[00:00] The butterfly effect is much stranger than we  thought. A group of researchers just showed that   if you take into account quantum physics, then it  doesn't just make predictions difficult, it makes   [00:12] predictions impossible after a certain time.  They worked this out concretely for one of the   most complicated chaotic systems we know, Earth's  atmosphere. Does this mean that we can't predict   [00:26] the climate either? Let's have a look. Stress can  damage your health both physically and mentally.   Ask me how I know. If you struggle with mental  health problems or know someone who does, today's   [00:40] paid partner BetterHelp can help. BetterHelp  matches you online with a credential therapist   based on your needs. You can talk by video,  phone, or messaging, whatever works for you. I   [00:52] think BetterHelp is a service that the world needs  because especially with mental health problems,   it shouldn't eat up someone's last remaining  energy to find a therapist. For me, this was a   real obstacle. Yes, I know they've seen criticism  in the past, but they've made changes and now have   [01:09] a 4.8 out of 5 rating in the App Store. More  than 6 million people have gotten BetterHelp this   way. Use betterhelp.com/sabine or my code during  sign up for 10% off your first month. And now,   [01:25] back to the science news. The name butterfly  effect goes back to meteorologist Edward Lorenz   who found out that the weather on our planet  is extremely sensitive to small perturbations.   [01:38] The flap of a butterfly's wings in Brazil could  set off a tornado in Texas. More formally,   if you start with two initial states that are  just a tiny little bit different, then over time,   the difference between them can grow dramatically.  This part of the butterfly effect is quite well   [01:55] known, but few people knew that this actually  wasn't Lorenz's point. His point was what climate   scientist Tim Palmer has called the real butterfly  effect. If you read Lorenz's original paper,   [02:08] you'll see that he was really arguing that weather  prediction suffers from a sort of Zeno effect. If   you improve the precision by which you know the  initial conditions by a factor of say 2, then   [02:20] you can increase the time over which you can make  an accurate prediction by one half. Improve the   precision by another factor of 2, and you add  another one half squared, then one half cubed, and   so on. This means that even if you could measure  the initial conditions to infinite precision,   [02:37] you'd not be able to make predictions beyond a  certain time horizon, even though the system's   fully deterministic. Whether this real butterfly  effect is actually "real" in the sense of happening   [02:49] in nature has never been settled, because Lorenz  didn't, strictly speaking, derive it. He just made   a compelling argument. I found it both interesting  and somewhat concerning that this real butterfly   [03:01] effect has gotten almost entirely lost in the  literature, because if it happens in nature,   it doesn't just happen for the weather forecast.  It'd be a general property of chaotic systems   [03:13] relevant for everything from nuclear fusion to  quantum computing to black holes. In any case,   the authors of the new paper now looked at this  question, whether there is such a finite horizon   for predictions for the weather, and they  found that yes, there is, and it's 129 days.   [03:31] After that time, they say weather forecast  isn't just practically, but even theoretically   impossible. Even if you knew the initial  state perfectly, had a perfect model,   and unlimited computing power, you still wouldn't  be able to get past the 129 days. Their argument,   [03:49] however, is rather different from the argument  that Lorenz made. Lorenz made a general argument   which said that inevitable changes on small scales  like molecules would spread to larger and larger   scales, eventually making predictions impossible.  The authors of the new paper have sidestepped the   [04:07] question of just what happens at the limit  to infinitely small scales by instead asking   what the physical causes for unavoidable  disturbances are and what we know about   them. Their answer is kind of obvious. Earth  constantly receives quanta of light from the sun,   [04:24] but we don't know the exact quantum state  of the sunlight. We can't know it. Quantum   physics forbids it. This quantum uncertainty  first affects very small scales, but from there   it spreads to larger and larger scales until  it has spread through the entire atmosphere.   [04:39] They calculate how long it take for the energy of  the quanta to spread, and that gives them the 129   days. What does this mean for climate models that  make predictions over years or hundreds of years?   [04:52] The maybe somewhat surprising answer is pretty  much nothing because climate models already   deal with averages. They'll tell you the average  ocean temperature as a function of average carbon   dioxide or the probability of certain extreme  events. They won't tell you whether it rains   [05:08] in Paris on July 14th, 2030. The calculation  for the butterfly effect tells you how long it   takes until average prediction is the best you  can do. It doesn't make the average prediction   [05:20] any better or worse. I give this paper a 5  out of 10 on the [ __ ] meter. That's because   I'm not super convinced by their argument that  quantum effects entirely erase the information   of the initial state. Still, I think that this  paper's conceptually an important contribution   [05:37] to the literature and I wish that physicists in  other disciplines would pay more attention to   this. The intersection of quantum physics and  chaos is maybe the most underrated research   [05:49] direction in the foundations of physics today.  When you want to know if there's a fundamental   limit to the precision your quantum computer can  reach before spending a few billion dollars on it,   maybe that's just me. If you try to  predict what someone will say from what   [06:05] they've previously said, how many words does  it take until you lose all predictive power?   Broccoli. Do you think ChatGPT would have made  this joke? Thanks for watching. See you tomorrow.