Workers creating AI systems to replace other people's jobs are complaining that their own jobs will be replaced by AI?

You are arguing that you should be able to have your phone so that in the case of a school shooting, during the last few seconds to minutes of your life you can feel better because you can call your mom? A quick search puts the chances of dying in a school shooting at about 1 in 5 million per year. Let’s say you are the unlucky one but do get a whole 10 minutes between the time you use your phone to call mom and when you die. The statistically expected outcome duration of a "better life" would be 0.00144 seconds (10 minutes times 60 seconds times 12 years of school divided by 5 million). I would argue that in the way more probable case that you don’t die in a school shooting, having your phone locked away will make the remaining length of your life better (due to a better education) for something like 60 years (life expectancy minus time in school fretting about no phone times nearly 100% probability). Expected time of a better remaining life is about 14 order magnitude longer with the phone being locked away. Even if you are the unlucky one that gets shot, you could have been telling your mother you love her every morning when you leave home – reducing the need for that "last call".

I am personally not that crass in real life and of course being a bit facetious, but since you have a six digit slashdot ID, you are not a kid in school and are making up a scenario. It’s the tired "think of the children" argument. I am countering with what would in the big picture be statistically “better for the children” with specific respect to cellphone access in school.

School gun (and other forms of school) violence do need to be addressed and the cloud of its possibility do weigh on children – even those who never see a shooting. It’s a serious issue. Let’s not dilute the issue by suggesting that cell phone access is some sort of compensating factor.

I believe you are incorrect. NPR reports a Pew study that indicates even every if everyone eligible to vote did, Trump still would have won. So while technically a majority of Americans didn't elect Trump, a majority of those who voted did, and a majority of those eligible to vote would have. Note that this is NPR reporting this, not Fox, and that Pew Research is generally non-partisan.

In today's political world 1.5% is not really a narrow margin - it's around 5 million voters. The Democrats need to realize that they are not in the majority and this is not due to voter suppression. What the the party needs to do is figure out how to appeal to more average Americans living real lives. Republicans seem to be doing this - yes they are doing it by lying - but they seem to be connecting better to Joe Sixpack than the Democrats. While personally I am not affiliated with either party, I am saddened by the trend toward anti-intellectualism. Being a reasonably critical thinker, learning the level to which I am in the minority really sucks.

It would depend on what is required to be "certified". If the apps I create contain objectionable functionality or content (they compete with the core platform or they contain pornography or something), but I am willing to put my name behind them (and take on any related legal liability), would they be allowed? If the answer is "yes" then requiring signed code (being a certified developer and being able to get a signing key) seems reasonable, however if there are other restrictions, then the developer program is about more than "safety". Also if the cost to become certified is unreasonable then the program is also more than about "safety".

This concept should get RFK foaming at the mouth.

The pandemic proved that it was possible to survive as a business without doing everything in person, it didn't prove that it was the best way to operate, and it didn't prove that businesses would thrive without some level of in-person communication. I support the concept of remote work for most tech roles, and I have primarily worked from home since before the pandemic. That being said there are some meetings that need to be in person, particularly when the the work is beyond being a pure coder or keyboard jockey - typically in the consultancy and sales areas. You might not ever want a job that requires you to deal with someone in person and maybe need to get on a plane for that in-person meeting to happen, and that's your choice. But there are jobs that occasionally require this, and requiring an in-person interview is one way to filter out those who can't or won't do this. Beyond filtering out fake resumes, in-person interviews are a good way to determine if a candidate is capable of functioning in the real world.

I get that at small and probably medium scale management of an array is a solved problem. I have a similar sized array on my roof with micro inverters. I have 28 panels and in their 12 years of operation I have been able to easily mange the four inverter failures that have occurred. I am curious about what configuration and monitoring you use for 1.3 million panels. Do you still go with micro inverters (while they do isolate you from individual failures, they also cost more and add more failure points)? Do you you use string inverters, and if so what is the typical string size? Do you keep it all DC between the panels and storage batteries and use massive inverters between the battery and the grid? Given that a system of this size is always going to have some level of failure, what level of failure do you just tolerate and what are the thresholds for repair? What kind of staffing do you need to monitor and keep a system like this running? I am not critical of such a system, more curious as to what it takes from a technical and operational perspective to keep such a system running.

1.36 million solar panels is a lot. I would love to see some photos of what that many panels actually look like and how much space it takes up. The photo in the article is clearly not that many panels. My first thought was that maybe the article meant that there were 1.36 million solar cells instead of panels as typical panels are built with 60, 72, or 96 cells per panel. This would result in a total number of panels in the 14 to 22 thousand range, which seems like a more comprehensible deployment. However if you divide the total megawatts by the number of panels quoted, you get about 550 watts per panel, so that would be talking about panels, not cells - that passes the sanity check for the numbers.

So how in the heck do you manage an array of that size (or the set of arrays that add up to that size)? The number of electrical connections and the number of failure points would be insane. I am not suggesting that there would be any single points of failure, and solar is extremely reliable, but even with very low percentage failure rates, at that scale you are going to have some level of failure. It would be cool to know how you monitor and maintain a setup like that, and what operational threshold for repair actions are (do you take action if a single panel goes down, 10 panels go down, 100, 1000, 10000, etc?)

Anyone here have experienced with managing systems at this scale or who can point to technical articles about such?

It is a bad deal and it's dumb, but if you don't take it, then you are over paying by 2% and getting nothing back. The figure is likely higher than 2%, but the point is the same.

I am not sure the is true, do you have a reference? Even if it is true, would those chips not need an external (to the chip) antenna (that could either be not installed, or defeated by grounding out)?

When my.yahoo.com went away recently, I switched to protopage.com - I found this very post on the slashdot RSS feed.

So how would a backdoor in a chip actually work (specifically and technically)? In the case of a CPU or GPU are we talking about a special instruction that when executed kills the chip (blows a fuse or something), are we talking about an extra hardware pin that when the signal is raised or lowered kills the chip, or what? I ask because the concept of a backdoor in hardware seems kind of weird. It seems that it would also need a significant software or firmware component, as well as off chip hardware in order to trigger the kill. If software, firmware, or off chip hardware is required, what would prevent someone from just not implementing that hardware or not installing that software in order to close the backdoor? Or are we really talking about a backdoor implemented in a reference design, chip set, or other tech that isn't really just the CPU/GPU? If it's a reference design, what prevents the adversarial country or party from not using the reference design?

I get that there are things like the management engine in Intel reference designs, and when using the reference design the built in network interface has an extra endpoint not seen by the operating system, but one could protect themselves by either firewalling the machine or adding their own network interface (and not plugging in the built in one). GPUs typically don't have built in network interfaces, so how would a GPU "backdoor" be remotely accessed or triggered?

I am not saying that it can't happen, but I would love to know the technical details on how it could.

While I get your point and agree with the general concept of "cloud only bad", and it's true that Enphase may still "alter the bargain further", I am still running an original Envoy and even with the latest firmware it still supports local access. With home assistant you do need to install the legacy support package to talk to it.

In my head I can hear George C Scott saying, "Mr. President, we must not allow and Ass-AI gap!"

We probably don't need gigabit, but we do need a reasonable lower limit to officially be "broadband" and to measure what counts as "Internet being available to everyone". I have a house where the best internet you can get is a dual bonded DSL connection that totals 12MB down and 2MB up. I don't count that as broadband as while its okay for basic video consumption when I am there, it's not enough for video conferencing, and it is certainly not enough for remote security cameras. There are lots of folks where DSL is the only option, and if you are on connections this slow, I wouldn't want that to count positively in "State of the Internet in America" statistics. It's a valid debate as to what the "minimum" should be to count as broadband. I would personally say somewhere in the 20-30MB up and down range.