90% accuracy is terrible
90% accuracy is terrible
Posted Jul 30, 2024 8:24 UTC (Tue) by cpitrat (subscriber, #116459)In reply to: 90% accuracy is terrible by passcod
Parent article: Imitation, not artificial, intelligence
https://newlearningonline.com/literacies/chapter-1/socrat...
"how could they possibly think that words that have been written down can do more than remind those who already know what the writing is about?"
Since then, many new technologies were faced with this kind of concerns: pocket calculators (people won't know how to count anymore), TV (people won't read anymore), computers, internet, ...
Indeed some skills disappear, but they disappear because they are not that needed anymore. It's not the knowledge that disappears (we still know how to add, substract, multiply ...) but the fluency because we don't need it everyday. If for some reason we start needing it again, we practice and get better at it.
Posted Jul 30, 2024 9:42 UTC (Tue)
by Wol (subscriber, #4433)
[Link]
The danger is that there is suddenly a big need for niche skills ...
As a (many people would say unlikely) example, imagine suddenly we lose access to fossil fuel. Okay, wind and solar will take up *some* slack, but most people will be car-less. We'll suddenly need LOTS more horses, and lots more blacksmiths, and and and. The skills are there, but scaling them up rapidly to meet demand will be pretty much out of the question.
There's some social evidence that something like this happened to destroy the Roman Empire. Historians think there was a smallpox epidemic about 200AD, that took out a lot of specialised workers and basically distorted the economy so badly the empire never recovered. Same thing about 1500 in South America. Compare that to the Black Death in the late 1300s, which seriously shrank the economy but because it took out workers pretty evenly across the economy, it didn't distort things that much. The resulting boom became the renaissance.
If anything happens to us, it'll be a Roman Empire crash, not a post-Black Death Renaissance...
Cheers,
Posted Jul 30, 2024 10:17 UTC (Tue)
by farnz (subscriber, #17727)
[Link]
The problem with this is the cost of practising; a good example is nuclear reactor engineering. From the 1950s through 1970s, France had significant programmes to ensure that it had a supply of people who were able to work safely on nuclear reactors, both civilian and military, and who had enough experience to distinguish "this is a transient state, and it'll be fine" from "this is not normal, and we need to worry".
They then let those programs atrophy during the 1980s through the 2010s. They're now training up a new workforce to work on nuclear power plants to replace the people who are retiring, and they're finding that a lot of the knowledge isn't written down - it's stuff you learn from experience and being told "no, that's not right" when you're following a textbook process.
Now, we can definitely recover all of that knowledge over the same timescale it took to build it up - if we started on nuclear power plants from scratch now, and maintained a skilled workforce of nuclear technicians, we could be back in the 1970s state within 20 years. But recovering lost knowledge is often slow, even when you've got it all written down on a piece of paper.
90% accuracy is terrible
Wol
90% accuracy is terrible
Indeed some skills disappear, but they disappear because they are not that needed anymore. It's not the knowledge that disappears (we still know how to add, substract, multiply ...) but the fluency because we don't need it everyday. If for some reason we start needing it again, we practice and get better at it.