The US military budget is 900 billion dollars. The government can afford a few hundred million for some sensors, it should not need private sector patrons
That number, incidentally, is the entire NATO military budget. Which scares me because I can imagine someone planning on taking action that would result in the dissolution of NATO thinking they can make up the difference that way and choosing that number with that in mind.
Is it actually coming out of the military budget? I assumed we were just going to get taxed for it, or they'd pull it from some other places that actually needs the funds.
Alternatively, they could just increase the military budget by 300 billion (or more).
That's only part of it, the US and allies are reportedly going to set up a $300 billion fund for Iran, the particular terms of accessing it aren't clear at this point. GP is correct in that the US is not putting up $300 billion, but it will be putting up part of that money. The US fought and lost a war and Iran comes out ahead, and the US and its allies come out worse and poorer.
> Don’t RMA it, and don’t solder it yourself. A local phone-repair chain with a microsolder tech can put a 3 mm SMD part back on a GPU PCB in twenty minutes for the price of dinner. The skill is in your city. You just have to look.
The trouble with this though is, what if that is not the only issue with the card? That’s normally my thought process on reaching for RMA. The unit could be an all-round lemon that should not have passed QA etc. (and as noted in the post itself, working for a week on various tasks is not enough to prove it good)
> However, having had much success with experimentation so far I decided to try something that would probably work: I copied the code from the magazine into Notepad and saved the file as .exe instead of .txt.
Glad to hear I’m not alone in these kind of early experiments. I remember having no idea what the concept of programming actually was, and opening game EXEs in notepad to try and understand how they worked. The demo of Majesty was one I particularly wanted to modify and had no idea how
Tried out doing this kind of stuff a while back to create a custom user space, and I found that suckless’ init, sinit, was quite educational for how to deal with init’s responsibilities in a super minimal way: https://core.suckless.org/sinit/
The problem with this study though is that it doesn’t really illuminate anything. Psychedelics restoring the default mode network in the brain is already somewhat understood (*that it happens, not the mechanism of how), so it’s not that strange a temporary reversion of the symptoms of Alzheimer’s would happen.
And it’s not even suggestive of eg making an actual medicine that could be taken long term, because Alzheimer’s physically destroys your brain. The restorative effect of psychedelics is just a bandage over not understanding why that damage is happening in the first place.
Yeah this story was on HN before a few weeks ago and I raised similar - it’s nuts to give that to someone who probably doesn’t have the mental state to be able to comprehend what is going on. Don’t understand how it is possibly ethical to do.
Especially with the effects being temporary - can you imagine how awful it must be to regain lucidity outside of your control and then lose it again for the sake of an experiment like this? Awful experiment.
Yeah well I was imagining how medical ethics work with respect to putting patients through unnecessary suffering. Temporarily restoring someone’s lucidity with the knowledge they will lose it again (having already suffered the progression of the disease the first time) is more helpful to you than it is to them
One advantage that does come to mind in light of the Iran war (and the loss of an AWS DC) is difficulty in attacking it, even when it’s directly above foreign territory. I wonder if one of the intended customers will be gov/military? Conjoined spy satellite/DC for some function maybe?
I don’t know what the end game of the US looks like but surely you’ll always find some territory around the world to relatively safely put datacenters? And if your opponent isn’t Iran but something like russia or china, they’ll just blow up your datacenter in orbit, too.
I see the point to maybe do some onboard data processing on spy satellites etc, but on the other hand, downlink bandwidth seems to be become less of an issue over time, so it doesn’t seem that important to me compared to just sending down the raw data over star link or the military equivalent.
> First of all, structs aren't used so you don't have to invent names for them (e.g. there is no IntVec)
But since it’s storing a void pointer any way, they wouldn’t need separate names right? You could use one struct everywhere regardless of the type of the items
Which IMO is a better idea than using an array here because the fields can be properly named and typed to prevent accidental misuse
Requiring at least an iPhone 15 Pro also seems like a mistake, unless it’s for actual hardware reasons. The 15 is only 3 years old, this requirement cuts off a lot of potential users I think
I think it's the requirement of having 4gb+ vram (for gemma+context) free at any one time, any phone older than that cannot materially satisfy that demand: https://iosref.com/memory-processor
It’s definitely for hardware reasons. They have been aggressively improving the vector math capabilities in their chips, but as anybody who has tried to run a local LLM will tell you, newer hardware works better and you’re always limited in what you can do.
> why older devices can't use private cloud compute thats 100% off device is just apple being anti-consumer.
I don't know if it's "anti-consumer" to NOT roll out free cloud LLM usage to everyone. The idea with only giving it to the devices with on-device AI capabilities is that ideally most of the tasks will cost Apple nothing because it will run on-device, and anything more complicated will start costing them tokens.
If they gave it to devices without on-device models, ALL Siri requests from people with older iPhones will suddenly be burning money.
Not to mention, if we assume responses from the cloud are better than the local model, then the older iPhones get an overall better experience than the newer ones.
The design is that there is always a local model capable of forming a remote query with just the subset of local data on your phone needed to answer that query.
They may have decided that local processing was a MVP feature either for faster responsiveness or to reduce cloud cost. It may have been additional memory pressure or a limitation in processing on the previous A-series chip. Or they may have simply decided it wasn't worth creating and validating Yet Another model.
They would have to build the product twice one for mobile chips and one for server, and then there would be functionality discrepancies. Or even worse, the on server one might work better than the on device one that newer phone users get.
If you want hosted AI you can already install the Gemini app or whatever. The only advantage Apple can offer is something that runs on device.
While I generally agree with your sentiment, imaging how they would say it to users: your ai works on iPhone 15 pro but some things will work if it’s a little less private we send things to a server then the regular 15 can do it. Image generation is server based so 15 is ok, but editing an image is not since 15 does not have enough ram but 15 pro does. Etc etc.
reply