Why not do this sort of thing in the license, whether or not it's practical to enforce it, at least it's on the legal record. We're just starting to see book authors take up cases regarding the scraping of their copyrighted content into LLM's. The reverse problem applies to the average coder using AI. Starting high level, but "how do I write a program to capture images from my [named PC model]'s built in webcam in C++ on linux 6". The fear for many in the openSource development community is that the offered answer to such questions, if used as-is, may be something someone else holds copyright to. So, let GPL4 include a clause stating "the covered material by this license may not be used for any purpose by any AI agent or tool other than as a reason to do no more than refer the questioner to the home site of this project". I already see how it breaks with the existing claims that the project is not provided or waranted for any purpose. Or, does GPLv3 already protect against AI abuse of copyleft in prohibiting the conveying of work for payment to a third party that is in the business of distributing software. I mean, the electricity supplier, the cooling water supplier, the building leaser and more are all in the business of distributing software even if it is not their primary business. By paying your rent on-line for an AI data center meanwhile returning answers that are rote copy of source segments from GPL3 source is your AI perhaps breaking the GPL3 anyway. Good luck to the book authors, perhaps it will have the side-effect of having AI platforms reduce their own creation of situations that abuse authors of any kind. I note already, none are proposing AI based government representation for individuals and AI writing of laws, there seems to be no reason they wouldn't be as capable of that as they are of giving webcam image capture code, I wonder why...

Siince i can't edit the incorrect quote already posted, crappy tablet book cover keyboard does it a lot, the corrected quote is "Hey farmer put away the DDT now. Give me spots on my apples but leave me the birds and the bees"

Grocery stores already do that as proxies. Two examples I know of are DDT and Paraquat. To quote a wonderful poet "Hey farmer, put away the FFT now. Give me spots on my apples but leave me the birds and bees".

The general discussion reminds me of a wonderful storage technilogy made from celulose fibres and cut into flat rectangular platters, typically sold in packs of 500 platters. The platters could store data on both sides and in any orientation of, any useful size and position on the platter. One of the best methods of storing data was thin wax sticks that were pointed at one end and often available in multiple colors. However, those were difficult to use for high data density and the data wasn't re-writable using those. A much better approach was a wooden stick, containing a graphite core and pointed at one end but often included a tool at the other end for erasing existing data, permitting the storage platters to be re-used multiple times. Unfortinately, copying data from one platter to another was generally a very laborious manual process for many hundreds of years before the 20th Century.

On the one hand I've had a fun week adapting a personally written multi-mode audio signal monitor to add seismic monitoring with some guidance on limiting requirements by Gemini, though I designed and wrote the code myself using Gemini to check a few corner cases that failed unexpectedly. But seriously, why have human representatives or leaders with personal prejusdices and prejudices that change solely from losing some "argument". Add lawyers, judges and jurors to that list. There's enough published material for LLMs to be able to write new law and law that develops as history progresses if the input material has a known date of acceptance into law. AI won't need a two week Easter break (insert your own preferred religious holiday instead of Easter when your representative is taking a break). AI can work 24/7 on governing. There are ways to ensure the outcomes "represent the will of the people", and it won't care if it wins or loses the vote. Ironic that nobody who is buying their preferred leadership in government has been observed making the case that AI could replace those who they are paying.

Absolutely true! The US has a very different population density environment from the UK and most of Europe. Just to visit the nearest metropolitan city from my US home is at least a six hour round trip by public transport. It's only 60 miles away. When I lived in the UK I was about 45 miles from the center of London and could travel to/from work there in under 3 hours round trip. I only needed to do that for work, the health and other services were all within a few miles. In the US, the only practical eplilepsy clinic I can attend is the six hour round trip to the nearest metro hospital, an eight or nine hour day including the time in the hospital. Without exhorbitant government investment the US couldn't create truly practical public transport for most states. In the UK, I lived in a two or three bedroom house but there was probably nowhere in the house that was more than 25 feet from the indoors of the neighbor's home on either side (high population density). In the US, every indoor position in even my modest five bedroom home (where I live alone) is probably no closer than 40 feet from the indoors of neighboring homes (low population density). There are miles of open land between cities (towns) between me and the nearest metropolitan city. Which is also the only metropolitan city in a state that is about 85000 square miles (about 220000 km squared). The area of the state is 93% of the area of the UK and has one metropolitan city. The population of the US is about 340 million in 50 states, most of which are nearly as big as the UK which has a population of about 70 million. Or, 340 million people in about 380 million square miles in the US, so a little under one person per-square mile. In the UK, it's 70 million people in about 95 thousand square miles, or 735 people per-square mile. On a cost basis alone, public transport works really well in the UK. Pickup trucks work much better on a cost basis in the US, honestly. Obviously there are huge metropolitan areas in the US with better public transport than I have but they are truly tiny islands of the country as a whole. Equally, obviously there are very isolated rural areas in the UK but they too are truly tiny islands of the country as a whole and it's easy for nearby higher density population areas to subsidise publilc transport through the truly rural areas. That's just not practical without exhorbitant government underwriting of it in the US.

That reminds of a radio story broadcast by a real news station about being stuck on hold for a long time with an automated receptionist. The automated phone call routing system "expert" being interviewed suggested that the best way to get through to a human is to say nothing more than a single word that makes no sense with respect to the business you are calling or any reason you might call. So, as long as it's not a banana related business say only the word "bananas" to the automated receptionist every time you are asked anything, no matter what you are asked. The interviewee reported that most systems will give up quickly and route you to a human. I assume plural nouns or even one word of a name unrelated to the called business form the best set to choose a word from because many verbs and adjectives could be used by the system to infer the experience making you call leading to an automated answer from the customer help database. The advice was to use the same ridiculous single word every time asked anything by the automated receptionist on a given call, i.e. even when the receptionist says "sorry, could you provide more detail" only say "bananas" again in response. I'm sure the automated support loop from the article was more online rather than a phone call. But even telling the online front-end to it nothing about your problem and nothing that could be used to infer a reason for your use would leave it helpless unless you were routed to a human. Please enter details of the problem you are experiencing? "bananas" That might get you help about how to use the default bananas picture when adding a desktop icon for something but there are will be more confusing words: "Asimov", "Gandalf", "toupees", "ligaments", "ankles", "joists", "cranes", "peppermints", "haddocks" and so on. I suppose there's more of a risk with an online support system since if it did give up and send to a human the text you entered as a description of your problem someone reading "haddocks" might be inclined to dismiss the message as spam but still...

For the majority of cases that could occur that only requires a way for the system to be updated with active "notices to airmen" (NOTAMs) before flight. A closed runway is (other than for cases like this landing itself) a planned event and commonly announced in advance via NOTAMs. So, during pre-flight, plug the system into your cellphone via USB and press the "NOTAM update" button and it should be aware of all practical runway closures en-route. That's likely to leave the cases where a runway chosen by the system is closed to be vanishingly small, though not quite zero. Take a look at notams.aim.faa.gov

Never mind long distance HSR construction. Use short rail stretches to the peaks of the high mountain rainges in the US, the continental divide being the perfect example, Accelerate on rails up those to a peak velocity that allows the connected vehicles to extend wings and disconnect from each other at the peak becoming individually airborne at the peak then glide to the destination, as much as possible using the glide angle to keep the speed high. It would have minimal environmental impact, only requiring sufficient land clearance and ramp smoothing either side of the mountaintop launch points to permit a comfortable ride and extension of the wings before the peak. The gliding part is essentilly energy free, though probably the cost is in reaching a suitable velocity to become airborne for a hundreds miles of gliding. The rail lengths would obviously be much shorther than needed for city to city HSR and there would be no concern about integrating crossings with other transpoort methods, just build the launch rails on "unused" mountainsides. Where I live in the US the launch angles would be excellent (easily averaging 75 degrees or higher), perhaps even making it possible to be used for rail driven escape velocities for orbital vehical launch. Steering could be done mostly after becoming airborne with only the need for the short rail sections generally point in the direction of any destination. Departure would be a thrill everyone would seek and the flight much quieter than any powered aircraft and leave no in-flight pollution. Children would demand their parents take them on a trip using this system in preference to any other vehicle. Where I live, a round trip to the nearest useful hospital is 60 miles each way for an extremely useful 20 to 30 minute bi-annual check-up appointment. By public transport it's a 10 to 11 hour round trip but by car it's about 3 to 4 hours, I have epilepsy so have to use bus and train. The main reason for the difference between public transport and drive times is the large number of widely dispersed discrete municipal communities in a single state as large as the UK making co-ordinated national or statewide public transport not a realistic possibility for more than a small percentage of the population. Therefore rather than building any HSR networks I vote for short length glider launch ramp construction each side of the major US mountain ranges, particularly the contental divide.

Things like the Google Thermostat, my fridge and dishwasher allow me to remotely control them when not at home, so the apps for them have central links that the devices and apps for them link to. My Amazon Echo devices do all language decoding only if connected to the Internet, I couldn't ask it to play my choice of music or radio and get a response unless it has an uplink. Most of my cheap smart lightbulbs and smart power plugs also have central "registration" servers that the control apps link through back to the device. The better quality ones will work via Bluetooth or Internet and can work when there is no uplink (I'm in the process of at least replacing the cheapo-foreign bulbs with a better quality US brand). There's no actual inbound connections from the Internet, all home networks are RFC1918 networks using NAT on a router with a static public IP but all the IOT devices make regular outbound connections, usually TLS encrypted using ephemeral source and destination ports. Which is what adds up to 2Mbps of outbound traffic. Since Apple Music lossless ALAC encoding has up-to 192kHz sample rate, 24-bit data and at least two channels and plays on the Amazon Echo devices the download data rate has to be at least 9.2Mbps to avoid drop-outs, with advice I've seen saying double that limit...just in case. Almost every recorded music in existence would be up-scaled to 192kHz but much modern recording is probably 24-bit/two channel. Apple Music probably requires 4x any useful data rate for normal operation. Some will claim interference patterns in the above audible-frequency-range can generate harmonic sounds within the audible-frequency-range so a 96kHz Nyquist frequency is useful, I can't say that's wrong. Radio streams, especially voice radio will be much less intensive. Even idle my outbound data rate from IOT devices alone is about 2mbps. Using 9.2Mbps down and 3072Mbps up tc limits configured via wondershaper only for the IOT NIC means I have no Apple Music drop-out on Amazon Echo devicees and tc stats show no dropped-frames or over-limit counts. So, sure, I limit the IOT use to 0.3% up and 0.92% down from my whole 1Gbps link but I suspect that even 3Mbps/9.2Mbps is productive for the headline post's botnet.

After getting worried about having smart lightbulbs, smart power sockets, a fridge, a dishwasher, some amazon echos and a Google thermostat on my home network with my computers I re-wired it all and put all the IoT devices on their own wifi, using their own IP network, their own cabling to their own NIC on the linux router and blocked traffic from that NIC to anything other than the Internet. Now I think I'll try to add a data rate restriction for the IoT network. Other than playing music on the Amazon echo devices I doubt it needs even 256kbps to be responsive. I supposedly have a 1Gbps Internet connection but it rarely delivers more than a few hundred Mbps downstream, which is fine even for video streaming but I figure after reading this why have the IoT devices able to have remotely close to as much Internet as I want for my computer networks.

But taking your simple example as a reference, what if slashdot wanted to reduce the use of the word FUCK in comments but used a generated data set with only the same distribution of ages as the real slashdot userbase that you suggest could be created as an equivalent data set with no association to individual users. Then, you "see" an apparant association between using the word FUCK in comments and the age of a poster being between 25 and 35 years. All very well, you put the extra CPU cycles into checking only posts from those aged between 25 and 35. However, what if the trigger is in another property not in that data set, e.g. the combination of gender and ethnic origin. What if the real trigger is white male readers, with no age correlation at all, the apparant age correlation only speaks of the volume of readers of a given age. What's needed ffor"clinical" analysis is not then in the synthetic data. Someone else already commented that this is a problem with synthetic data sets. The interpretion of it is walled in by the set of parameters it contains but a true medical record has a very large set of parameters, including some that would be expressed almost as implications in medical records rather than as absolute facts. These attributes might be the difference between someone diagnosed with gout due to swollen feet.heels and their real illness being radiation poisoning from having growwn up living close to the tipping site used by a Uranium refiner. I'd be less concerned about whether the generated data set is ethical but truly representative and more concerned about whether it is capable of truly containing all signals and markers if being used to create a new generation of medicines for gout...for example.

The only real way you can get "value" or a "bargain" on Amazon is to go there knowing what you want but assuming you don't "need" it delivered tomorrow and monitor the price maybe for weeks or months. Sometimes you can get what you "really want" at as much as 50% off if you can live with being patient but that does mean checking the Amazon web site pretty much daily. Any time that happens I can't help wondering if I'm looking at the pre-sell-off before version next of the same product is about to be released and ask Gemini if there are any rumours about a new version release. I wonder if the cost of that exceeds the value of any one day price reduction... Otherwise the benefit is that the typically quite small extra cost of using Amazon is the saving from not having to leave the armchair to go get it locally. Agreed on the Chinese trash pit though, it wastes a lot of time scrolling past junk clone versions of something to find decent quality products...unless you know a specific brand and model name.

Sorry, that has few typos from the tablet keyboard that I can't edit but you should get the point.

Reading the wikipedia article on "Seismic wave" is worthwhile. Ambient noise can be filtered (not stated but I'm sure pocket rocking is included). Since the differerent seismic wave types travel at different speeds depending on the material between source and your phone then, as was said , it only really would need a "mesh" of 3D accelerometers at known locations (GPS phone position) because that indicates the material between any source and any given phone (from basic 3D earth maps) and the decay in amplitude to predict whether a given phone user is likely to feel ground sshakingfrom a given earthquake. The speed of travel of the waves can be up to 5000 meters per second (or about 1100mph) so the ground motion propagation is probably magnitudes shorter than the comms time between an two devices within the mesh. I have to agree about how the implied disturbance to my privacy. I've used my phone accelerometers to measure Ataxia when I had it recently, I could see in a 3D accelerometer app that it was on a single axis (my elbow only), regardless of hand rotation and at 9Hz IIRC. If that can also be filtered from ambient noise then how many other "back and forth movement" actions might be filtered with range and direction idenitified from multiple sources, e.g. two partners' phones in the same bedroom while....well, you know. I predict the app Google Infidelity Detector within 10 years.