Science desk
< June 1	<< May \| June \| Jul >>	Current desk >

Welcome to the Wikipedia Science 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.

June 2

Why females produce androgens

If human embryons of both sexes start off from a female blueprint and given that females lack the male Y chromosome, how it came that women also produce androgens (even if in small quantity), with related limb and facial hair? 212.180.235.46 (talk) 19:59, 2 June 2024 (UTC)[reply]

The article you linked says that the ovaries also produce androgens. ←Baseball Bugs ^{What's up, Doc?} carrots→ 22:33, 2 June 2024 (UTC)[reply]

Biological systems weren't built by any sort of logical designer. They in no way resemble a computer program, a computer, or, for that matter, anything else in the universe. In the case of androgens, the article mentions that androgens are the precursors to estrogens. Males need estrogens too, btw. All these are steroids, which are fundamental to life and are derived from cholesterol. Abductive (reasoning) 23:18, 2 June 2024 (UTC)[reply]

"Biological systems weren't built by any sort of logical designer." Which is why so-called "intelligent design" is just nonsense. The "design" is emphatically not "intelligent". --User:Khajidha (talk) (contributions) 12:07, 3 June 2024 (UTC)[reply]

I'm not sure biology and computing are as far apart as they used to be now that systems essentially create the gigantic opaque function that transforms input to output themselves in response to their environment/what they have seen and the objectives. Generative adversarial networks for example seem a bit closer to biology than systems used to be. Sean.hoyland (talk) 13:06, 3 June 2024 (UTC)[reply]

Not in a billion years. Abductive (reasoning) 17:51, 3 June 2024 (UTC)[reply]

The technology for synthetic biology, still in its infancy, is advancing with large strides. Whether you like it or not, sooner rather than later it will become possible to design and create complete viable and functioning biological organisms. --Lambiam 06:40, 4 June 2024 (UTC)[reply]

Further, the idea that it is so complicated that nobody understands what it does because it thinks for itself is a farce. No matter what is being used for the computing hardware, be it electronic or biological, the mechanism of operation is very well understood by the engineers who developed it. It just sounds cool to say that it is beyond comprehension. It doesn't sound cool to say that the engineers understand it very well and could trace input through to the output if they wanted to, but simply don't care to do because they have other things to work on. 12.116.29.106 (talk) 14:42, 4 June 2024 (UTC)[reply]

Tracing the input through to the output is IMO not a helpful concept. Not only do we not know why the most advanced chess or go playing programs make certain surprising moves, but it is not even clear what it means to "understand" why they did this. The computing platform performs a calculation with a certain outcome. The engineers can perform the same calculation by hand, or using abaci, and if they make no mistake they may arrive at the same result in a few billion years: 42. But can they say more than that the answer is 42 because this is the consequence of the rules applied to the input? They knew that already. If someone wants to know why it is the consequence, they can tell them to repeat the calculation. --Lambiam 15:51, 4 June 2024 (UTC)[reply]

IP, LLMs for example involve activations in very high dimensional spaces. Trying to map those activation patterns to things we can understand, like concepts etc., is the whole field of interpretability, and it is in its infancy (and safety critical). Engineers are still far from understanding why input A to a model is transformed into output B. If this is something that interests you, have a look at the work being done in Anthropic's lab. Sean.hoyland (talk) 17:16, 4 June 2024 (UTC)[reply]

Overgrown Markov Chain models are pretty much useless in any technical field. BTDT Greglocock (talk) 04:48, 5 June 2024 (UTC)[reply]

I guess that's one of the reasons why "Attention Is All You Need" turned out to be such a great title for a paper. Sean.hoyland (talk) 08:39, 5 June 2024 (UTC)[reply]