Re: Don't get cocky kids
Well, my research was about 25 years ago, so i would have thought they had come up with some better tools in the mean time :)
357 publicly visible posts • joined 4 Jun 2008
Cool, i was wondering if this would be possible for detecting AI generated text.
My university research was on automated detection of plagiarism, using a fairly simple Bayesian classifier to pick out stylistic features of authors and then check a new text supposedly from that author. People really do have detectable stylistic features in how they write, this is often how the papers like 'Computers say Shakespeare probably didn't write King Lear' come about. I trained my classifier on novels from a particular author, then picked a chunk from another novel not used in the training, and inserted paragraphs from a completely different author. it could pick out the inserted passages pretty well.
What i was surprised by was it was also pretty accurate when it hadn't seen the authors work before, if you trained it on the single text and asked it for the outliers, it could find them.
So, based on that, i was wondering if chatGPT actually had a discernable style. If the classifier couldn't find anything useful to build the model with, then you might think it wouldn't work, but it always managed to find something when trained on human text. You could also see the model it was using and what features it had picked for that author, so if it found nothing, it might not be a human.
It would be interesting to see what it did on ai generated text where a human had edited it... One of the main ways to cheat in uni was apparently taking something from the internet, writing a new introduction and conclusion and slightly editing the text. This is pretty easy for the classifier to pick out, but also... lecturers said it stuck out like a sore thumb anyway :D (particularly on non native english speakers attempts to cheat.)
We're all made of protons and every physical thing in our world is too ... OK, I'm off to have a beer full of protons, I guess they must taste great!
So, you must be a fan of lambics... The sour tastebuds basically just respond to the H+ ions floating about in acidic substances, and H+ is just a proton. :)
Technically a Geuze is a young lambic beer mixed with an old one, eg. one batch that has been fermenting for a year with another that's been fermenting for 3 years.
An unadultarated lambic is a wild fermented beer from the Zenne valley near Brussels. Apparently the yeast that makes it a lambic only lives there. And it has to be wild brewed, you can;t export the yeast from the valley and make it somewhere else or it's not lambic.
There is also Faro, which is a lambic with added sugar, but it used to be a way of serving lambics, where you poured it slowly over a sugar cube on a spoon to sweeten it to taste. It's lovely on a hot day.
charcoal fired and ~450C on the stone, not sure about the air temperature to be honest as the air thermometer doesn't go high enough, but the stone is measured with one of those laser things.
Takes maybe 90 seconds- 2 minutes, the dryer mozzarella is melted and nicely browned in that time, and pulls extremely well, but the wet stuff is gooey and browning round the corners but certainly not stringy.
We normally make 5-6 20cm pizzas in a row as the last one is cooked by the time the first one is finished...
Also, cooks bacon from raw to crispy in 90 seconds.
You aren't supposed to use the wet mozzarella on pizza, it should be the grated dryer stuff. The wet stuff takes too long to melt compared to the cooking time of the rest of the ingredients...
The wet stuff is for eating raw. Preferably with a sprinking of fresh basil. But straight out of the packet is fine.
(could be a lie, I was told it by a Neapolitan though, so you never know.)
indeed, but when you go past the end they tend towards 1 sized again but in a different dimension. Egg shaped is actually a kind of infinite hyperbolic surface where the front and back sides of the surface are not connected to each other, and only close to the origin is the back side actually behind the front side.
I did :) I read the documentation from NASA and MIT about mechanical counterpressure suits, which work pretty much as described, the pressure on the body is provided by foam that wants to expand but can't due to a non-stretchy layer over it, this provides enough inward pressure on the body to allow humans to breath air at a much lower pressure than normal atmospheric pressure, and also not explode. It also contains a bladder inside the suit to provide back pressure on the breathing gear to stop all the gas escaping in one go.
I suggest you provide a recommendation for a useful alternative to the breathing apparatus to enable such an unpressurized suit to work instead of saying 'that won't work', because that's how science works.
10 meters depth is double atmospheric pressure, 20 meters treble, etc. 40 meters for a deep dive is 5 bar, which is pretty high pressure. (for me anyway, it's enough to drive some pretty serious air tools. it's not high pressure like in the tank though) So, as mentioned, the breathing gear would be scuba gear that delivered lower than atmospheric pressure. The pressure in a pressurized space suit is about 1/3 atmospheric pressure, and they do breath pure oxygen at this pressure.
I'm wondering if you could have an elastic suit that compressed your chest cavity to 1/3 atmospheric pressure to provide the ambient pressure, probably that can be tightened and loosened in the airlock so it's possible to breath once the air pressure raises again.
Ususally scuba gear has to be quite high pressure to balance the much higher water pressure outside your body, so you'd need much lower pressure scuba gear, enough to balance whatever inward pressure could be generated by the elastic suit mentioned below but not enough to inflate the astronaut?
Sublimation of the gel would only occur if the material is directly exposed to a vacuum though, so some kind of insulated sealed elastic outer suit will be required then, to prevent heat loss to space, evaporation and maintain some pressure on the gel inside by squeezing against the astronaut? Something like a cross between a wet suit and a dry suit. Maybe an air tight inner elastic layer like neoprene, then the gel, then an air tight outer layer squeezing the gel. Quilted in such a way to prevent the gel from moving about to much, no point squeezing it all into the gloves every time you move.
Basically, if you put one of those air free foil pouches of cat food into a vacuum chamber, how long would it take for the cat food inside to freeze dry?
Fill the new spacesuits with gel, then they don't need to be pressurized, which makes them much easier to move in :)
But much more messy to get out of. Showers being a bit of a luxury, if you make the gel from that powder that turns into a gel when wet, then turns back into a powder when dried, you should be able to get clean with a hair dryer and a vacuum cleaner.
Apparently it keeps quite cool too. And the volume of water involved might make a pretty useful radiation shield too, but would probably be quite heavy.
There's probably a million reasons why this won't work in reality, but it's fun to speculate :)
100% agree :) I have an old FinePix S200EXR which lives in my bag, It's not to heavy for carting about everywhere (800 grams) but its zoom and low light performance are amazing, for something that cost me 100 euros 5 years or so ago. I'm not saying everyone should carry around a big camera, but if you want to take good pictures in all conditions, then take a camera because your phone will just disappoint. If you don't, then take a phone, it'll do what you want.
Train multiple CNNs on different things. One on texture, one on sihloetes, one on shape, one on edges. Train them all with a dataset that has multiple views on the same object, so a cat from the front and a cat from the back and from the side and above looking down when the cat is looking up etc.
Same image processed multiple ways each time.
Then train another one to use the output of the others to weight the responses. This supervisor is trained by giving it the set of processed images as one example, and the rotations and viewpoints as sets that are related, so it's tagged as this set and these sets are all the same thing, not these images are all of related things. If one quickly comes up banana and two come up toaster then it's more likely to be a toaster. you could even feedback from the supervisor to say try again, no one else thinks it's a banana and see what I says then.
"But then the disc's rotation speed has to be super-constant and the read:write head movement super-precise so that the positioning, depth and timing of head laser firing can match the nanocrystal sizing and layering."
Why spin the disc? Have a flat square that the head can move across, or better, have the head mounted so the read/write point is scanned by adjusting the optics, rather then move the disc or the head. Or am I missing something?
"Just box it in at the bow with 2 reasonably sized vessels and it'll have to slow down or stop for safety reasons"
A big containership or bulk carrier can take an hour or so to slow down*... don't park too close in front.
An hours warning for the ship's owners to know that something is deliberately slowing the ship. Given that pirates will probably be closish to shore, that's probably enough time to scramble a couple of planes loaded with parachute and gun toting security personnel.
*I learned this from Ship Simulator, may not be accurate :)
Accountants were automated out of a job years ago, but they are really good a lobbying the government to keep the tax and accounting rules changing often enough to ensure that the people who write the automation are always several steps behind and so need a trained lobbist to make sure it's doing it's job...