Artificial Intelligence Predicts Earthquakes With Unprecedented Accuracy.
Researchers at the University of Texas have developed an AI that predicted 70% of earthquakes during a trial in China, indicating potential for future quake risk mitigation. The AI, trained on seismic data, also ranked first in an international competition, underscoring its effectiveness and open
Read in full here:
This thread was posted by one of our members via one of our news source trackers.
One of the better and useful uses of AI. Glad to hear they are using AI for the good.
Now imagine what happens if the AI makes one single mistake.
Not sure what you mean. Wouldn’t this help minimize loss of human lives?
Yes and no. Look that we don’t have only 2 cases here, for example:
If algorithm is right then it minimises losses
If algorithm makes mistake and warns about earthquake then, depending on who have access to it, local economy of some province may have unbelievable losses (like temporary suspended contracts). Every business that can’t be moved could have terrible losses.
However 2nd point is only about money. In case algorithm makes mistake and do not wan about earthquake then people, used to algorithm, would not be prepared or even ignore alarms until it would be too late.
Please pay a close attention in what kind of world we live. It’s not “black and white”. People are rushing from one edge to another one (politic, healthcare, trust). 70% is extremely high value to minimise loses (point 1), but 30% is extremely low, in our world unacceptable, value (point 2 and 3).
Let’s take a look at Japan. They have earthquakes very, very often (if not everyday). There are so much earthquakes that people do not react on the weakest ones, but are terribly scared if they would become stronger. The common practice there is to have an escape bag.
Now consider that such disciplined people would suddenly start relying on such tools as if they were oracles. Depending on earthquake thousands of extra people would die. I would say that discipline in Japan is above said 70%, so if we would compare 2 alternative worlds with and without algorithm then it may be that in a world with an algorithm there would be more deaths.
So we already have potentially more death + money loses on fake alerts. However it’s not all! Each earthquake is different. One is stronger than another. One takes (relatively) short time and after another one aftershocks may happen even few months after initial one. One earthquake in Japan caused the eruption of Mount Fuji that took 66 days! Currently eruption like that could cause a collapse of Japan economy. Now think that in such case “AI” says:
There would be a tiny earthquake soon, please ignore.
Well … earthquake “guessed” correctly! Wasn’t tiny? It have 70% chance to guess correctly and it did so and rest are details - of course from the algorithm side. It’s only around 100 km from Tokyo. Over 40 million affected people. Let’s say the chance to die as same as chance for a fake i.e. 30% - that’s over 12 million of people. Of course that’s a dark scenario, but it’s really possible to happen. I have no idea what real losses could it cause with or without AI, but better if people would think for themselves in such important cases.
Ah and by any chance … If you think that some people would be smart in such case then think again what their boss would say after seeing same alert. Again in Japan there is a very big discipline. If people would rely on possibly false results the consequences would be terribly hard to count.
Exactly! I couldn’t have said it better myself
Good explanation on this. And I agree.
As a geologist I know how hard it has been to predict earthquakes accurately. If AI can achieve even just this, it would be a game-changer.
You are a geologist and a software engineer? Cool 
Yes.
