Welcome to the Wikipedia Computing Reference Desk Archives
The page you are currently viewing is a transcluded archive page. While you can leave answers for any questions shown below, please ask new questions on one of the current reference desk pages.
I just changed the battery on my mouse and now the right button doesn't do anything. I've tried turning the mouse off and on again, and taking the battery out and putting it back in. Windows 11. Help! DuncanHill (talk) 10:35, 19 March 2023 (UTC)[reply]
Happened to me with the left button, and it turned out merely to have broken physically. You can get a right-click using the menu key, although this doesn't work perfectly: for a more genuine right-click, if you have a keypad, turn on "mouse keys" (which lets the keypad emulate the mouse), and press -, and now the 5 key will right-click. Mouse keys is under "accessibility" in the Windows 11 settings. Card Zero (talk)16:27, 19 March 2023 (UTC)[reply]
Oh, so really it was Edge that had capriciously decided to disable right click? Or was it just that you were looking at one of those irritating sites that try to protect their content (and shepherd users around) by disabling it? Card Zero (talk)17:52, 19 March 2023 (UTC)[reply]
AI can answer with text. So, suppose you've got some text from AI; what answer would you expect to your question with respect to that text? --CiaPan (talk) 12:05, 20 March 2023 (UTC)[reply]
Well, it depends what you mean by "artificial intelligence". If your definition is loose enough, it’s really easy.
Here’s one:
#1: the coin flip
import random # huh oh
input_text = "Put your text here"
if random.rand() < 0.5:
print("Methinks it was written by a man")
else:
print("Methinks it was written by a woman")
It flips a coin to tell the answer, but it’s computer code, so it’s artificial intelligence, right?
That "AI" had a roughly 50% success rate. Can we do better? Enter candidate #2:
#2: it’s always men
print("It was written by a man, definitely, no question about it.")
Most well-known books were written by men (for obvious socio-cultural reasons that we need not delve into here), so that performs above 50%.
You want better than that? OK, here’s candidate #3:
#3: the know-it-all
input_text = "Put your text here"
if "Ein Gespenst geht um in Europa – das Gespenst des Kommunismus" in input_text:
print("That’s the Communist Manifesto, written by Marx and Engels, two men.")
elif "I ought to be thy Adam, but I am rather the fallen angel" in input_text:
print("That’s Frankenstein, written by Mary Shelley, a woman.")
elif (...snip thousands of lines...)
else:
print("no idea, that’s not in my database")
That is simply a (poorly-formatted) database of known quotes and the associated answers. Obviously, it only works on text that the programmer knew and incorporated in the program. It seems silly to call that "artificial intelligence".
To produce candidate #4, have a postgraduate student program a neural network and feed it a bunch of texts for which the answer is known (including the Communist Manifesto, Frankenstein, etc.). Now we’re talking, right? The issue is that depending on how you do it, it’s really easy for that method to degenerate into candidate #3 (see overfitting).
My point is that there is no fundamental difference between candidates #1 through #4. "Artificial intelligence" is a nontechnical term that means "emulating human intelligence by mechanical means", but that definition is a moving target. Few modern people would think a calculator to be artificial intelligence, but the difference engine certainly caused quite a stir. Today, we do not think it weird that a free phone app can beat the world champion at chess; less than 30 years ago, Deep Blue versus Garry Kasparov was a shock to many. TigraanClick here for my talk page ("private" contact)14:02, 20 March 2023 (UTC)[reply]
Note that per Card Zero's point below, people have been doing this sort of thing with some claimed success since before Gmail existed and when Deep Blue was still a relatively recent thing. See e.g. [1][2] Maybe even before then. Nil Einne (talk) 15:56, 20 March 2023 (UTC)[reply]
I think the question is not, "could an AI theoretically determine the gender of the author of a given text?", but, "has anyone actually trained an AI system to do so?". I do not know the answer. It does not seem a terribly difficult thing to do, but the success rate may strongly depend on whether the collection of texts it has been trained on is sufficiently representative of the texts submitted for gender identification. --Lambiam19:09, 21 March 2023 (UTC)[reply]
One would think that actual humans would, at least currently, be better at this task that "AIs" (I don't think a true AI has been invented yet, so deprecate the term's use). I am reminded of the speculations about the real identity of the rumoured-to-be-pseudonymous author James Tiptree Jr. and the widespread dismissal of suggestions that the author might be a woman, with Robert Silverberg going so far as to assert, in an introduction to a collection of Tiptree's stories, that he found the theory "absurd, for there is for me something ineluctably masculine about Tiptree's writing." Shortly afterwards Tiptree was revealed to be the not-unknown writer Alice Sheldon. {The poster formerly known as 87.81.230.195} 5.64.160.67 (talk) 01:23, 22 March 2023 (UTC)[reply]
true aiThe foot-in-mouth Silverberg anecdote (for Silverberg himself about the issue of whether such things as “masculine” and “feminine” fiction exist, see here) goes to show that humans are not terribly good at this, so I would not be surprised to see a news headline any day now telling us that "AI beats humans" at this task. --Lambiam14:15, 22 March 2023 (UTC)[reply]
This is a difficult question. You can probably do alright for historical texts, but that may be based on incidentals. Women write who-dunnits and romances, men write steam engine manuals and battle reports. If an AI can systematically determine the gender of the author of modern texts written to the same topic is a different question. --Stephan Schulz (talk) 15:20, 22 March 2023 (UTC)[reply]