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.

Well, it's a survey so I'll add my YES to the headline question. I got fed-up with ChatGPT but quite like Gemini. I mostly use it for things I could probably do myself but just don't want to invest the time getting it close to right. Last night it was comparing the output signal level on two RF transmitters that should have the same designed output power but one is reported received better than the other. Seeing a 10dBm difference between their output on a Spectrum analyzer I knew the well received one has more power output than the less well received one. So I just asked Gemini to give me a number of Watts from dBm. I know how to do the math but it was just quicker to see the approximately 10dBm difference expressed using the actual peak dBm values from the spectrum analyzer as 3 Watts versus 26 Watts for the transmitters. I'll probably do two tone PEP measurements in a similar way today to verify it. Gemini has also been good when asking for an example developing with something like Qt and an object or feature of an object I've not used before, or even being more confident about basic C/C++ assumptions, CPU behavior: For linux, in c++ with Qt 6.8 show me an example of layered drawing on a QGraphicsScene (I have to convet it for my use case but it saves me hours of reading just to get started, and I do turn to the docs but with a better undersdtanding than NULL of what I'm reading); or reminding (re-assuring) myself that no element of a for() loop is executed if the conditional expression is false at the start of the loop; what allignment is required for the destination of the Intel x86_64 movaps instruction (the cause of the customer's random SIGSEGV verified from the answer and the registers seen in gdb at the point of failure). I have asked Gemini to write a loan contract between me as the lender and a couple I loaned some money to but I read it in full, it was basic but functional, so used it because it set limits in my favor I wouldn't have thought of. I have epilepsy and and have asked many times for what I'd call starting information on very detailed topics of interest to me (what are sodium channels; and what are potassium channels) which gave me enough starter information to use when reading search results that are published medical papers and improved my understanding, so that I know at my next neurology visit I'll be asking if it's most likely I have an untreatable (probably genetic) EAAT-2 protein disfunction that results in periodic (28 day) EAAT-1 protein build up that becomes easier to trigger into seizures as it builds up. I liked the beautifully hedged answer to does playing soduko rugularly reduce loss of mential function with age and opted to play crosswords as well based on the hedge. Yes, I have asked why the sky is blue a few times on ChatGPT and Gemini, but I already knew the reason. Alexa probably isn't really AI by it's own admission in answers but I haven't been able to resist asking "Alexa, what's that smell" a few times.

Well said, the sole thing I used to feel I was missing from the Windows arena was MSFS until I tried FlightGear on my linux systems earlier this year. I ordered a USB joystick and throttle lever the same day and have never looked back. The Airbus A320 simulation in FlightGear seems a thing of beauty to me. Believe it or not the only thing Windows has that my sole use platforms of linux, Android and iPod don't have is the Android/Wear Watch Face Studio program for making watch faces for Wear based smart watches, but I really must try it on WINE before calling that a failure. Winwhat? I never use it and don't feel I'm missing out on anything. /SingleUseCase

AI is in dire need of a bot that speaks at once to groups of people together with shared problems, has a welcome prompt that is the text "this machine kills fascists" and instead of telling each person how easily they can be the winner over all others helps the group use their combined capacity because it's greater than the sum of their individual capacities. openFolk AI, it's got to happen or we will all just retreat further into our own corners dying assured that we beat over everyone else. It's easy to do because the answer my friends is "blowing in the wind". Hippocrates is probably turning in his grave at the headline news story for this and it can only get worse in medicine (and other life necessities) unless we work to put it away somewhere in it's own corner of irrelevance and find an AI model that more brings us together as communities instead of isolating each individual of us. /Offered seriously with intentional sardonicism and satire.

Agreed, I did claim the tax rebates for installing my rooftop solar, garage battery system and high UV reflective windows but the tax rebates were never a consideration when deciding to install any of them. I had the available savings to pay cash, wanted to stop paying up to an unknown few hundred dollars per-month for grid power, i.e. I simply re-position existing savings into my property, raising it's resale value and it permits me to deposit more to savings per-month thereafter from reduced power bills. My background usage is 0.5kWh (1706.07 BTU) and peaks at about 2kWh (6824.28 BTU) for minutes long periods based on home appliance usage. I admit I now use all the data from the IOT parts and schedule appliance usage to minimize grid consumption. I have a dual-flow electricity meter and submit a varying amount into the grid up-to about 4kWh (13648.6 BTU) that obviously varies all daylight hours. I run off the battery for most of the night and get about 97% self-sufficiency from April to October with regular electricity bills of around $10 per-month (cost of having the meter) even running my air-conditioner in the hottest days. The new windows should reduce that usage and seem to help reduce winter gas bills as well, the final few should be installed in the next month, they were also paid cash from savings with an assumed improvement in my home value (though probably not the whole cost of the windows) but reduced monthly costs meaning more avaailable for me to save per-month thereafter. The peak winter electricity bills are about $40 because the heating source is LNG and electricity just runs the fan on short, lower light days (winter heating runs mostly at night anyway). The power company chooses to pay a lot less for received energy from my solar system than it charges for delivered energy I use from the grid. I'll probably double the size of my battery system and maybe add a few more solar panels some time after the last of the high UV reflecttive windows. I'll do it regardless of tax rebates, aiming to hit 100% self-sufficiency for electricity usage at least.

Add creator/owner/intended-audience identity as a property of the entry in the file-system directory and that isn't necesarrilly true, a single English/Western language speaker would more sensibly tend to use Temp1, Temp2, Temp3, Temp4; or tempA, tempB, tempC, tempD; etc. Add multiple user access to the directory with no restriction on the language spoken by any user and your point isn't necessarilly true. A better approach is that those properties exist outside of the filename itself, though still as directory properties for the file (filename metadata) but we are effectively shuttered by the idea of a fopen() function taking only one argument that specifies the file's entire identity as a filespec or filename. Why shouldn't it be possible to open file "Temp", "English", "UK"; then open "tEmp", "German" "CH"; then open "teMp", "German", "DE"; then open "temP"; "Lemerig", "VU". Where each of the four filenames means something in it's native language/country/audience combination while appearing nonsense text in the other three language/country cases. All the time having a desktop file manager that shows the four files as paired icons, one representing the content type and another being an icon of the flag for the country of the creator/owner. I should have no care that the Lemerig language has a meaningful word spelled temP (which isn't a true example of Lemerig, but illustrates a big part of what the problem really is). At the tty console it would require a dir/ls command to show the language/country information as part of the output in order for each filename to be properly understood. Obviously, I agree you are probably correct about the four names you used but the problem doesn't end at such a concise set and there will be combinations that come very close to having meaning in other language/country combinations, perhaps differing and even opposing meanings. Who am I to say that Temp is not the exact line representation of a very personally directed abusive swear word in the Lemerig language on the island of Vanatu.

Isn't the proper terminology for this known as "circling the drain"?

Amen! Just one example, a few years ago I bought a scientific calculator noting the package description of the "lifetime software license" it came with and quite decent software that requires the calculator firmware be maintained if it developed, "tear here to accept the license terms and access the installation CD/DVD". I noted they recently upgraded to python for user coding on the calculator and upgraded. The manufacturer just dumped using the lifetime license with no regard for anyone that had already purchased and agreed to the existing terms. They switched every user to a four year plan with no negotiation, "we don't use a lifetime license any longer". I called to negotiate and they gave me an extra four years...at least I expect to retire in 8 years.

Yes, that's the point of the OP and I agree with you. My points were only about the presumption of guilt in a case where it could be shown that the accused had used an AI system to gain sufficient knowledge to commit a crime. Which is a different matter from the OP's point but the OP and my point raise substantial questions that will have to be answered in court regarding whether an AI system can found liable in a crime (defamation in the OP's point) but aiding and abbetting in my point.

I love Hubert Dreyfus... I only discovered him after Googling for examples of philosophiical research into artificial intelligence. There's a greap wikipedia article specifically on his views on artificial intelligence: https://en.wikipedia.org/wiki/.... Goog luck to all AI research but I suspect it wiill utilmately fail as a concept if it doesn't heavily use actual critical philosophers during design.