I would hope that any AI would be on-device. I think anything else would be a show-stopper.

We can move goal posts all we like. All modern companies writing software are using AI. My company is heavy in AI use and we are opening more positions not less. The problem is that these new roles are for much more senior engineers simply because the need for entry level engineers is being replaced by AI. The bigger question is how do you train new senior engineers.

Slop companies were making slop before AI, they are going to continue. There is a huge market for it. Just like spam, scammers, and all other sorts of nonsense. AI might change the pace, but it doesn't change that shit companies will produce shit things and good companies will produce good things. Both will use the same tools.

I mean, if it is lasers, the cat might chase it, and use up all nine immediately.

It's a trivial task that wastes one to two hours per person per month. For a four-person household, that's the better part of an entire person-day wasted every month. So it might be trivial, but that doesn't mean people want to do it.

No, this is *exactly* the kind of thing I want AI to do. Clean my house. Straighten my house. Put things away. Fold clothes. Free up those hours so I can do something more fun, more interesting, or more productive.

That said, there's one very critical design consideration that Apple may not be thinking about, but that will make or break this as a product: upgradability. No one will tolerate throwing away a hundred pounds of e-waste to get the latest version every three years. People are going to want to buy this and keep using it for literal decades, because they'll see it as an appliance, not an electronic toy. This means a modular design, both at a hardware and software level:

• All of the parts should standardize on CANBus for control purposes and something faster like 10-gigabit multimode fiber for video delivery.
• The central core should be a large open space with adequate fans to ensure decades of upgrades.
• The central core should contain a computer in a box that has its own fans to vent into the larger enclosure, and any network switching if necessary should be a separate unit with separate fans. These should be user-replaceable if future design changes require updates.
• Cables should snake through conduit for easy replacement of large components like arms or head.
• Physical connection points (e.g. arm attachment) should be standardized up front and never changed.
• Upgrade units should include: battery (individually, external, quick swappable), cameras (individually), hands (individually), arms (separate from hands), legs (individually), head module (separate from cameras mounted on it), brain, network switch.
• Each of these individually upgradable units should be easy to detach and reattach with a minimum number of screws and zero small connectors or thin cables.
• To the maximum extent possible, each of these pieces should be easy to take apart for component-level repair.

And so on. The software should be designed under the assumption that new software versions should be compatible with older brain units to the maximum extent possible, for as long as possible (minimum 10 years), and that nothing more than a brain transplant should be required for backwards compatibility forever, so long as you are not using new features that require new versions of components (e.g. you might need an upgraded camera module for the eyes for some future feature, but that should block only that feature, and only until you swap the module).

In other words, for this to actually sell, it has to be built more like a Windows PC than a Mac. And this is why as much as I'd love to see the level of polish Apple would bring to something like this, I have my doubts about whether they can distance themselves from the hard-to-repair form-over-function design concepts that they are used to, and create something that behaves more like an appliance that people will pass down to their grandchildren when they die.

Thank you for this sensible comment.

Often in copyright cases for song melodies, if the artist being sued for infringement could have possibly had access to the music they’re accused of copying, even if it was something they listened to once, they can be accused of “subconsciously” infringing on the original content. One of the most notorious examples of this is Tom Petty’s claim that Sam Smith’s “Stay With Me” sounded too close to Petty’s “I Won’t Back Down.” Smith eventually had to give Petty co-writing credits on their own chart-topping song, which entitled Petty to royalties.

That eventually lead to this https://archive.org/download/a... where programmer, musician, and copyright attorney Damien Riehl, along with fellow musician/programmer Noah Rubin, sought to stop copyright lawsuits that they believe stifle the creative freedom of artists. They did that by just using a algo to make ever possible 8-note, 12-beat melody combo within a single octave. They then copywrighted it and released it to the public domain.

So clearly written page is not required when assessing a music recording. Hell even John Fogerty was once accused of stealing his own music and won the case by playing a guitar in front of the jury.

In 2003, Trump a card holding liberal sent Epstien a birthday card. That was a liberal attack to hurt today's Donald Trump, the most conservative man on earth and our greatest president.

Fucking liberals man, ever time!

Using AI is a wide spectrum.

You can easily use AI to make great video games, just like you can use AI to write any other software. Just like you can use intellisense for code completion and still make great video games.

The thing that determines if a product is great is the people involved in the process who understand the business case, the expected work product, and the spirit of the goal. So I'm 100% for a game company using AI to help it write code and create the building blocks of a game. Like any product code needs reviews, graphics need tweaked, the software needs to be tested and product tested.

So yes, if you one shot a game with AI and do no testing or tweaking it's going to be shit. If you use it appropriately like any other tool you can accelerate your deliverable and make fine products.

I think we will see some games crafted with love that leveraged AI for the parts that are less interesting. For example someone with talent in graphic design and writing who has a wonderful idea for a turn based RPG, but isn't sure how to make Unity work leverages AI to turn that story and art into something fun. It would be great if everyone could find people who believed in their ideas and worked with them, but often all you find are people who tell you nothing will work. Sometimes you have to go it alone and AI is making that easier.

Software development as we know it is ending. AI tools are 'good enough' that any idea can be made fairly quickly with even limited knowledge. This opens up a lot of opportunities for creativity and itch scratching while at the same time impacting the commercial value of software. I'm shocked the open source community doesn't see this as a pivital moment for free software.

For example, recently I wanted to play nethack on my m1 mac. I also wanted to use a tileset. Finding no good OSX arm based clients, I spun one up in a week with a custom tileset that I tweaked to my likings and even added graphical features to represent the true 'size' of items and creatures in the game.

Now I'm fully capable of building that solution by hand, it would have taken a month or two, but I would have needed to find a like minded person who knows how to build graphics and shared the same visual inspirations and goals I had. That would have probably just caused me to end the idea. Instead I scratched that itch and have been playing nethack a ton more and I'm happy. I released the code as open source so anyone who cares to can play with it.

To me that is real value and not 'slop'.

model weights already more or less "contain" several partial OS kernels,

No, they don't. They were trained on that data, but they are not that data. You can't extract it from weights, and I highly doubt you'd be able to reproduce it from weights, even on an exceedingly large model.

At best, it's a work inspired by the original, not even a derivative. There's no "substantial or material" replication.

Why is this modded insightful?

Is it "stolen work" to read a book and then recite it, or paraphrase it, or create a story with the same themes? No. Those are new forms of work. The actual work isn't being recreated. Is software not the same as free speech? Can I not use phrases from other speeches in my speeches, and still have them be original work?

Who owns the copyright on the Klingon translation of Mein Kampf? Does Hitler retain those copyrights? No.

This "AI is all theft" argument is as tired as the old Microsoft lie that compatibility with their shitware was the same as copyright infringement. It's a tired argument. Retire it.

Look, I get it - a lot of the anti-AI folks don't take the time to inform themselves of how AI works, and it's generally scary, because it's threatening your identity as a coder. It's unfortunate, and not that different than the stablemaster being threatened by the automobile. That's the way things work, and there's no putting this cat back in the bag. You've got to adapt or die, just like the COBOL and FORTRAN developers of yesteryear. Or you could become a Maytag repairman.

Are we really going to trust something security-oriented by a guy who conceived of systemd? Instead of creating something which improves security and isolation, he created something renowned for its architectural and management complexity, privilege boundaries (or frequent lackthereof), attack surface, and the frequent consequences of concentrating system management in a large privileged software stack.

I'm reminded of the SSH/Pulse Audio privilege escalation through systemd some years back... due to their obstinately poor security design.

PowerBook Pismo: Display cable sliced by Airport antenna connector. PowerBook Wallstreet: Hard drive connector board attached by a single screw that didn't hold it in place, causing constant disconnects. The iBook G3 graphics chip disaster. So many mechanical design mistakes, so few column inches to write about them.

You mean 24x7 solar power? Now calculate how many centuries an on-earth data center could go, even with having to use batteries at night, even with replacing the backup batteries every couple of years, before the cost would equal the cost of a single launch. In other words, that's not an actual advantage, at least in practice.

Besides, you can't repair or upgrade the hardware when it is in space, which is such a colossal disadvantage that the advantages would have to be almost unimaginably great to balance that out.

The only theoretical advantage I can think of for a satellite would be that if someone launched one from international waters, if some random government wanted it to stop existing or stop doing something, they'd have to launch something to destroy the satellite, because from a legal perspective, it wouldn't be within their jurisdiction.

But in practice, as far as I know, nobody is launching anything into orbit from international waters, so this is entirely theoretical, at least right now.