-12
CoPilot is a good feature and will be built into every major desktop OS (lemmy.world)
submitted 10 months ago by to c/unpopularopinion@lemmy.world
84 comments fedilink hide all child comments
 

Pressing the copilot button to instantly bring up a text box where you can interact with an LLM is amazing UI/UX for productivity. LLMs are by far the best way to retrieve information(that doesnt need to be correct).

If this had been released with Agentic features that allow it to search the web, use toolscripts like fetching time/date and stuff from the OS, use recall, properly integrate with the microsoft app suite. It would be game changing.

We already have proof that this is a popular feature for users since its been integrated in every mobile phone for the past 10 years.

you are viewing a single comment's thread
view the rest of the comments
[–] [S] 2 points 10 months ago (1 child)

I agree that phone operating systems are way to locked down and hostile to user privacy. I wasnt holding them up as an example of what we should strive to replicate. I was just pointing out that I find the assistant feature on phones to be useful with how it can handle natural language query and preform different actions and I've heard many people say the same thing. With linux yeah we have enough choice where there will always be non AI distros but I think once the tools become good enough they will get adopted by ubuntu, fedora, mint etc. A tool like https://github.com/qwersyk/Newelle could one day be shipped with ubuntu and I think it would be good. Giving users a local first private AI that can help them do things would be a huge usability improvement in my opinion. Just the calendar event booking alone would sell me.

I also agree with you on AI productivity, sometimes its better and sometimes its worse and sometimes its catastrophically wrong. I'm mainly trying to make the argument that Copilot(and other AI assistant implementations) are a good feature/workflow for the users. I accept that their current state is unpolished and copilot is marketed to do way more than it can do. But I think its core concept is solid and the features are being built out and they will get to a point where its commonplace and in every major desktop. I've been following people who are using AI on their linux systems directly linked up to the terminal and it looks useful to be able to say "book and event on x day doing x" or "send John and email saying X" or "what was that file I downloaded yesterday". These kind of actions currently work with Claude + tools but unreliably at the moment and the safety aspect is yet to be solved.

I'm sure someone will read this and say "but you could just send that email with a single command" but a normal user isnt going to send an email by typing in like thunderbird -s -t test@google.com -bcc mydad@google.com -m "hello" for sending an email but an AI can easily turn a natural language request into this command. So from the normal user perspective they go from having to open up a gui and enter out all the fields to pressing a button and typing out what they want or even saying what they want into a mic.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 2 points 9 months ago* (last edited 9 months ago)

    Seems we're basically on the same page.

    I'm fairly sure Linux tends to lean more towards tech-savy people, who in turn tend to be the more pragmatic ones and think in a more problem-oriented way. So I'm positive it's gonna be more about productivity in that community. They'll adopt something based on usefulness.

    It's just the companies who don't operate like that. Their AI tools are more pushed in a top-down way because of the investment bubble all the companies take part in. It's not necessarily about productivity or anything. That's some desired side-effect, but I think all of it is more about what their investors want to hear.

    As if now, I'm not sure, maybe it's still net-negative for us, the Free Software community. Our servers get hammered by AI crawlers, our projects swamped with fake AI bugreports. While the AI tools aren't good enough to be of proper help in more complex projects. And we don't have an infinite amount of money to just push for it anyway and care about profitability in 10 years... So I think we're bound to do it the other way round. And AI has to actually prove itself, and that takes some more time.

    For example, I hope some day I'll get some modern AI tools in my image editor. I mean I'm of the pragmatic type myself, I'm gonna use it if it contributes to my life and doesn't come with a devastating cost on society and the environment and other individuals. Same with chatbots. I don't think we can tell yet. I think we first need to make it way more "intelligent" and come up with new regulations, ways to deal with the negative aspects... Currently it's a bit of a train wreck with the flood of fabricated things that displace human conversation on the internet, Americans like Peter Thiel who makes big bucks inventing Skynet and push for doomsday. And we can tell it's not a positive balance yet, because almost all AI companies aren't profitable. But maybe we can tackle that. And it's the promise. We'll see. At least on the technical side we seem to be making progress each new day.

    Yes. And these AI tools with terminal access seem fun to explore. I think they're called coding agents. And we get quite mixed reports. Some people use it and it (roughly) gets their job done. For other people it just casually deletes their harddrive or does other weird things. We really want something like this, though. So we as humans transition from doing the coding to being software architects and the AI does the actual coding. I think it's very difficult to have things in-between, copy-paste all the time and argue with AI, then nobody has a look at the code, so we miss the security issues and only learn about it after the company has been hacked... I think Instead we want some end to end solution that just reads the specs and does everything including some testing, integration and security and factors all of it in. And for more than gimmicks, it needs to do the job to some acceptable level. But that's to a large degree a technical problem, and we might be able to figure it out in the future.

    I'm also looking forward to AI being able to do proper useful stuff, like clean up on my messy harddrive, do my personal bookkeeping and paperwork... I don't think we're there yet. At least I haven't heard people do that (successfully). But that'd be a nice job to delegate.

  • permalink
  • fedilink
  • source
  • parent