content made in the future. It isn’t scalable; if everybody pirated content without paying a single cent, there would be no content made except by hobbyists who don’t want to make a living out of it.
This has already been mentioned a few times. Just find the relevant issue on GitHub and give it an upvote.
I knew I recognized him from somewhere. He was the developer of lib.reviews. https://github.com/eloquence
I believe Claude 2 is the best LLM option currently, if you live in the US or UK or have a VPN.
I’m still waiting for a local autonomous AI agent with search. I don’t understand why most autonomous agent projects use GPT-4 without incorporating search capabilities. Allowing the model to continuously hallucinate is not productive. Instead, it should be able to discover factual information and perform genuinely useful tasks.
Imagus is a browser extension for Firefox that allows users to enlarge images and display images/videos from links with a simple mouse-over. It is similar to the Hover Zoom extension, which is a popular tool for browsing memes and gifs on desktop. However, Hover Zoom has been flagged as spyware, so users may want to consider using Imagus instead. Imagus works with a wider range of links than Hover Zoom, including those from imgur. The extension also has customizable settings, including an expandable set of rules for getting larger images, media, or other content.
I’ve saved this, but it would be nice to see the syntax somewhere in the search engine
This is so cool. It’s exactly what I had in mind when thinking of a modern bash alternative.
Which is why we should make the implementer’s job easier by providing only the relevant information instead of wasting their time with a flood of comments on the issue tracker.
I don’t think local models have enough context yet. I don’t know how perplexity manages to use so many websites as references. As soon as there is a model like WizardCoder that has a 10k context or higher, I’ll give this a try.
I can confirm that the Opera method works. I couldn’t create an account before but I’ve managed to create the account using the built-in VPN of the Opera browser. Here’s an example of a Python script written with Claude 2 that GPT-4 wouldn’t have enough context for, https://lemm.ee/post/2595655 Really impressive.