Computing desk
< January 13	<< Dec \| January \| Feb >>	Current desk >

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.

January 14[edit]

Can ChatGPT design a experiment?[edit]

I ask myself however ChatGPT is able to creat a scientific experiment. 2A02:908:424:9D60:B4AA:6DE:2CF7:65DA (talk) 14:32, 14 January 2023 (UTC)[reply]

Why not ask ChatGPT instead of asking yourself or us? I suggest that you come up with one or more scientific questions, such as, "Is the boiling point of sparkling water higher than that of plain water?", or, if you want to make it more difficult, "Are neutrinos Dirac fermions or Majorana fermions?" (List of unsolved problems in physics may provide some inspiration for tough questions.) Then ask ChatGPT to design corresponding scientific experiments to find the answers. --Lambiam 02:01, 15 January 2023 (UTC)[reply]

I asked it "Design a scientific experiment to determine if the boiling point of sparkling water is higher than that of plain water", and it responded:

1. Materials:
Two beakers or pots (one for plain water and one for sparkling water)
A thermometer
A heat source (such as a stove)
A timer
2. Procedure:
3. Fill each beaker or pot with the same amount of water, one with plain water and one with sparkling water
4. Place the thermometer in the center of each beaker or pot, making sure it is not touching the bottom or sides.
5. Set the heat source to high and place the beakers or pots on the heat source.
6. Start the timer and record the temperature every 30 seconds.
7. Continue heating the water until it reaches boiling point.
8. Compare the temperature at which the plain water and sparkling water reach boiling point.
9. Repeat the experiment 2-3 times to ensure accuracy.
10. Analysis:
If the boiling point of the sparkling water is higher than that of the plain water, this will be indicated by the thermometer recording a higher temperature at boiling point for the sparkling water compared to the plain water.

It seems a fairly reasonable experiment design, although measuring the temperature every 30 seconds doesn't really seem necessary to answer the question. CodeTalker (talk) 02:06, 16 January 2023 (UTC)[reply]

Maybe it's a moot point, but in this case, we might contest whether this experiment was "created," or "merely repeated..."

Quoth the the Plato Encyclopedia article on Scientific Discovery - which I, a lowly human, merely repeat here -

"Philosophical issues related to scientific discovery arise about the nature of human creativity, specifically about whether the “eureka moment” can be analyzed and about whether there are rules (algorithms, guidelines, or heuristics) according to which such a novel insight can be brought about."

One wonders whether the software, in its present state, can actually create a scientific experiment that it has not ever encountered previously... which is, of course, not an original insight, but one that has been explored in great depth in professional discourse and in entertainment media. My mind wanders toward certain quotes in Elementary, Dear Data - "fraud!"

Nimur (talk) 14:22, 16 January 2023 (UTC)[reply]

The question, as put to ChatGPT, was to design an experiment, which is how I interpreted the intention of the question posed by the OP. So we should not be disappointed that the suggested design was quite conventional. (Still, I keep being astounded; see also my request for a more detailed technical explanation of ChatGPT over at the Science section of the Desk.) Can you think of a question, not absurdly difficult but one that many people could answer, that requires humanlike creativity for a satisfactory answer? --Lambiam 18:32, 16 January 2023 (UTC)[reply]

As a software engineer, I am more than astounded at its ability to write functional programs based on requirements specified in (often imprecise) English. I asked it to write a quite nontrivial program that I would often use as an interview question when interviewing job candidates, and its answer was correct and more complete than any answer I ever got from a human candidate. CodeTalker (talk) 18:49, 16 January 2023 (UTC)[reply]

Hopefully you've avoided any of the ones in [1], a lot of computing interview questions and answers are on the web. NadVolum (talk) 13:39, 19 January 2023 (UTC)[reply]

Can you set a human an essay assignment which requires humanlike creativity to complete? Probably not. Card Zero (talk) 09:13, 17 January 2023 (UTC)[reply]

If a cheap VGA phone fell on a corner[edit]

And the upper-left 2cm is a white triangle with black rim of varying width, lines with corner damage are black and normal alternating and all other lines are white and normal alternating with a few minor black rectangles on left edge, then will my keeping it on max brightness for readability cut screen life?

The rectangles grew from nothing since it fell ~12/27 but if area grows linearly "all black" would take a decade, if exponential the screen might be black before the 23 Samsungs cut the prices of the rest.

2. Does anyone here know what causes these interesting partial screen failure patterns? Some vertical striping and white lines in the black lines, some vertical bleeding of what's on screen that instant that actually improved (stopped happening)... Most of the screen is fine right now except every other line is white. Sagittarian Milky Way (talk) 21:18, 14 January 2023 (UTC)[reply]