Scan on end user's devices, but never transmit the result of that. Only report it ON the device itself to the user. A false positive when you send a pic of your naked kid to your spouse might show a warning icon asking you if you are sure you want to send it.
Also: for minors (Who is a minor not determined by some central age verification, but by me specifying in the Apple/Android family settings who my kids are) you could make sending certain things it blocking or subject to parent approval. E.g. if my daughter is tricked into sending nudes, it's something that's handled the same as if she wants to install an app or visit a specific web page.
No encryption is ever backdoored. Anything beyond this, e.g. reporting any user actions, would be allowed only through a court, just like any wiretapping always was.
What constitutes a replaceable battery in this regulation? I don’t want hatches on things. Especially not phones. It’s fine to have some screws between me and a battery, if that makes it have 1% more capacity, 1% more rigid or be 1% more water proof.
Not sure this is possible, or provides a meaningful ”lightness”. The 4.x CLR is an OS component as expected as Win32. You could have it as a separate download but I don’t really see the point.
I agree with the overall premise of the article though.
I’d like to go a step further. It shouldn’t just be ”light”, it should be power user focused. There are a thousand little annoyances that plague the OS (like defaulting to hiding file name extensions) which every technical-above-average user needs to adjust in every new install.
But who insures the ship and its cargo? And what's the premium? No one cared about sailors before either. But if the ship sinks then you cash out from Lloyds. But if the risk premium increases by a lot, then that adds to the cost of the cargo.
And eventually it's just not worth transiting the strait no matter how "open" it's claimed to be, if there are still unacceptable risks.
I hope so too. But I fear that it will feel inadequate if we know there is always a $20 online model that is an order of magnitude better. I don't think there will be a "good enough" local model so long as frontier models look so much better.
I wonder what the reason is to not go humanoid but super-human? Lack of training data?
For example, having 3 arms would help a lot of tasks. Or having fingers with twice the length of human fingers and 4 joints on each finger could enable them to switch a headlight bulb on a French car.
I did 81/100 (not my first language) but I probably only knew 60 from before. But I speak other languages and so I can usually decode an origin of a word or I have seen other words in English or another language.
So it’s not a test of how many words you know but how good you are at guessing what words mean.
Without the MS account, it is _much_ harder to sell OneDrive, Copilot, Office365-whatever subscriptions etc to their users.
I get it. I get that they need to upsell their customers OR their product would be more expensive. I'd be happy to pay that premium though, and I'm not going to buy any of those additional things.
But let's not pretend that this is purely an evil or thoughtless design choice that isn't economic in Nature. Windows has a cost and that cost gets subsidized by all the people who buy additional services.
I think something similar applies in many parts of the world, _but_ I also think that the response from the OEMs is basically "fight the warranty claims hard when there is third party service involved, and just honor the warranty with a smile otherwise". Basically malicious compliance is what they have to resort to. And as a car owner I'm lazy and afraid of conflict. So I happily pay X% more for the service to get the no fuss warranty (And the OEM shop is even happier).
The right to repair thing is a spectrum. On the one hand you have the John Deere inkjet model. But on the other end you have the normal car manufacturers who don't make it impossible, but who incentivize using 1st party repair. Are we talking about that being against right to repair? Because that feels somewhat hard to legislate.
E.g: A car has 1 year warranty. But the warranty is extended by one year so long as you service it with manufacturer, not a third party. This doesn't prevent me from servicing with a third party for half the price, but it might not be worth doing that. And I would think twice about buying a car which is only a few years old unless it's serviced completely at first party, due to issues that could arise with warranty claims.
Could legislation be made that says "If you repair at a third party, or even repair yourself, and you know what you are doing/do it according to the recommendation, then the manufacturer has to give the same warranty they would have if you serviced it at them"? Because that feels insane. The warranty is just a kickback for using their expensive service, getting you to the service place/showroom regularly etc.
This is especially nice in the age of AI. I (the graybeard senior developer) does all the risky refactoring. I can take a performance issue and turn it into six regressions in half a day ($100). Then everyone is impressed when I let Opus fix these regressions in 20 minutes and $2 worth of AI.
No one notices when you cut 20% of some expensive process but cause no regressions.
Yes it can't be used 24/7 for 5-minute charges because then the buffer does nothing.
But if there is never back-to-back charges then I'd argue it's also kind of pointless because when the speed most needed (when there is a queue) is when the charger starts going the slowest. The balance is to have N charges (say 3, 5 or 7) in the battery. That way you can churn through the peak with N charges (say 07-09) and then charge for several hours until the peak hour returns at 16-18 when you can once again at least serve the first few cars without falling back to whatever you can suck from the grid continuously.
An aerial bomb is equally indiscriminate. You make a judgement about the target value and the estimate collateral damage/civilian casualties and you fire if the equation is reasonable.
Same here. If it's a muddy square kilometer of trenches do you KNOW there are no civilians? No. But it's not indiscriminate. The target area selection is the key. You can't release an indiscriminate autonomus killer drone into a city of course. But I doubt it'll be long before these drones make target distinction of humans (Vehicle recognition is old at this point).
Just like with land mines it's the military value vs risk of civilian casualties that determine if it's a war crime or not. Some countries have signed the Ottawa treaty, but notably not the US or Russia. And the trend isn't going towards states signing the treaty but rather towards leaving it.