Science & Technology Q&A for Kids (and others) [Part 47]

okay hi there welcome to another episode of science and technology q a for kids and others and i see we have all kinds of questions here uh let's see where should we start well there's a question from uh travel here why does silver farms consume so much power and what advances are possible with parallel computing to reduce the power consumption or get more computations with the same power okay so why does computation take energy it's a good question and actually it doesn't need to there are so what happens in a in a computation you might be saying there are uh there's a uh there there's a piece of my program which says if this is true if if this thing is a one and this thing is a one then the result is a one otherwise it's a zero that would be an and gate that's a that's a particular uh thing that exists on on a microprocessor that would be a particular operation that could be performed by a microprocessor and the question is why does doing that actually take energy and in the present day the the reason well that is i'm realizing it's a slightly complicated question to answer the um there is a in order to kind of push the electrons through the so these ones and zeros and so on they're represented by clumps of electrons in order to kind of push the electrons to the place you want to be able to do these computations you have to put energy into pushing these electrons to go to the right place now there's a question of why can't you set it up like you would in like a mechanical system if you had sort of perfect billiard balls the billiard balls bounce off each other and there's they don't they the um the energy that you put in is kind of uh regained when the billiard balls go to a different place and so it's um so it is possible to have sort of processes that happen that don't sort of use up energy um as as they run and so it was it was for a long time it was thought that when you do computation you necessarily have to use energy it became clear by about the 1970s that this wasn't the case that it was possible to do reversible computation see see what happens when when you kind of push those electrons through to to get results from the computation you're ending up turning some of the energy that you put into pushing those electrons through into heat heat is sort of random motion of atoms and once energy has been turned into heat it's hard to get it back again into something more systematic that you can use so so the traditional way of doing computation involved pushing electrons around and some of the energy that was used to do that pushing would end up as heat actually part of the reason for that is is the following so imagine that you have a um imagine you're rolling a ball you have a ball that's that's going into some bowl or something you have a marble that's going into a bowl the marble is is going to the marble goes in the marble runs down one side and the marble might just run up the other side the reason it doesn't run so far up the other side is because of friction because it's losing energy if it was the energy that it has the potential energy it has from being a certain height on one side of the bowl it will use up that potential energy turning into kinetic energy of motion when it gets to the bottom then it will be able to turn that kinetic energy of motion back into potential energy and climb up the other side of the bowl so it will be just if there was no friction the marble would just be oscillating back and forth back and forth across the different different uh sides of the bowl but because of friction the thing will eventually lose energy and come to rest at the bottom of the bowl what is friction friction in the end is the conversion of of mechanical processes of motion into just heat it's it's something where you're taking sort of the the motion of the marble and you're you're just making things heating things up and you're you're taking that systematic motion of atoms and it's being turned into kind of random motion of atoms that corresponds to heat okay so people thought that if you want to get a definite answer from a computer when it works out that and computation for example they thought you needed something like that idea of friction to get it to come to a definite answer they thought that if there wasn't something like friction it would be like oh i think the answer is this i think it's that i think it's this i think it's that it would never come to a definite answer that to get it to come to a definite answer would need to have something like friction which would dissipate heat which means that you put energy in and you don't get anything out from it you just get heat out from it that was what people thought now it turns out that isn't true turns out that it is possible to make a computer that is completely reversible and you can do it conceptually by just imagining that your computer is made from a bunch of perfect billiard balls that are perfectly elastic that just bounce off each other and with the exact right pattern of billiard balls you can do whatever computation you want it turns out well so this idea of reversible computing is theoretically possible um it's never actually been done in practice practical computers right now still use the fact that kind of you have to you have to have some sort of friction you have to you have to dissipate heat to get the thing to come out with a definite answer and the question is how fast do you want the thing to come out with a definite answer well the the more you can have the the faster you can push the electrons effectively and the more you can just allow that pushing to turn into heat the faster you can get an answer now in practice and the design of microprocessors and so on that turns into questions about capacitance and and uh uses of of uh voltages and and uh sort of details about about the way that you're sort of charging up capacitors and letting them uncharge and so on let's not get into that doesn't so it's it's more details of the electrical engineering um but uh the the main point is that the the the more energy you put in more or less the faster you can get things to run given that there is a certain uh degree of sort of physical character to these electrical signals and all this kind of thing so the reason that servers tend to use lots of power is your you know the kind of the harder you push them the um uh the the faster they'll go now there are details so for example in the detailed design of microprocessors there are trade-offs about how exactly you'll set up the wires whether you'll have these these sort of talk between wires when that happens uh whether you'll have um uh certain features of of the way that the characteristics of the individual components work and so on and by tweaking the engineering you can end up with using less power um because things sort of just go more smoothly in terms of the way that the the underlying physics of of the of the devices is set up and sometimes by by letting the thing run slower you can use less power but so typical server farms have this trade-off between how much how fast they go how much power they use exactly what design does the thing have and sometimes when people have really redesigned microprocessors from the ground up it's been possible to design them to work with lower power so for example in the in the whole sort of story between intel and arm two different major microprocessor design design companies um there's sort of been been issues for example there was uh for a while the issue that sort of arm had in simpler designs not as sophisticated as intel's designs but those designs were sort of created from the ground up and it was possible to make lower power versions of them which is why arm processors one of the reasons arm processors became popular in cell phones because that was a place where power was at a premium whereas in servers it's like the thing is connected to the electrical grid and you can just put power through it and get computation done in terms of of whether things like parallel processing will help with the sort of the power that's used in service probably not so much because once when you're computing you're computing you're doing it in parallel you don't sequentially doesn't really make much difference you know i know for my computers they have many cores in them and you start to when if you start running a program that is making use of many parallel cores so normally when you run a program in the simplest way of setting up a program the program has a series of instructions in it and those instructions get executed in series and so on your computer there's just one the the the computer is just saying i'm at this point in the program okay i'm going to move the program counter to another point in the program then another point it's sequential but in modern computers there are often multiple cores multiple different uh pieces of of microprocessor that can separately run programs and so it's it's pretty common to want different programs to run in different cores and for example your operating system and your computer will be running the different programs you're running in lots of different cores on your computer but sometimes you have a single program where you want to divide up the work into multiple cores let's say for example you're going to process a whole collection of photographs and you're going to i don't know you're going to turn every uh um every cat in every photograph green or something you want to go through and you want to use some machine learning algorithm to turn every cat and every photograph green and you've got a million photographs how are you going to do it well you can have a sequential program that just opens each photograph does its thing and you know decide turns the cat green saves the file again goes on to the next one goes on to the next one but let's say your computer has i don't know uh 20 cores or eight cores or four cores that it makes more sense to have those separate cores each separately go out and deal with their group of photographs and do all those things in parallel but when you do that and when each core is doing some sort of computation intensive thing you'll often hear your computer start to make more of a noise what's actually happening what's happening is the cpu is all those different pieces of the cpu all those different cores are starting to be used and the computer is using more power and it has a fan typically that is or many computers have fans that are kind of blowing air across the top of the cpu chips to keep them cool um because they're using more power and so they're going to heat up and what happens is usually there's a certain operating temperature range of temperatures for for microprocessors it's determined by the fact that if you make the electrons in the microprocessor too hot if you make the microprocessor if you take make the semiconductor that the microprocessor is made out of too hot the electrons will start kind of moving around in the material even in ways that you sort of didn't tell them to move around by putting voltages on and so on so there's usually a a maximum temperature at which a semiconductor device can operate and so you have to if you're starting to put more power into this device you have to be pulling power out by having by blowing air across it and having sort of heat exchange happen and so on and so that's why when when you're making your computer do more work the fan will start up as it's trying to blow air across that microprocessor to keep it cool so that it doesn't go to a temperature that's too high that um uh uh that that where it switches off you know i i discovered once i i remember when was i thinking when i first had an ipad a very very long time ago when ipads first came out i i left the ipad on a um on a dash in a car it was a sunny day and this ipad because it had effectively a black screen when it was switched off was just absorbing sunlight like crazy and it got really hot and i was kind of impressed that when i tried to switch the thing on it came up with this warning thing saying this ipad is too hot it will damage itself if it switches itself on you know let it cool down before it can before it can be operated and that was that was a microprocessor that had gotten outside of its normal operating temperature so that's why why these things have to be cooled in in elaborate ways and so if you go into a you know in a typical if um if you go into a server room uh server farm for example they're usually called they're often called server farms um i'm not sure that's a good comment on the way that animals are kept in some kinds of farms but because they're just rows of servers um that uh are just doing computations and so on and a typical server room we set up in so-called hot aisles and cold aisles so you have all these rows of servers and they're they're all stacked up they're usually the the pizza box form factor is a pretty common one where the the actual computer is the shape of a pizza box basically and uh uh it's it's fairly thin sometimes they call blades um the uh where the the circuit boards for the computer are just lie horizontally and um uh you you're stacking them all up to make this big big um uh big server uh big big configuration of server room and usually what will happen is there's a side of it that's cold where where it's being cooled by the by the uh air conditioning and the server room and then there's a back of it that's hot uh where it'll just be blowing out waste heat from the operation of the server and that's sort of a typical setup and it's often the case that the amount of computation that can be done the number of servers you can put in a server room is limited by the by the rate at which you can get heat out of the server room um and uh that's and so sometimes people do crazy things like they they have uh uh they try and have several rooms sort of embedded in glaciers or there was a there was a trend at one point of having you know underwater servers um that were uh kind of dissipating heat uh to an ocean there's a there's a more amusing version of that which is if you have um for example high altitude drones that um for example are providing things like cell phone service you can do computation in a high altitude drone where it's getting power from sunlight and so on and because it's really cold at 60 000 feet or wherever the drone is operating you don't have to have the kind of cooling that you need to have in a server room the drone can just as it's flying through the air and it has a server on it the server will just be cooled by the by the low temperature of the air at the altitude that the thing is flying at and so you can you can kind of optimize things in terms of cooling but in terms of what will the future be in terms of uh power consumption of computers my impression is that the engineering progressively gets better and it's progressively possible to sort of remove all those little uh stray things that that that use more power than they should it will take some new technologies i suspect to dramatically reduce the amount of power that's consumed and certainly some of the technologies that people use for what they imagine to be quantum computers even if they're not really quantum computers operating really really according to to the laws of quantum mechanics some of those technologies particularly ones with optical with light and so on instead of electrons instead of electricity some of those will likely lead to things that are computations that take a lot less power um so that's a a comment um uh about that um let's see the questions here boy um [Music] okay there's a question from aaron here have you ever had ever had a situation where uuids collided and caused some sort of issue well that makes me talk about what uuids are um i've sometimes said when the uuids collide the universe will end um i say that somewhat uh somewhat facetiously but let me explain what uuids are and uh and and and how this works so uuid stands for universally unique identifier so let's say that you are going to you have a whole network of computers and uh somebody you're going to have a uh a file that's created on one of those computers and you want to make sure that that file is named in a completely unique way so even though you didn't check with any of the other computers oh do you have a file named this do you have a file named that you want to make sure that the name you come up with for that file is going to be universally unique that no other computer will come up with that same name you might also do this if you were for example if you were generating addresses for little devices and you you're going to have a trillion devices in the world and you want each of them to have a unique address one way you can guarantee that they all have unique addresses is to have a registry where you're listing all those addresses and you check for every new device you add is that address i'm trying to add in the registry or not if it's in the registry i can't use it if it isn't in the registry i can use it but the point of a uuid is to avoid doing that to use the fact that if you have a long enough number and you pick it randomly the chances are that that will never be picked at any other time so in other words if you were to say i'm only going to i've got a thousand possible ids that i can use well if you by the time you've got maybe about 500 different um uh different things where you're picking the ids the chances are you'll have a collision with chances are two things will both randomly get picked as id number 205 or something and then whatever system you had that relied on the fact that no two different things ever have the same id will just get get uh will fail so the idea of the uuid is um that you have something where it's a long enough randomly enough picked number that the chances are it will never be picked the same two name number will never be picked ever again in the sort of history of the world and typical uids are these days what is it 256 bits i guess um so they are they're usually written out in hexadecimal um so hexadecimal is base 16 we usually use base 10 and so in hexadecimal they're numbers 0 through 9 and then there's a number there's an a for 10 b for 11 etc up to f so those are the digits of hexadecimal and usually uuids are numbers that are written out in hexadecimal that's the most common way they're presented but they're fundamentally just numbers but they're numbers between uh 0 and 2 to the power 256 minus 1 and those numbers that's the typical choice of uuids that that number is big enough that it's for example larger than the number of atoms in the universe so the if you were to number uh if you were to use the uuid to label every atom in the universe there's a good chance that if you picked that label at random that every atom would get a different label so that sort of seems enough that the uuids aren't going to collide um and the question is to make sure that uuids don't collide you have to pick them at random if you say every tuesday i'm going to pick uuids of this form well that's bad news because then the chances are you'll end up picking two identical uids it's the same kind of problem as if you're trying to pick a password a key for a cryptography system you want to pick it sufficiently randomly that that that same key won't ever be picked again or that people won't be able to guess what you picked um so there's sort of a challenge of how do you make how do you pick the uuid randomly enough that it won't that the same thing won't get picked again it's not that difficult in practice to do that so i've never seen a case where uuids collided so many systems in the world today depend on uuids not colliding that i have no idea what would happen it's a good kind of doomsday scenario to figure out what happens if the uuids collide i don't know i think a lot of things which where the identity of a thing is determined by uuids would get very confused and people where they think they're two different users of a computer system it would be like oops the uids have collided they're actually the same user and that would get very confusing and things where only one user is supposed to have permission to see something well actually another user with the colliding uuid will get permission to see it and so on a lot of things would go crazy in the world so we have to sort of hope that uh that you ids won't collide and we can know that so long as they're picked randomly enough they won't collide i might tell a piece of history that some of you may know some of you may not know about kind of things people assume about computer systems and what goes wrong with them this is about the y2k problem and this is supposed to be a q a for for kids and others so the kids may not know about the y2k problem the others may know about the y2k problem what happened is uh heading for the year 2000 there was a big sort of panic about the way that computers stored dates because computers are typically stored year month day and the way they'd stored as a date and the month they store as you know zero eight or something for august the day they store as you know one five for the fifteenth of the month and it's like uh programmers had thought well let's just store the year as two digits it's all good that was the decision that was made in the 1960s and it's like let's store the year 1998 as 98 let's store et cetera et cetera et cetera use just two digits well the problem is that a lot of programs said is this date after that date and so they would say uh we compare the dates so first we compare the years is this year after this year okay well you know the year 1998 stored as nine eight is correctly identified as being after seven five which means the year 1975. but if you're storing things only with two digits by the time you get to the year 2000 it's that doesn't work anymore the year 2000 is after the year 1999 but zero zero the number zero zero is before the number nine nine so the sorting of dates we got wrong so people really started to panic a few years before the yi2k year 2000 uh moment of oh my gosh all the computer systems in the world are going to fail and so a bunch of companies that sold sort of systems that deal with particular transactions and databases and things they were like this is the most fantastic opportunity we've ever had we can finally get people to upgrade these systems they've had the same system for 20 years they've never been upgrading it you know it's going to fail you know let's not patch the thing let's just have the thing fail and say the only way you get to have a system that works when y2k happens is to have a new system so it's a huge amount of activity in sort of patching up computer systems so that they didn't fail at the moment where where y2k happened people really thought that there might be a possibility you know the the air traffic control system would fail the nuclear power stations would fail the electrical grid would fail that that at that moment where the uh where y2k happened the world would end it's kind of like the uh the mayan calendar which has this feature that it eventually wraps around and people sort of wonder if the world is going to end where the calendar wraps around people were really wondering what was going to happen at that moment when the the the beginning of the year 2000 uh when two digit dates uh would have failed in that way and i remember we had done a lot of stuff at our company to try and be ready for y2k some of it was probably overdone actually because there was such a sort of generated panic about y2k and i remember at the at the uh at the moment where y2k happened i was like uh you know let me send a piece of mail to all people the company you know happy new millennium type thing and uh let's let's get that right at the moment of y2k and let's see if anything actually fails well it didn't um actually interestingly enough there's another y2k like problem in the year 2038 that's coming because uh in unix common operating system common sort of foundational operating system for many computers there is a uh it's often the way that unix stores times is our seconds after the year 19 after january first 1970 and it turns out that the number of seconds after january 1st 1970 by the year 2038 will be more than the number of bits that you can store in a 32-bit word um and so that if for systems that have 32-bit uh that use 32-bit numbers there'll be a sort of y2k like problem in the year 2038. i don't think it's going to be very serious because everybody will move to 64-bit systems and uh i think um uh the sun will no longer be shining when um when 64-bit systems ran out of seconds so it's uh so we don't have to be too concerned about that um all right it's a question here about um uh i think coming back to servers and things like that um from reem why does bitcoin use so much power uh let's see well all right basic point what is bitcoin bitcoin is a cryptocurrency that is based on the idea of blockchain what is blockchain blockchain is a way of recording transactions it's a way of having this chain of recordings of transactions where it's set up so nobody can fake it it's set up so that you you will have you store this person paid this person such and such a number of bitcoin this person paid this person and if somebody says i've got this bitcoin i can pay you somebody can go back the system can go back and automatically check is it really true that you still have that bitcoin or did you already spend it and so there's this ledger there's that goes all the way back to when when bitcoin was first uh deployed in what 1998 am i getting my years wrong around that time is that right um that uh has it been that long no no no and maybe i'm i'm off by 10 years maybe it's 2008. um in any case when um yeah i think it's i'm off by 10 years there um so when bitcoin was first deployed the first transactions were recorded on the bitcoin blockchain and transactions have been being recorded on that blockchain ever since and the way it works is there are maybe 10 000 maybe 20 000 copies of the bitcoin blockchain and different on different people's computers all over the world and all those copies are consistent and so when somebody says do you really have the right to give me this bitcoin you can go check any of those copies of the blockchain and you can make sure that yes that bitcoin hasn't been spent it's good you can go ahead and give it to me type thing okay so the question then is how does though how do you decide when a transaction happens how do you decide that that was a valid transaction that really happened and it should add it to the ledger of transactions to this blockchain of possible transactions and there are a variety of different schemes but the one that was picked for bitcoin was a thing called proof of work and sort of the idea is people can oh let's see how deep do i want to go into this people the the notion is you have to decide who is going to add the next block to the blockchain for bitcoin a block is being added about every 10 minutes these days and what's happening is within each block there'll be this whole big list of maybe i don't know a thousand transactions that happened that are being recorded on the blockchain for all time and the question is in order to put that block onto the blockchain you have to make certain cryptographic things connect with the previous parts of the blockchain there has to be sort of a valid way that you add that block to the blockchain but the question is who gets to add that block and you want to be the person adding that block because every transaction that's in that block being added to the blockchain comes with a certain commission and if you get to add that block you get to take all those commissions you get to be paid in bitcoin all of those commissions and so so you want to be the person who adds the next block to the bitcoin to the blockchain question is how is it decided between these 10 20 000 uh different uh places where the blockchain is stored these different who's going to add the next block okay so in order to add a block the rules of bitcoin are such that you have to solve a particular essentially computational puzzle and the puzzle is basically there's all of the data that corresponds to a sort of a hash code hash codes are you take some collection of data and you kind of mash it up and you get a number out of it and so you might have a huge amount of data and you mash up that data you just get this number that's a 100 digit number let's say and uh hash codes are useful because the chances are two different kinds of data that you mashed up will end up with two with different 100 digit numbers so if you want to find out whether these two things are the same you just compare their hash codes and that will tell you well sort of an abuse of the idea of hash codes is what happens in in bitcoin because what's happening there is one's asking can you find a way to take all the data that exists on the blockchain plus a little extra piece it's called the nonce and to arrange the nonce so that when you take all the all the data on the blockchain plus the nonce that the hash of all of that is a number of exactly a certain form that uh starts with a bunch of zeros okay so the the problem of blockchain is that the way you decide who gets to add the next block is who's the first person who picks the nonce to get the hash that has the right number of zeros at the beginning i forget how many it is now maybe 14 zeros i think it's got to now um and and the difficulty of finding that hash is being made more difficult every few years as you're required to find a hash which has even more zeros at the beginning so so what's being done bitcoin mining is the process of trying to find those nonces effectively that will generate a hash that has the right number of zeros now so and and to do that the only way people know to do that is you just try a lot of hashes you just try and try a lot of monsters and you see what the hashes are and suddenly if you try enough of them chances are you'll say ah i'm the winner i've got the one that starts with a bunch of zeros um and so that process of bitcoin mining that's what it involves and the the sort of concept is that these are so-called cryptographic hashes and they have the property there's really no way you can find out whether the hash will have the form you want except by just running the hash and seeing what happens this is very related to this idea of computational irreducibility that i developed back in the 1980s and i i think it's very likely that the original nature of bitcoin knew about that idea and that was sort of uh the origin of this notion that you can have to do irreducible computational work to achieve something but so that the idea is that the so what's happening is uh people have made all sorts of special purpose computers and things and they count the number of giga hashes the number of billions of hashes per second that those computers can do to try to find the find the nonce to find the you know to add the the um uh to add the next block and so people um uh that that's some um uh that that's so so what's happening the the energy that's being used in in mining bitcoin is just running all these hashes to try to see whether you can get the one that is going to allow you to validly add a block to to the blockchain so this is an incredible waste of electricity it's a you know when i invented computational reducibility i never imagined i could never never imagine anything like this as a sort of application of computational irreducibility why does it make any sense it makes sense because you want something that is sort of hard to make you don't want something that doesn't take any commitment to make you don't want something where everybody just say i'm i got the block i got the block i got the block i have to have some way of deciding it's kind of like you imagine people digging up gold somebody is the person who gets to find that gold deposit and go to all the effort of mining it and that's the person who gets that piece of gold and it's the same kind of idea here in in in bitcoin and you know i think people say that the the total amount of electricity used in bitcoin mining is comparable to the electrical output of denmark and it's it's predicted in a couple of years it's going to be comparable to the electrical output of italy i don't know if this is all correct but lots of uh lots of electricity um is being used in bitcoin mining a question you might ask is could bitcoin mining be actually solving a useful problem instead of doing this kind of uh computational puzzle of solving hashes i have wondered about this i've never figured out how to make it something where it's solving a useful problem because you kind of have to know in a sense you have to know the form of the answer in order to validate the fact you've got a correct proof of work so to speak so it's hard to have something where it's doing something useful where it doesn't know the answer though as i'm thinking about it telling you about this i'm thinking about whether there's a way to get something where there is an answer and it gets to that answer but there is an intermediate step that was necessary to get to that answer and the intermediate step is useful the answer is known that the intermediate step is useful the other issue with computation that are useful is it may be more difficult to know for certain just how difficult the computation is expected to be although the truth is that with cryptographic hashing it's a question of did you guess the right thing by by good luck or not um there are other approaches used in other blockchains the most popular is a is there are many different ways to decide who gets to add the next block sort of another one that's becoming popular is proof of stake where you're basically showing that you own a bunch of that cryptocurrency and you get to sort of vote on what the next block is that's a that's a different way in fact ethereum is just transitioning to something uh close to that maybe it it's just happening right now ethereum 2.0 i think uses that kind of method rather than the just burn computer time to to uh to solve these uh these computational problems let's see there's a question here from slayer very different kind of question what is the correlation between sleep and weight loss boy i don't really know um i think a lot of that information is so deeply anecdotal i mean i i am uh it's always some uh you know we don't really completely know what the function of sleep is people uh you know i've always assumed that sleep is necessary to clear kind of junk out of the brain much as when muscles get tired because they build up lactic acid from uh from sort of the repeated uh um nerve uh um the the the repeating of that um um of the operation of the the the nerve uh on the muscle um that uh you know just as one builds up sort of the waste product of lactic acid in muscles someone must be building up some kind of waste product in the brain as a result of having a sort of waking day and i think people say adenosine is one of the possible things that is produced as a waste product from the uh the sort of electrical activity of neurons in the brain and then sleep is sort of a time to clean that out and to their sleep because of the way biology tends to work once it has a capability it's going to pile a lot of other pieces of functionality on top of that capability so it's like okay given that you've got the animal asleep let's have it uh do this let's have it do that let's have it uh you know have a time when it's digestion is doing this so when it's um something else is doing that and i think um the uh so sleep ends up getting used for many purposes even if it's primary purpose might be sort of cleaning out gunk from the brain it's getting used for lots of different things and some of those things probably have to do with some of the kinds of things that relate to one's metabolism and so on certainly one knows that you know temperature goes down at night heart rate goes down at night um i'm not sure how all of those uh things play together and i think the um a general finding is people need a certain amount of sleep um and if you don't get it bad things happen uh things like your immune system stops working as well you're more likely to you know catch a uh catch a virus because your immune system is not as aggressively uh attacking it uh your uh you know all sorts of different things happen you you don't think as clearly because you know something is kind of um uh gunking up your brain perhaps but so there are different things that happen when you um uh don't have enough sleep and i think it's one of the the uh the things of modern times i don't know what's happened to people's sleep during this pandemic i i kind of think it might have gotten better that uh you know there's people in many periods of life tend to have a sort of chronic lack of sleep where they're like well i've got lots of things to do and then i gotta get up very early in the morning and and certainly uh school systems have had a terrible habit of getting kids to get up super early in the morning even though you know the way that we are all synchronized you know we have this kind of circadian rhythm that tends to synchronize when it is that we think we should sleep and it's so synchronized to the sun um and uh it also tends to be aligned relative to the sun and i think people generally believe that particularly kids are actually quite unsuitable as teenagers and things at least to be getting up super early in the morning and it's kind of a crazy thing that uh that for example schools often start very early in the morning and i think that's been something there's been some pressure against that but it's something which sort of the societal structure of things is such that well it's really good to get the kids off to school because then the parents can go into work or whatever um and so that's how that develops but it's probably not the optimal thing in terms of cognitive performance of the kids and so on but uh so i think um uh so i don't know the answer to um uh the relation to uh uh to weight loss um us aaron is pointing out that the number of atoms in the universe is actually larger than the size of a 256-bit uuid you know as i was saying that i was remembering that i i think i mentioned this in this little book i wrote called elementary introduction the world from language i think i discussed this question of just how big is the typical uuid and i i'm i'm remembering um uh the um um uh this point you're absolutely right that the it wouldn't quite make it to label all the atoms in the universe um the question of how we know how many atoms there are in the universe we know how big the universe is uh we know um they're um uh let's see they're about some uh yeah they're about a hundred billion galaxies in our universe each one has about 100 billion stars in it each star has about 10 to the 57 atoms in it and so we we know from that uh roughly how many atoms there are in the universe um it's about 10 to the 80th atoms in the universe that's a that's um uh 10 times 10 times 10 times 10 80 times so it's uh if we counted out a trillion as 10 to the 12 so if we counted it out um uh let's see can i do the math here uh it's a a trillion trillion trillion trillion trillion trillion trillion trillion uh atoms um in in the universe which is quite a few but um and uh and more as you are correctly pointing out then you fit in a 256 bit uuid so at the point uh which is not going to happen in our universe with the evolution of our universe at the point in the in the very very very distant future of our universe um the uh uh that you know when when the time the universe the 14 billion years of our universe existed so far when that time looks like something infinitesimal when the universe has evolved you know many many many orders of magnitude more than that time then maybe we will be at a point where the uuids that we have today will have sort of more serious problems so to speak um of course it's not clear there are many other weird things that can happen in the universe before you reach that point um gosh okay all sorts of interesting questions here there's one from mahmoud how to convert black and white video to color so this is something people have been doing in recent times you know you've got a black and white movie that was made in the early 1900s and it's like let's colorize this and the traditional way to do that is people with paint brushes effectively you take every single frame and you go and you kind of change the color of the of the black and white areas in every single frame that's a traditional way to do it um and you know by using human judgment you say uh you know this um uh you know this cat was yellow type thing and let's just decide in this cartoon that was made in you know 19 you know 1925 or something let's just decide this coat was red and then we color all those frames in the right way okay the thing that's become possible recently is more automated colorization and the way that works is as follows it's it's using machine learning and using neural networks and essentially here's the idea imagine that you could show a computer examples of zillions and zillions of different pictures maybe you could show 100 million pictures and each of those pictures has things that have certain shapes and certain colors you have a tree it's got green leaves on it the tree is a certain shape and the leaves are green you have the sky it's up there and it has it's sort of plain and it's tends to be blue you have different kinds of things and they tend to be certain colors so essentially what one wants to do is to say given the shape we can we know 100 million examples we can look at the shapes in those examples and we can say what color did a thing of that shape happen to be in that image and so then if we can learn from those examples we can say given this movie frame that has a picture of this this and this in it we can reasonably guess based on the examples from those hundred million images we can reasonably guess the sky up there is blue the grass down there is green et cetera et cetera et cetera and so the kind of the the idea of colorization automated colorization is that idea that you learn from all these examples and then you implement it in the particular case that you're showing now how does that learning work and so on that works using these things called neural nets which were kind of invented as a as a idealization of what's happening in our brains and essentially what what goes on there is that well it's a fairly complicated story altogether but roughly you've got a program that has a lot of different variables in it millions and millions of variables and what's happening is you're saying you want to train this program to follow certain examples so you might show the picture it says that's an elephant you say no it's not it's a cat and then and what it's doing then is it setting those different numbers inside the program to be such that okay it said it was an elephant but actually it should have been a cat so let's tweak those numbers to make it say it's a cat in this particular case show another picture it says uh you know that's a car and you say no it's an elephant you tweak the the things a little bit more to say okay make that be an elephant as well and you keep doing this tweaking and you do it billions and billions of times and after enough time you can have ended up tweaking it to the point where it'll more or less get things right and it'll more or less say yes that really is a cat in that picture and so you can do the same thing with colorization you're progressively training using existing images and you're getting it to the point where you where it correctly says the grass is green the sky is blue etc and that's how this automated colorization mechanism works and and you can find in more from language for example we have some colorization uh neural nets that will do this kind of automated colorization and it works fairly well it's not perfect but it works fairly well um and uh it certainly can bring some black and white pictures somewhat to life there's a question from heather here how much sleep do i get at night you know i'm one of these people who who knows a disgusting amount of information about myself and uh i have all that data the answer is about seven hours usually and um i i usually make sure that um uh i kind of budget for eight and expect to get seven um and uh the um uh and it's i think uh over time i have probably needed a tiny bit less sleep um and uh it's kind of a feature of the of the horribly aging brain but i kind of think that um in the effort to preserve it as long as possible uh getting enough sleep is an important thing and i recommend it to everybody and it's kind of like one of these things where you can say well let me get less sleep you know i think only oh gosh i'm trying to remember how many all-nighters have i pulled in my life i can remember one of them when i was a teenager um the uh um uh and um have i pulled other ones hmm i've certainly uh you know i've had things where changing time zones and flying around the world i've had sort of somewhat strange sleep schedules but i think i think i've been convinced that all-nighters don't make a lot of sense uh partly because what i find is if i'm really tired i don't think very well and i can't do the things i want to do and it's much better to just sleep and then i'm actually alert and i can figure stuff out and things happen fairly quickly and i can be like tired and i'm like um i don't really know and uh you know can't figure anything out so i in in in worst case i'll i'll take a um i i one of the things that's been an advantage of getting older is it's getting i i it gotten progressively easier for me to like take a nap for 15 minutes and have that work i think when i was much younger i used to have much more trouble you know falling asleep and i i made an active effort when i was a teenager actually to you know lie on one's back close one's eyes just fall asleep and i finally got to the point where yes i can really do that it's sort of a reflex to to do that and that has the terrible feature that you know in all sorts of strange situations where uh you know i'm at a dentist's office and i'm you know i have to say i'm sorry you know don't don't mind me if i fall asleep because when i'm in that kind of position i sort of have this reflex of falling asleep but it's been a useful thing to learn and uh something that that helps me um okay there's uh let's take a look here um let's see um question here from aaron ask what kind of car do i drive and what's the difference in technology that fascinates me versus technology that i just see as utilitarian oh interesting question i mean i've never been a big car enthusiast i haven't driven the car very much at all in the past year but um i have what do i have i have some kind of bmw x5 i think is the designation i think that's right the fact that i don't even know exactly what kind of car i have is kind of a sign of a lack of profound interest i i got this car i think i think my wife basically picked this car up for me and um the uh uh i've i've generally not been um uh and it it sort of has the technology is okay um and uh doesn't have anything that sort of horribly in your face not um um uh kind of you know badly designed technology i had a previous car that was nice car to drive but it had just really challenged technology um i have to say when i got this car it was just a personal silly story but i like what was it a couple of two two three years ago i'd realized i had a previous car for 10 years and i still didn't know how to do certain things with that car it had all kinds of controls and it's like i don't really know how you get this to work and that to work and so on and in fact some things just didn't make sense and you know i i tried to figure out how do you get it to you know have its gps work without having the radio on or something and it just like you just can't do that um and uh so this time i got a car i decided i would do a thing that was sort of a um uh a look am i gonna really read the manual am i not gonna read the manual and i was like okay just get the car dealer to spend you know an hour and show me how the car works and i have to say they were a little bit surprised it seemed being asked that and it turned out it wasn't as educational as i hoped but it did allow me to know oh yeah you press that button there's this little door that opens here and it's like i would never have found that i could have driven this car for a decade and i would never have known that so so thanks for telling me that it was it was a useful piece of education but um uh i would say that um uh sort of a big trade-off between what technology do you really sort of uh put effort into understanding what technology you just use for the sake of using it i tend to always try and get the kind of most leading edge gadgets and devices and so on unless they just don't work very well and there's sort of always this trade-off if you do have things that have been well polished but don't don't do as much or do you have things that do all kinds of cool stuff but have all kinds of rough edges and it's always a trade-off for me which of those things i will end up being uh you know uh dealing with and like for example you know that uh you know the thing i wear on my wrist is a you know that measures heart rate and and motion and so on is is not actually the very best and latest uh technology because the best and latest technology doesn't bundle in things like uh getting text messages sent there which and i don't want to be wearing two different things i mean there was a period of time when i was wearing a thing on one wrist a thing on a little wrist and a thing on my ankle um and that was like this is too too crazy i'm not going to do this um so you know that that's a uh that's the trade-off there for me but i think it's it's always um i also tend to be somebody who when i'm using some technological device and i find some kind of bug with it i make a point of actually reporting that bug to the company that makes the device because i know from being on the other side of that story people will say oh i've been using wolfram language for years and oh there's this really annoying thing that happens and you say really you know and maybe it's a bug but maybe they just didn't understand how to do something but even if it's a bug it's like i wish you'd reported that 10 years ago we would have fixed it nine years ago then but nobody ever reported that because nobody ever had you know a game pad together with uh uh you know a zulu keyboard together with this together with a vat and this just never came up before but sorry you know you've been the person who is the one person who has that particular combination and so you see this all right let's see um it's a question here from s leia uh asking can i explain the differences between dc and ac all right that's a that's a uh he says famous last words a fairly straightforward sort of piece of physics and engineering so what's that about it's about when you're going to uh you're going to have when you see this is this is what always happens the things that i think are like super super straightforward to explain there ends up being a little bump um before one gets to them but okay so this is about electrical power and it's about how one delivers electric power to things so for example when you have a power outlet in a typical place in a typical country uh you're plugging a you know you plug that plug into the wall you're getting a certain electric current from the electric grid um that you can use to to power your devices your computer whatever else um the uh also when your computer is connected you know you have this adapter that goes plugged into the wall then you plug the other end into your computer your the wire that's connected to your computer is also there's electric current going through that wire but the electric current that goes to your computer is a little different than electric current that comes out of the uh out of the the wall so to speak and the difference is that what's coming out of the wall usually is ac or alternating current what's going into your computer is usually dc or direct current okay what's what's the difference between these the issue is there is a voltage the voltage is kind of what pushes electrons through wires when you have a voltage it's like there's a high electric potential lower electric potential it's like you're making a ramp and you're forcing and the electrons go down that ramp so when you have when you're you're having a high a plus voltage on one side a minus voltage on the other side that's defining a a potential difference which is kind of a thing that pushes electrons from one side to the other and um the uh because of a vagary of the way things have worked out you'll actually push electrons from the minus side to the plus side because electrons are negatively charged things but but that's sort of a a detail you're you're basically pushing uh you know voltage is pushing electrons through wires so direct current is you're always pushing the same way you just you just push push push the electrons through the wire and batteries for example produce direct current there they produce a voltage it's always pushing in the same direction so another thing is alternating current and what's happening with alternating current is you're pushing the electrons one way for a short time then you reverse and you push the electrons the other way you push them the two different ways and for many purposes it's perfectly sufficient to just have electrons be pushed for example if you're trying to heat something up you just want electrons to get pushed into something the fact that they get pulled out later pushed in pulled out doesn't matter they they still by by this sort of motion of the electrons effectively you're still heating things up doesn't matter that you have alternating current for other purposes it can matter for example if you're running a computer it can matter that you always are pushing the electrons the same way you need direct current so why is alternating current so in in actual electrical supply in the us for example it's at 60 hertz which means 60 times a second you're reversing the direction of the voltage and it's a it's a sine wave so it just wiggles up and down uh 60 times a second in some countries like england is 50 hertz uh some other places it's usually 50 or 60 hertz is the typical frequency of the electrical supply and usually it's pretty precise actually the there's there's a considerable effort is often made to arrange the the uh the frequency to be pretty precise it used to be the case that electric clocks would actually uh use the uh electrical frequency as as the way of of um of deciding uh of measuring time that's no longer done it's it's just oscillation oscillations of quartz crystals or alternatively just data that comes from the internet that sets time but um but back in the day that it used to use the frequency of electrical supply okay why do people use alternating current well the main reason is that when you're going to make uh you're going to send electrical uh electricity over a long distance it is it uh less um uh you you lose less power in heat with alternating current than you do in direct current and so if you're going to have an and actually it's even more than that the higher the voltage the less you dissipate uh heat and so when you're sort of taking electricity long distances when you're for example you have a power station that's a hydroelectric power station that's sitting right there at the position of a waterfall the um uh then you you're generating power at the hydroelectric dam but you want to use power at the city that's 500 miles away so you have to you have to have wires that take the electricity from hydroelectric dam to the city and usually the the main wires are operating at high voltage because that's the way you lose the least power um in transmitting so usually i think um 20 000 volts is common 100 000 volts is also common as the voltage that you're you're sending and that's always alternating current because that loses less power now back in the day there was a a big sort of competition when when electrical devices in the electric grid were first getting developed uh thomas edison and was it george westinghouse that's right uh westinghouse versus edison big kind of competition of um uh who would uh uh who would end up being uh you know which system would end up being the winning system am i am i remembering this history correctly the um and i think edison was not on the right side of i'm trying to remember who was on which side of that story and i remember all kinds of crazy terrible things about about demonstrations of electrocuting uh poor animals with with direct current and things and and showing that alternating current uh didn't work as well but it was sort of a fake experiment and all kinds of skull duggary but but um having to do with with which got picked but in the end uh for the main electrical supply partly for this reason of electric of power dissipation and for other detailed reasons of of which technology got developed first and so on alternating current was the thing that got used for large-scale supply alternating current has to be converted to direct current for things like computers and that's what that little box that is um between the uh uh the electric uh supply and the computer that little box is uh is the thing that does the conversion and i have to tell a little story against myself there that um uh many years ago as a physicist who claimed to have invented some zero flux transformer that would transform alternating current to electric current and somebody i knew was um we really wanted to know is this going to work should i invest in this invention and i'm like i don't really know but you know this physicist is you know is a person who does particle physics i don't know why they would know anything about electric transformers and so on is a very different design from any transformer design that's existed up to this point and uh you know i kind of it's like i don't really no i don't i don't think it's going to work and you know that would be my my assumption well many years go by and then i realized that actually that physicist to their great credit pushed this idea forward and it became the standard idea that's used in essentially all um uh has been used until recently there's one new idea sort of come on the horizon that's been used in sort of essentially all adapters that go from um uh for from maine's ac mains supply to the dc power that's used in computers and one of the things that's actually clever about this particular form of adapter is it used to be the case that if you wanted to go from 110 volts which is in the us what the what the voltage what the peak voltage actually it's the uh is that the peak or the rms um it's the the voltage that characterizes this alternating current if you want to go from the 110 volts that is the the mains power to the whatever it is six volts or nine volts that you need for your computer um that you have to pick you have one transformer for 110 volts if you go to england where it's 220 volts or 240 volts um you um you need a different transformer but this methodology for making a transformer allows you to have one transformer that doesn't care what the uh what the input voltage is it can it can do the the transformation of the electric current so that's kind of the the story of um another question of whether you can transmit whether there are ways to transmit direct current over long distances one of the things that happens is if you could make an electrical wire that has much lower resistivity where the electrons you know usually when electrons go through a wire they keep on bumping into atoms effectively in the wire and they keep on being sort of not allowed to just keep just flowing without running into any interference so to speak and in superconductors they're special materials which are sufficiently low temperatures the electrons can just flow through them without effectively running into any obstructions at all so they have zero resistivity zero resistance so the electrons just go through without without running into anything without dissipating any heat so the problem with superconductors is they only exist the only ones we know about are ones that exist at low temperatures so you have to have liquid nitrogen or even liquid helium um to cool down this uh this material to the point where superconductivity happens it's long been thought one day there will be a room temperature superconductor at which uh sort of just in the everyday world it'll be possible to have a superconducting wire if that happens the electrical grid will be transformed i'm sure dc will be what's used most of the time and the electrical wires we see which are now right now made of copper usually um will will transform to being made of high temperature superconductor at which point um the the the economies of of uh ac versus dc will change okay well i should probably um uh oh there's one question here i have to i have to maybe address from motu if i walk into a very powerful solenoid will the magnetic fields affect my nervous system or have any other physical effect on me that is a much debated question so in clearly in us humans so okay magnetic field if you're in a very big magnetic field uh for example mri machines usually use very big magnetic fields so if you're if you're getting an mri done and you're in the mri machine there's usually a pretty big magnetic field that's going through you because that's necessary for the way that an mri machine works that's the m in magnetic resonance imaging is uh it's all about the magnetic field so the um uh so the question is uh if you have a big magnetic field going through you does it have an effect on you if you have a magnetic field that's rapidly changing is it different from having a magnetic field that's just a a static magnetic field that you would get from just a big magnet the answer is nobody really knows there isn't the probably the most obvious things nothing terribly obvious happens nothing terribly obvious goes wrong with us humans in high magnetic fields um but it's quite likely that there are actually things happening so it used to be said when i used to be involved in particle physics uh i used to go visit particle physics experiments and uh this was in days when they were a little bit less fancy than they are today when they only cost a few million dollars rather than tens of millions of dollars but they were they would often have these huge magnets and physicists would say you know what if i stand inside that magnet when it's on i get a funny taste in my mouth and that's not a completely crazy thing because the way that taste works it involves electrochemistry and it can involve electrical magnetic there could be an interaction of a magnetic field with some of that electrochemistry i have to say that that um uh i have had the belief which i don't know if it's correct that um uh when when i pick up a a strong magnet my fingers start tingling and uh my children at various times have tried doing experiments on me to see if it's really true and the the results of the experiments are are let's say they're not they're not definitive um but there's sort of a question if you pick up a very heavy magnet and you know you have that magnetic field particularly if you wiggle the magnet around can you when you have a changing magnetic field from wiggling a magnet around you can induce electrical uh currents and the question is can you induce electrical currents in your nerves that will cause you to like feel tingling in your fingers just from wiggling that magnet around i would say that in me the the jury is still out on whether that's that really happens or not now if you ask do magnetic fields have an effect on anything in humans we're not sure about humans in pigeons it's pretty likely magnetic fields have an effect it's pretty likely that in the brains of pigeons there is something that detects the magnetic field of the earth and that's used by pigeons to as their sort of gps-like device to know how they navigate to to be able to be a homing pigeon and navigate back to to where you were coming from and so on and uh so it's believed that there are devices in a pidgin's brain that are sensitive and magnetic fields and which presumably would be completely scrambled um by uh by a big magnetic field i mean presumably if we humans have that kind of thing it's it's kind of like how does the pigeon feel about its view of the world you put the pigeon in a big magnetic field and you know the pigeon is going to go you know look around and you can have a big debate about whether this pigeon is having a good time or a bad time whether it's having the time of its life getting this big magnetic field it's like i'm having a great bath of magnetic field this is wonderful whether it's having a terrible this is horrible i'm you know i've got this this terrible sensation in my brain it's hard to know because it's it's tough to know what pigeons are thinking but um uh whether we humans and so the the mechanism by which uh magnetic fields affect uh can affect neurons in a pigeon's brain it's not well understood it's thought to be something to do with it's thought to be something quite quantum mechanical it's thought to be something where it has to do with electron spins it has to do with the the best current theory i think is that certain chemical processes well certain processes in biomolecules and the molecules that exist in proteins and things like that that there are places in proteins where the direction of an electron spin can affect the rate of certain chemical reactions and so what happens is when there's a magnetic field that causes the direction of the electron spin to turn and it usually electron spin doesn't have any effect on the rate of a chemical reaction that the claim is in some particular cases in for some particular proteins and things that it can have an effect and that that's how you can have sensory uh you can have a sensor that is sensitive to magnetic fields and you know the fact is that that one of the things that happens is proteins tend to pick up all kinds of like a protein will have a it's made of amino acids always made of the same kind of stuff but it can have these cages that can enclose particular kinds of atoms like hemoglobin in our blood has a cage that is specially set up to enclose an iron atom and that's that's how we get the iron that transports oxygen in our blood is the hemoglobin uh protein has this cage where as soon as you get any reasonable amount of iron the iron will get stuck in this cage in in the hemoglobin and then it can be used to transport oxygen and there are similar theories about various atoms getting stuck in cages and that's some related to the way this magnetic field sensing works all right i think it is time for me to go to something different here um so thanks for lots of uh lots of interesting questions and i look forward to addressing more of them um next week and so um uh come back again next week and i think next wednesday i will also be in a live stream and we're doing those in a two week cycle and so next wednesday is about business and innovation and then this time uh next week i'll be doing another science and technology q a so thanks for joining us and

Channel: Wolfram

Views: 1,110

Rating: 4.9148936 out of 5

Keywords:

Id: ti5kid6cUUA

Length: 77min 29sec (4649 seconds)

Published: Fri May 07 2021