There have been good writeups on why Apple doesn’t provide gyro data — it can be used to physically track people. This is mostly an issue in apps that embed Safari, such as a store loyalty app that can track your movement while you’re in their store — or in a competitor’s store. Since Firefox isn’t embedded in apps, it’s not an issue there.

“You know you’re Generation X when….”

I find it odd that people now reflect fondly back on “old Facebook” when I remember Zuck and his friends going on about how they got hot girls to give them their photos and personal info.

Hopefully the pendulum will swing again; Gen Z doesn’t actually seem very impressed with AI and the digital services economy.

Hybrid Electric Vehicle

There was an x prize for a tricorder about a decade back; they found that to accomplish all the tasks, they actually needed three sets of hardware, two of which needed to be strapped on to the person.

That’s all great, but all it takes is to unalign a single parameter and it appears to unalign the entire model.

So this is great for ensuring you’re testing what you think you’re testing, but it’s not going to actually secure a model you’re going to make open.

I’m confused. Aren’t you the one who referred to LLMs In a thread that was conflating LLMs with AI? The parent’s comment seems to be right on point.

It’s kind of like how we’ve lost the war on hacking.

Large language models fall under the current definition of artificial intelligence just as much as Cyc or Cog did in their day, or various expert systems and machine learning models, diffusion models, etc.

Pretty much any non-deterministic inference engine can be classified as an AI, including LLMs.

But which 10 minutes?

One sec, maybe ChatGPT knows….

Yes, but Teslas are spying in everyone they are in the vicinity of, too. Most cars don’t send video from their onboard cameras back to the manufacturer for analysis.

Cheaper for them, that is.

What I want to see is throttleable models, kind of like progressive JPEG, where the default model is “nano” and it has a watch function that analyzes if more tokens might be needed for a certain task and scales up as needed — identifying if the resources are too much for the device and offloading to the cloud (with explicit permission) only if (but always if) needed. Over time as the technology improves, larger models move to the endpoint.

And then people could have a basic set of sliders: on-device only, on-cloud only, or somewhere in between, based on the user’s preferences.

“There was a long time where the United States government advised other countries on how to protect people within their territory from foreign oppression,” Perloff says. “And it is appalling to realize that now other countries may have to do that about us.”

I use my phone as a media platform; I rarely stream any content on it unless it’s off my Jellyfin server.

I also have a Sony Walkman I got in 1986, but as I don’t have any audiocassettes anymore, it mostly operates as an AM/FM radio (tell me again why our phones don’t have AM/FM when Japan and S Korea do?).

One thing I HAVE been considering is a LoRA Meshtastic setup; plug in a router with an extender board at home, and carry a handheld unit or two with me. That would mean that I’d have a non-cellular connection to my home network within a 25km radius of my home, and mesh networking connection to the Internet in most densely populated areas outside that range, AND the two devices could link any phones or work as walkie talkies in areas where there’s no available WiFi or cellular signal.

To me, that’s a step forward, where a dedicated digital audio player doesn’t really solve any legacy or current problem.

But if you’ve already got one, why not use it? The batteries last a whole lot longer than a smartphone.

I still have an iPad 2 I bought in 2011, and amazingly, the battery still lasts 5+ hours.

It’s got 128GB storage, and that’s used to store a movie or two, 30 hours of music, and a bunch of books and PDFs.

Of course, I’m not going to carry it with me everywhere; that’s what my phone is for. Which also has those things stored on it.

Good luck restricting it for this purpose… but as you’ll see, there isn’t actually a creative commons license that is just “CC” anyway, so that wouldn’t be an issue.

We also have .zip but no .rar, .ace, .gz, .bz, .7z, .xz or even .tar.

Actually, you’re on to something there; have the TLD represent the license agreement to access ANYTHING served over it.

.gpl .lgpl .agpl .mpl .mit .bsd .apache .epl .cddl .cc0 .ccby .ccbysa .ccbynd .ccbync .ccbyncsa .ccbyncnd

I’d pick .

And then people can register whatever TLD they want on top of that.

Wait… this is specifically about websites?

Easy solution: stick your website behind a CDN. That way, people are using a VPN to contact a CDN, and only the CDN ever connects to your website.

And if Utah thinks two degrees of separation isn’t enough… well, it’s likely that every legislator in Utah is two degrees away from someone who will break this law, so they should obviously be the first to be subject to its penalties.

Have there been actual studies done against the various facial recognition technologies out there right now to see what the minimal adjustments are to break it, and following that, to successfully and reliably be recognized as a specific other person?

Who thought to name it Epstein-Barr?

What I dislike is that the term AI has gone the way of terms like “hacking” and “the web” — I’ve been working with AI for over two decades, and suddenly LLMs are getting VC funding that’s enough to buy up the world’s memory and CPUs, use up entire communities’ water supplies, and set back responsible use of fossil fuels by decades. And now we’re being told to use these massive models or be replaced by them.

In other words, the problem isn’t AI, it’s people (as usual).