If you shipped most software from 2024 back to 2000 it would be (correctly) labelled spyware. It also wouldn't run at all because system requirements even for simple applications have bloated beyond belief.
AI is just the latest iteration of these trends - now the vendor gets to record everything you put into their software and a text editor requires exotic new hardware so that the AI "copilot" can run and help you type every character.
Third-party calculator apps on mobile are absolutely bloated because they're just ad-delivery systems. It's parasitic behavior from developers looking to cash in on basic software we've had for decades.
Worse, they load slower now on today's hardware than they did 20 years ago on hardware from 20 years ago. There is no excuse for doing what Windows 95 did, with one hundred times the system requirements of Windows 95. I'll give them a pass up to Windows 2000 or XP, because that's when it became stable by removing all the DOS compatibility stuff.
> Microsoft's minimum requirements for Windows XP are a 233 MHz processor, 64 MB of RAM, 1.5 GB of available hard drive space, and an SVGA-capable video card
Is it any wonder, when most modern text editors are a giant ball of javascript running in a giant browser? Seems like hardly anyone is making desktop applications anymore, everything is webapps in browsers masquerading as desktop apps.
There are clear benefits to building web apps vs native apps. You only have to write a single web app and it runs virtually everywhere. Also, there is a huge pool of talent that corporations can tap into, people who know web development. It's a less specialized skill.
What I really wish existed is a kind of web browser with a greatly limited feature set, so many web apps wouldn't need to be so bloated.
There are benefits to native apps too. Being able to work offline and not having your every action and input spied on in real time is a big one. It's all a question of trade offs. Users who prefer privacy, lower-resource consumption, offline use and more familiar/native interfaces will probably not want a web app. Users who don't care about their privacy and want a program avilable to them anywhere and anytime they have an internet connection might want a web app.
You should never use an LLM/AI Chatbot with a third party service for anything that could be confidential, private in nature, or may have to do with your security.
All of the LLM/AI providers can read your contents. The only thing between them and your chats is "trust," and we've all had promises broken on us before.
Trust is not security.
Shameless plug: You can definitely use IPv6.rs and Cloud Seeder [1][2] to run Ollama and OpenWebUI on your own computer (confidentially) and access it via TLS remotely (via phone, etc.).
That advice is not limited to chatbots. You wouldn't believe how many times I've seen people paste sensitive data into online tools (formatting json, validating xml, etc) without giving any thought to what happens with that data. Or pasting error messages into Google, verbatim, including more sensitive data.
This is an important concept that I’ve never seen described before.
It’s not just big tech. The government is also heavily depending on this creeping change in expectations.
Setting aside what it has been interpreted to mean legally, “unreasonable search and seizure” in English can either be taken to mean “searching without a reason” or “searching more than is normally expected.” Obviously the latter is how most people would read it.
In this way, huge legal invasions of privacy such as the Patriot Act and its descendants can never be viewed by the general public as unreasonable search, precisely because of how many people they affect. Everybody faces it, so it is incredibly commonplace, so everybody expects it, so it can’t be unreasonable.
Last week, the company that I work for rolled out Copilot for all Microsoft Office apps.
AI does not really help me for the kind of work that I do, so this really didn't seem like it would affect me much. Except one day I open Outlook and there's this giant-ass honking big "Summarize this thread" bar that they added to Outlook. It's a full inch in height and full width, which means that it takes away valuable space needed by the message list and message pane. All for a single button. I would have preferred they fill that space with ads for Candy Crush, because at least then it would be doing something arguably useful!
Further, the company's legal team drafted a list of 11 things you CANNOT use Copilot for. I won't paste them here, but basically it amounts to not being able to use Copilot on anything you wouldn't share with the press or a competitor. I can't actually think of a single employee in the whole company who can follow these rules AND use Copilot effectively for their job. IT says that Microsoft enabled this for everyone using O365 and that they can't turn it off for individual users.
So basically I now have this fire-alarm sized button in the middle of the screen on my mail client that I must NEVER click.
I was recently gifted an old dslr, and after finding that Canon still makes the original picture viewing/editing utility available on their website, was blessed with a beautiful pre-ribbon era interace that does exactly what it says on the tin, offline and without needing an account or subscription or cloud storage or advertising.
Playing with that was the most fun I have had in front of a computer in a very long time.
>The natural conclusion was that Microsoft was spying on its AI users, looking for harmful hackers at work.
openai and azure routinely ban people for other reasons too. it's clear as day that all your prompts are logged, evaluated and stored indefinitely to be shared with a number of third parties, three letter ones and otherwise.
I have to assume this is true for every cloud service. Unless something can used locally and entirely offline it's probably spying on you constantly and pushing ads.
I kind of knew this already, I think many of us did, the idea that for kids growing up today, invasive technology is just normal for them. It's great to see the idea being formalized like this though.
20 comments
[ 2.6 ms ] story [ 58.6 ms ] threadAI is just the latest iteration of these trends - now the vendor gets to record everything you put into their software and a text editor requires exotic new hardware so that the AI "copilot" can run and help you type every character.
Note: my expectations may have changed over that time, but still, I think my expectations run behind Moore's Law.
https://pcalc.com/index.html
> Microsoft's minimum requirements for Windows XP are a 233 MHz processor, 64 MB of RAM, 1.5 GB of available hard drive space, and an SVGA-capable video card
What I really wish existed is a kind of web browser with a greatly limited feature set, so many web apps wouldn't need to be so bloated.
All of the LLM/AI providers can read your contents. The only thing between them and your chats is "trust," and we've all had promises broken on us before.
Trust is not security.
Shameless plug: You can definitely use IPv6.rs and Cloud Seeder [1][2] to run Ollama and OpenWebUI on your own computer (confidentially) and access it via TLS remotely (via phone, etc.).
[1] https://ipv6.rs/cloudseeder https://github.com/ipv6rslimited/cloudseeder
[2] https://www.youtube.com/watch?v=g_JyDvBbZ6Q
It’s not just big tech. The government is also heavily depending on this creeping change in expectations.
Setting aside what it has been interpreted to mean legally, “unreasonable search and seizure” in English can either be taken to mean “searching without a reason” or “searching more than is normally expected.” Obviously the latter is how most people would read it.
In this way, huge legal invasions of privacy such as the Patriot Act and its descendants can never be viewed by the general public as unreasonable search, precisely because of how many people they affect. Everybody faces it, so it is incredibly commonplace, so everybody expects it, so it can’t be unreasonable.
AI does not really help me for the kind of work that I do, so this really didn't seem like it would affect me much. Except one day I open Outlook and there's this giant-ass honking big "Summarize this thread" bar that they added to Outlook. It's a full inch in height and full width, which means that it takes away valuable space needed by the message list and message pane. All for a single button. I would have preferred they fill that space with ads for Candy Crush, because at least then it would be doing something arguably useful!
Further, the company's legal team drafted a list of 11 things you CANNOT use Copilot for. I won't paste them here, but basically it amounts to not being able to use Copilot on anything you wouldn't share with the press or a competitor. I can't actually think of a single employee in the whole company who can follow these rules AND use Copilot effectively for their job. IT says that Microsoft enabled this for everyone using O365 and that they can't turn it off for individual users.
So basically I now have this fire-alarm sized button in the middle of the screen on my mail client that I must NEVER click.
Playing with that was the most fun I have had in front of a computer in a very long time.
openai and azure routinely ban people for other reasons too. it's clear as day that all your prompts are logged, evaluated and stored indefinitely to be shared with a number of third parties, three letter ones and otherwise.