“In any moment of decision, the best thing you can do is the right thing. The worst thing you can do is nothing.” -Theodore Roosevelt
It's been an incredible week here at Starts With A Bang! We've covered everything from the night sky's newest supernova to the possible fates of the Universe to what might be our window into the next unexplored mysteries of physics. Moreover, I'll be giving a public talk in Seattle this Wednesday, May 24th, for Astronomy On Tap, at Peddler Brewing Company! Join me if you can, and I promise you'll be left with plenty to wonder about! With that out of the way, here's what we've covered this past week:
- Can the Universe still end in a Big Crunch? (for Ask Ethan),
- After 50 years of missions, we're finally ready to know: is there life on Mars? (for Mostly Mute Monday),
- The odds of your unlikely existence were not infinitely small,
- Closest supernova in years brings cosmic fireworks to Earth's skies,
- NASA's idea for a space station in lunar orbit takes humanity nowhere, and
- Nuclear physics might hold the key to cracking open the Standard Model.
After a lengthy series of exchanges, I've decided to not continue the discussion on climate science, since I think everyone has made their perspectives quite clear and I don't think there's anything more to be gained. But as new topics come up, so do new discussions, and that's part of what makes this series such a wonderful source of extra science on top of everything else we've discussed this week. So let's get right into our comments of the week!
From Denier on the approaching singularity: "Machines are going to become smarter than humans. It has nothing to do with society being a predictable linear system but instead draws from the reliability of Moore’s Law. It is going to happen. Your cell phone will be a smarter being than are you or me or any human who ever lived, and sooner than you think."
So you may not know this (or care), but I spent years working for an artificial intelligence software company. I've seen firsthand what artificial intelligence is good at and capable of, but also what its limitations are and how reliant it is on human input in order to avoid getting total nonsense out of it. You may have seen, recently, that three well-known actors took the first AI-written script for a science fiction film and acted it out. I think it does a good job of exposing what AI can and cannot do today.
What about when we get there? How will Moore's law continue? We don't have a physical concept of how that's possible; we are limited by particles and the speed of light. Quantum entanglement can't beat it. But if somehow we do construct a machine that's more intelligent or capable than human beings in every regard, what will we do with it? Will we just let it make all the decisions for us? Sabine Hossenfelder weighed in on this, saying:
I think it's highly unlikely we'll ever let an AI decide what is "good" for us, but we'll almost certainly ask it how to best achieve what we think is "good"
Given how poor we are at agreeing with each other, even when presented with the same suites of evidence, I don't see the same future that you and Ray Kurzweil envision. But you never know.
Possible fates of the expanding Universe. Notice the differences of different models in the past. Image credit: The Cosmic Perspective / Jeffrey O. Bennett, Megan O. Donahue, Nicholas Schneider and Mark Voit..
From John on the difference between the rate of expansion and the acceleration of an individual object: "“There are two things we can measure when it comes to the Universe’s expansion: the expansion rate and the speed at which an individual galaxy appears to recede from our perspective. These are related, but they are not the same.”
Good stuff! Thank you for the clearly expressed review of the differences."
Thanks John. If the Universe were empty, which would be the same as if it were curvature-dominated or filled of cosmic strings, the apparent recession speed of a distant galaxy would remain constant over time. As it gets more distant, for its speed to remain the same, the expansion rate (a speed-per-unit-distant) must drop. It would eventually drop to zero asymptotically, but never get there. (For the physics/dark energy buffs, its equation of state, w, would equal -1/3. For matter, w = 0, for radiation, w = +1/3, and for cosmological constant, w = -1.)
A standard cosmic timeline of our Universe's history. A series of extremely unlikely events all needed to occur in order so that you would exist. Image credit: NASA / CXC / M. Weiss.
And from Jonathan, wanting to know a little more detail about how it all works: "Hang on! You say at one point that the expansion rate goes down, but near the end you say that current evidence overwhelmingly supports that expansion continues at the same rate forever."
Let's examine the possibilities for how the Universe would behave if it contained something other than pure emptiness. If you're tipped towards matter/radiation, your expansion rate drops to 0 more quickly and your distant galaxy speed drops; if you're tipped more towards a cosmological constant, your expansion rate drops more slowly and your distant galaxy speed rises. If you're equal to a cosmological constant, your expansion rate asymptotes to a finite value and your distant galaxy speed accelerates; if you're more negative than a cosmological constant, your expansion rate reaches a minimum value and then rises, and your distant galaxy speed accelerates at an accelerated rate.
What we observe is consistent with the Universe being dominated by a cosmological constant, but it still contains appreciable fractions of matter. So the matter gets less important (and the expansion rate due to that part drops), but the dark energy portion remains constant, and that's why our expansion rate is asymptoting to a finite value: around 45-50 km/s/Mpc.
From Frank on the future fate of our galaxy: "I think if expansion of universe stops someday all matter would join into a single black hole (surrounded by a cloud of dark matter) in the end, because of gravity."
We observe this scenario in miniature: all gravitationally bound objects have done exactly this on their own small scale: overcome the expansion of the Universe. But what does that mean for these bound objects? Do they collapse down to black holes? No, and they never will entirely. Sure, there are black holes, and some migrate to the center of the galaxy where they become supermassive black holes. But they're only a small fraction of the mass of the galaxy, and will never become even the majority, much less "all" of the matter. Gravity is just one force, but other phenomena such as angular momentum and gravitational interactions between various masses will cause the overwhelming majority of masses to be someday ejected from the galaxy through the process of violent relaxation. It may take ~10^20 years for most masses to get ejected, but they will, including (probably) whatever's left of our Sun.
From eric on the possibility of Martian life: "Mars’ loss of atmosphere, loss of magnetic field, oceans, change to temperature, etc…. occurred billions of years ago. IMO if some microorganism survived those disasters I think evolution over the intervening billions of years would’ve allowed adaptation to the conditions above, on, or near the surface, and we’d have easily found evidence of it ."
Some argue that we have found evidence for it. The hematite spheres, the methane vents in the surface, the 1-out-of-3 successful tests of the Viking lander... and some point to the microscopic (very Earth-life-like) fossils in meteorites that originated from Mars but have made their way to Earth. There is evidence that one can construe as pointing to either current or past life on Mars.
But it's not overwhelming evidence, and it hasn't convinced everybody. In fact, many would argue that this is consistent with no life at any point on Mars, with the only so-called evidence coming from Earth-based contamination. In principle, we could find out immediately by sending a crewed mission to Mars with state-of-the-art equipment. But finding the evidence requires building the instruments that would perform the critical experiments, and there has been an active push against doing exactly that for a variety of (IMO) misguided reasons. I want to know, and I bet most of you do, too, no matter what the answer actually is.
From Denier on betting on failure: "I would caution against counting on the ExoMars Rover. So far no one but NASA has landed a functional piece of hardware on Mars. Others have tried and all have failed. Specifically the ESA has tried and never been successful. The most recent was the Schiaparelli lander late last year which was supposed to test the landing method to be used for the ExoMars Rover. It didn’t work."
The success rate for landing on Mars in general is still appallingly low, and the ESA, JAXA and the RSA have never yet had one successful lander between them all. But sometimes, you put a new system in place and it works brilliantly, as it did with Curiosity. I would accept your recommendation to not "count on" the ExoMars Rover, just as I'd remind everyone that space exploration of all types will always carry a risk of failure. I think this applies to crewed spaceflight as well, and while it might be a good soundbite to say, "failure is not an option," it's always a risk. One well worth taking, IMO.
When spiral galaxies merge, their black holes form a pair inside the nuclear disk at the center of the greater galaxy that results. Simulation from University of Washington.
Also from Denier on star creation (related to supernovae): "In the somewhat related topic of firing up some star creation here in the Milky War, one of the simulations Nvidia did when testing their new processor was to model the collision of Andromeda with the Milky Way."
What's important to remember about these simulations is that they are all limited in what they can track. You want gas, dust, heating, star formation and feedback? You have to model it as an average. You want the individual stars? You can only model about 0.1% of them as particles. You want the dark matter halo? You have to model those as well. And so if you're asking, "will the Sun be flung out of the galaxy," you can only give a statistical probability based on those simulations. (The answer, by the way, is not usually.) While it's fun and pretty, the conclusion they arrive at is not very robust; I wouldn't count on the Sun being ejected in any meaningful way.
What I would count on, though, is that this future merger would lead to an incredible burst of star formation, and that the new burst of star formation would lead to a large population of high mass stars, which would significantly increase the rate of Type II supernovae in our galaxy. And that is a sight to behold! Even if we do get ejected, we'll have a tremendous vantage point of our home galaxy, and that will be a fireworks show worth watching.
The Orion capsule would be one of many components on a proposed space station that orbited the Moon, but the scientific and technological payoff would be extraordinarily low. Image credit: NASA / flickr.
From Andrew Dodds on one possible lunar strategy: "The aim should surely be for a functioning moon colony that has some productive capacity – namely the ability to make iron, aluminium, and titanium from lunar basalt, and hopefully extract volatiles from shadowed craters to make fuel."
I got a lot of pushback after my article here. Many claimed that a lunar orbiting station wouldn't be a mere "middleman" as I asserted, but would be an important base where lunar colonies could stop off before returning to Earth, making the whole operation more efficient, just as a port or rail hub is essential for resource transport on Earth. And while there's some validity to that, it only makes sense if you have at least one established colony first! If you want to do something that takes humanity forward, you need a tangible result, not checking off an item on an imaginary wish list. Show me the value -- and putting humans into harms way from radiation isn't it -- of a crewed lunar orbiter and then we'll talk.
An artistic impression of humans on the surface of Mars. Image was taken on Earth, obviously. Image credit: OeWF (Katja Zanella-Kux).
From eric and denier on some interesting and differing viewpoints on colonization: "I think human space exploration is currently better performed under public/government systems (whose purpose is to spend money for long term, large-scale benefit) rather than private/corporate systems (whole purpose is to pay off investors on a quarterly basis). We are not yet to the point (or barely at the point) where corporate interests can realize a profit for anything beyond LEO satellite delivery." -eric
"What is the point of human exploration in space? At this point what can humans learn that robotic probes can’t also learn and for far less money? Humans might be better than probes at learning how to get cancer outside the Van Allen belt, but that is about it." -denier
There are a number of problem-solving tasks that a human brain is far better suited for than a robot. Humans can assess situations and make inferences from a limited set of information in a way that no computer program can. Robotic missions are very, very good at performing a small set of tasks that they were designed to perform, and very bad at performing pretty much anything else. Not everyone places value on human exploration, and I think there are different and valid perspectives behind both sides of the argument. But many do value it, and if it has the potential to improve the quality of what it means to be a human for a great many individuals -- and it does -- perhaps that's worth exploring, too. (Again, not everyone will agree, but many will.) There's more at stake than just profit and/or individual points of knowledge.
From Sean T on human colonization: "...should we not find out everything it is possible to find out about a potential colonization destination before sending the humans out to make a colony? Space travel is risky, and likely always will be. Should we not reduce that risk as much as possible?"
No. No we shouldn't. I think we should decide on what type of risk-tolerance we're willing to take and once we get past a certain point, we should be honest about the risks as we've assessed them and then we should try and go anyway. "Failure is not an option" is hugely hampering to the enterprise of human exploration of anything, and has never been a part of our exploration efforts before. Most of Columbus' crew died. Magellan died. The first explorers to reach the south pole died. Many mountain climbers die. And yes, sadly, some astronauts have died, too. More will die, but that's not nearly as sad as those who live because they never got the chance to make the attempt. We can reduce the risk as much as is reasonable, and we will likely have arguments about that threshold, but not as much as possible. That's too much. If we value exploration, we need to go. We need to try.
Colour flux tubes produced by a configuration of four static quark-and-antiquark charges, representing calculations done in lattice QCD. Tetraquarks were predicted long before they were ever first observed. Image credit: Pedro.bicudo of Wikimedia Commons.
And finally, from Elle H.C. on what other particle states are out there: "“… there’s almost definitely an entire spectrum of these new sets of particles in existence: tetraquarks, pentaquarks, and possibly more!”
Uh, there can be only one more, a hexaquark, when you’re colliding 2 protons with each 3 quarks."
So there can be many more, and what you're calling a "hexaquark" is not a hexaquark at all, but is rather known as a dibaryon. Six quarks (or six antiquarks) is what you have when you get, say, a deuteron. Instead, think about the fact that you only need a colorless combination. Here's how you can do it with a total number of quark/antiquark particles that equals:
- 4: two quarks and two antiquarks. (Tetraquark.)
- 5: four quarks and one antiquark, or four antiquarks and one quark. (Pentaquark.)
- 6: six quarks or six antiquarks (dibaryon), or three quarks and three antiquarks. (The latter is a hexaquark.)
- 7: five quarks and two antiquarks, or five antiquarks and two quarks. (Heptaquark.)
- 8: four quarks and four antiquarks, or seven quarks and one antiquark/seven antiquarks and one quark. (Both are octoquarks.)
And so on. As long as you have a colorless combination, you can have these states. Whether they live long enough to be detected is still up-for-grabs, but theoretically, they are possible. As we have high-energy, exotic-quark producing factories with large numbers of quarks available, we may yet detect these states and/or infer them from their decay products. Nuclear physics isn't a "done" science just yet.
Thanks for a great week, and looking forward to making the next one spectacular too!
Ah well, is virtual quark still a quark?!
On mars life:
with the only so-called evidence coming from Earth-based contamination...
Oh, just to be clear I don't think the things we're discovering are purely or even mainly a result of contamination. I think they're martian. I just don't see them as evidence of organisms being alive right now as opposed to billions of years ago. I think if some organism had survived Mars' catastrophic changes in climate billions of years ago, then today you'd be able to find living microorganisms pretty easily. No inorganic structures they left - they themselves.
But I will walk my claim back a bit. AFAIK, the deepest we've dug on Mars is just a few inches; less than a foot. Given the radiation and temperature, it seems reasonable to think that surviving martian life would exist deeper than that and not on the surface. So I guess it's entirely possible that the subsurface of Mars has microorganisms and we just haven't seen it because we haven't dug 3 feet down. We will have to see!
BTW according to theory there are three valence quarks and an infinite of sea quarks in a proton.
"All Quantum Field theory interactions involve us having to account for an infinite number of created an annihilated pairs that could occur in our calculations. These don't technically exist in the way valence particles do but they effect the interaction all the same and give us infinities in our calculations, but theoretical methods like Lattice QCD offer us ways around this. The sea quarks may spontaneously pop in and out of existence in finite numbers, but these have no measurable effects outside the hadron."
So what is the LHC going to discover? This seems like we are done.
> six quarks or six antiquarks (dibaryon), or three quarks and three antiquarks. (The latter is a hexaquark.)
three quarks and three antiquarks call baryonium.
haxaquark vs baryon-baryon molecule denote to a wave function of a 6q object: If all 6q form one white object - it is a hexaquark. If we have two white 3q clusters it is a molecule.
If 6q form one white object and none of the 3q configurations within this object are white, we have "hidden color hexaquark".
"what you’re calling a “hexaquark” is not a hexaquark at all, but is rather known as a dibaryon."
This is how it is described at Wikipedia:
"for example a hexaquark might contain either six quarks, resembling two baryons bound together (a dibaryon), or three quarks and three antiquarks. - https://en.wikipedia.org/wiki/Hexaquark
Fair point, but perhaps I wrote my comment poorly. I probably left out the word "reasonably" from it. Should we not do as much as REASONABLY possible to find out what we can about a potential colonization destination before sending humans out?
Like I mentioned in another comment on the topic, Columbus, Magellan, etc. lacked the technological capability to send out unmanned expeditions. Don't you think it would have been reasonable for them to send such expeditions out were it possible? After all, Columbus got lucky; he seriously underestimated the size of the earth and only survived because he ran into a previously unknown landmass. Had he been able to send out an unmanned expedition, it would either have found that landmass or confirmed that Columbus' estimate of the distance to the Orient was seriously flawed. In either case, this would have been valuable information for him, don't you think?
I'm not talking about reducing the risk to zero; of course that isn't possible. I am talking about not flying completely blind. If it will take a few more years or even a few more decades, but will significantly increase the probability of success, is there not at least a reasonable argument for delaying the final human mission?
You might think that computing power would continue to increase ad infinitum, but there’s going to be a limit once we start encoding 0s and 1s with single particles
It is true that Moore’s law has to come to an end. In fact current manufacturing is already behind the curve so Moore’s law in purest terms of transistors per square inch is already dead. That said, it hasn’t stopped the increase of computing power. Right this moment you could purchase off the shelf components to assemble a single 4U server for under $30k (Supermicro 4028GR-TRT2 + 10x Nvidia GTX1080 Ti + 3TB RAM + SSD storage array) that is an order of magnitude more powerful than the IBM Watson room-sized cluster than won Jeopardy!
If you give it about a year the same amount (or a little more) will be able to purchase a new server that is an order of magnitude more powerful than the one you built today. The end of Moore’s Law hasn’t stopped anything. They’re just making the silicon die bigger. In the case of the Nvidia Volta processors which exist today but won’t be remotely affordable for a year, the die size is 815 square millimeters. By comparison the ‘large for their time’ POWER7 chips used in the IBM Watson Jeopardy! Cluster were 567 square millimeters.
I think it’s highly unlikely we’ll ever let an AI decide what is “good” for us, but we’ll almost certainly ask it how to best achieve what we think is “good”
I think you vastly underestimate both the drive for profit and the laziness of society at large. There are already examples where we are letting AI decide what is “good” for us. Cars are being sold today are equipped with sensors and computers loaded with AI trained algorithms that will override the human driver to apply the vehicle brakes for our own “good”. It is foolishness to believe this automotive AI intrusion into our decision making is the pinnacle of what people will allow.
If an AI takes a 'what's good for us' decision than it was trained by (other) humans to do so, humans that set the rules of what is good and what is bad.
AI works often with giving a reward to the program when it makes a good choice or a bad one. Without these points the AI wouldn't know what is good and what is bad. You could even program the AI model in such a way that killing everyone by accident is good.
What Sabine is saying is that we won't let AI make the choice if killing us or saving us is good or bad. Of course good or bad can differ from one person or regime to an other.
On the flip side an AI might run into a situation where some bug influences its decision making, mutates, and goes rogue.
There are already examples where we are letting AI decide what is “good” for us. Cars are being sold today are equipped with sensors and computers loaded with AI trained algorithms that will override the human driver to apply the vehicle brakes for our own “good”.
Speaking as someone who was an AI grad student back in the days before the term itself became passé, by this definition, one might as well claim that a ground-fault interrupter is "AI."
@10: or "the clapper" of cheesy commercial fame. It also makes decisions for us!
There are good ethical discussions to be had over systems such as computer-driven cars. Should they preference their passenger's safety over other vehicle passengers? If so, how much? How do we want "trolly car" scenarios solved by such computers, because the chances are reasonably good that they will, at some point, have the data and milliseconds needed to make decisions about whether to steer left into someone or right into someone or go straight and kill their own passenger when faced with an impending head-on collision.
Second-order ethical questions also arise. Who decides such programming decisions? The auto manufacturing company? The state? The owner of the vehicle? Is it important that all such vehicles have the same rules programmed into them, or are there libertarian reasons to allow variation and choice of vehicle type? Do we want to make such computer calculi deterministic or stochastic (i.e. calculate comparative values down to significant digits, or 'roll a die' as long as values match to x significant digits?)?
But those are ultimately decisions about what we, the humans, decide to program these systems to do and the rules we make the programmers follow. They are not decisions about AI. At least, not at our current level of computer system sophistication.
Second-order ethical questions also arise. Who decides such programming decisions? The auto manufacturing company? The state? The owner of the vehicle?
I strongly suspect that this subject has come up in the context of "fly by wire" aircraft. Then again, in a purely hydraulic scenario, how do you know that the pilot's not an "AI"?
I strongly suspect that this subject has come up in the context of “fly by wire” aircraft.
Aside from there being no "AI" used in the development of any commercially released aircraft controller, the entity who gets to make the decisions is the manufacturer. The completed product is then certified by the FAA as compliant with the Federal Aviation Regulations.
Product Development Decision: Manufacturer
Compliance Certification Decision: Government
Standard Requirement Decision: Government
the chances are reasonably good that they will, at some point, have the data and milliseconds needed to make decisions about whether to steer left into someone or right into someone or go straight and kill their own passenger when faced with an impending head-on collision.
Easy one. If an accident cannot be avoided then take the path that does the least damage to the car.