• 20 Posts
  • 446 Comments
Joined 3 年前
cake
Cake day: 2023年7月13日

help-circle

  • Not really, it sends a combination of 3 really fast button combination on press and doesn’t report to the os when your finger releases the key. You need to run a daemon that detects that specific combination and emulates a finger release after a specified time

    Assholes. Couldn’t they just remap one of the useless “internet” keys from the 00s like KEY_HOMEPAGE or XF86Mail??

    Or create a brand new scancode since they could change the kernel without any problem, especially because the manufacturers would need to change the keyboard firmware anyway AND because copilot requires special hardware, it doesn’t need retro compatibility with older windows versions that don’t have the scancode in kernel or in the keyboard driver













  • I’m using Gmail as an email client. I tell Gmail to go to fetch emails from my other accounts and so I have unified email. I don’t use @gmail.com as my primary address.

    Problem: from today Google discontinued the feature and they warned me only yesterday.

    Which webmail has almost feature parity with Gmail?

    I need filters and fetching emails from multiple accounts. I have a dedicated server so the part of fetching email can be done from an external program


  • You need to do a custom program if you want to do that. I mean a traditional program where variables are stored properly.

    The models have no memory at all, at every question it starts from scratch, so the clients are just “pretending” it has a memory by simply including all previous questions and answers in your last query. You reply “ok”, but the model is getting thousands of words with all the history.

    Because each question becomes exponentially expensive, at some point it starts to prune old stuff. It either truncates the content (for example the completely useless meta ai chatbot that WhatsApp forced down the throat loses context after 2-3 questions) or it uses the model itself to have a condensed resume of past interactions, but this is how it hallucinates.

    Otherwise it will cost like $1 per question and more







  • I like the idea of a federated version of that, but there are many problems:

    1. When people scroll, they need an algorithm to choose content for them.
    2. 99% of average TikTok content is low quality shit and it’s the algorithm to find and promote what works
    3. All that 99% of shit is occupying exabytes on bytedance servers. Users are uploading terabytes of shit every hour, it’s impossibile for a federated instance to keep all this shit on disk. Look how Lemmy is designed for example. I upload an image and then it’s stored forever in hundreds of instances. For videos is untenable. Only companies that are profiting from this (for example using such videos for training ai models, or using this videos to hook people and serve them highly personalized ads) can host a TikTok clone