SGU Episode 882

This episode is in the middle of being transcribed by Hearmepurr (talk) as of 2022-08-22. To help avoid duplication, please do not transcribe this episode while this message is displayed.

Template:Editing required (w/links) You can use this outline to help structure the transcription. Click "Edit" above to begin.

SGU Episode 882
June 4th 2022
(brief caption for the episode icon)
SGU 881                      SGU 883
Skeptical Rogues
S: Steven Novella
B: Bob Novella
C: Cara Santa Maria
J: Jay Novella
E: Evan Bernstein
Quote of the Week
When I was a kid, people wanted to be an astronaut. Today, kids want to be famous, and that's totally the wrong approach. You have to have authenticity in what you're doing. You have to really care about the core message of what you're saying, and then everything else will fall into place.
David Copperfield, American magician
Links
Download Podcast
Show Notes
Forum Discussion

Introduction, Skipping horrible news

Voice-over: You're listening to the Skeptics' Guide to the Universe, your escape to reality.

S: Hello and welcome to the Skeptics' Guide to the Universe. Today is Wednesday, June 1^st 2022, and this is your host, Steven Novella. Joining me this week are Bob Novella...

B: Hey, everybody!

S: Cara Santa Maria...

C: Howdy.

S: Jay Novella...

J: Hey guys.

S: ...and Evan Bernstein.

E: Good evening folks!

S: So here we are. We're trying to figure out what we're going to chat about at the top of the show. And like there's just there's just like horrible news happening in the world, you know what I mean? Like there's really nothing good that we would that we wanted to talk about or that wouldn't sound tone deaf or insensitive. Like you can't talk about something happy and there's only a bunch of sad stuff to talk about. So we decided just not to talk about anything.

C: Get on with it.

J: Let's talk about science.

C: Oh yeah.

S: Let's talk about science. Let's get right so let's, why don't we just get right to our news items and make people feel better about some awesome cool science.

E: Okay.

S: Not to diminish all the bad stuff but this is what we do. All right here we go.

News Items

(At some point in this episode, the Rogues mention the word/term consilience. This is arguably a good word to have tagged for the vocabulary section.)

Exascale Supercomputer (1:05)

• Absurd U.S. Supercomputer Becomes First to Officially Enter Coveted Exascale Status

S: Bob tell us about I know you've been waiting what for years for this Bob, right? An exascale super computer.

B: Okay. So, on SGU Episode 151--

C: --Whoa!

B: --June 9th, 2008--

E: --Oh my gosh!

B: --I announced the first petaflop supercomputer: a quadrillion calculations per second, 10¹⁵. I have been waiting 5105 days to announce this. That an exascale supercomputer has finally been create. Meaning it can do at least one quintillion calculations per second.

C: Wait Bob, why petaflop but exascale?

B: It's an exascale meaning that it has breached the realm of the exascale computing of quadrillion or quintillion calculations the second. So that's an exascale is that scale of─

S: Order of magnitude.

B: ─quintillion or more. The flops is just like you have to have essentially at least one exaflop to be in the exascale.

J: Bob can you give us some perspective on like how much more powerful this is? Like give it on the spectrum where is this.

B: I mean it's hard to put numbers like quintillions in context, right?

C: Yeah.

B: I mean what is a quintillion. I am told that not everyone is obsessed with such large numbers as I am and I of course respond I don't believe you. But let me see if I can put a quintillion into context. It's 10¹⁸. 18 zeros. It's a billion billion. Or this is cool. It's a million million million. Now, those are multiplied. They're not added. When I say a million million million that's not three million. That's a million times a million, times a minute. Just throwing that out there. Okay here's another one. The Milky Way galaxy is one quintillion kilometers wide. Okay, that helps a little bit. One more. Every human, including Jay, doing calculations for four years - the frontier supercomputer can do all those calculations in one second. So it's huge. It's like four times more calculations than number two on this on this TOP500 list of super computers. So yeah it's. It's a milestone. And it's interesting because often if something is ten times bigger, or better, or whatever, an order of magnitude that's dramatic. But for super computers once you have three of those orders of magnitude a thousand times better the number changes from 500-600 petaflops to one exaflop type of deal. At that thousand. And often with it with that three order of magnitude increase in speed you can do dramatically, dramatic new things. Faster things. Many things that you probably couldn't have even thought of with previous computers because it would just be too hard and take too long.

S: Bob there's going to be a new Netflix series called faster things.

J: (laughs) Oh my god.

B: Is that based on the book?

S: It's going to have a whole 80s vibe to it.

B: Oh okay. (laughs)

J: Bob what did they do with this computer?

B: Jay you're throwing it off my flow totally.

J: All right nevermind. Wow Bob. That's this is really awesome man.

B: Okay.

E: Tell us more what you were gonna say next.

B: Okay but I gotta say but listen to this. China already has probably a couple true exascale systems. Exascale super computers. Already have it probably. But they mysteriously, to me anyway, decided not to publish about it and get all those bragging rights. All of them. I mean that would have been all over the news..

E: State secrets man, state secrets.

B: They did of course-- but it hasn't really been that way in the past. They did of course make sure the news was leaked. So it's just kind of weird and but still I don't wanna detract from this accomplishment. Did I say that it's called frontier created by the United States department of energy's famous Oak Ridge National Laboratory.

E: Oh Oak Ridge.

B: Now Frontier. It's an astounding creation. It took more than 10 years of a collaboration between the United States Department of Energy's exascale computing project. But also with national laboratories, private industry, academia. And I think maybe there was a deal with the devil in there as well. So the key thing here, the meat, the meat of this announcement is this. The HPL score was 1.102 exaflops per second. That's it. That's all you need to know. We're done. No we're not. So flop. When I say exaflops a flop is floating point operation. It's just a calculation.

C: I don't like it. I don't like flop.

B: It's yeah. It just─

C: Makes me uncomfortable.

B: ─like yeah. I kind of like exascale as an alternate way to--but exoflop is very specific. And HPL is also critical that stands for High Performance Linpack. It's a benchmark test. It's very important. But it is not this crazy esoteric test to determine how awesome the super computer is. It's really just a dense system of linear equations. That's it. You go through these linear equations and if you do them at a certain in a certain amount of time, bam! You're doing 1.102 exaflops per second. It's a common benchmark and it's it's widespread. That's why they use it because it's used all over the place. And the vast majority of relevant systems can run this test. So that's important because if you had an obscure test that only a very specific type of computer could run it wouldn't be very helpful and it wouldn't be able to put all these computers in context. Remember this is number one on the TOP500 list. So 500 of these super computers need to be able to run this test, this landmark test. So how does it even do this? What's under the hood? You open the hood, what do you see? You don't see a hemi you see Hewlett Packard Enterprises Cray EX platform.

E: Oh the Cray.

B: Yes. yes. And Evan you remember that because─

E: I do.

B: ─that to me that was very nostalgic for me because I remember─

E: The 80s my gosh.

B: ─oh dude you nailing it. I remember the Cray-2 in 1985. The first supercomputer that reached the milestone of one gigaflop. Cara. 1 billion operations a second. That was the fastest computer on the planet. I remember that day and it seemed fast. Frontier is a billion times faster. A billion times faster. The phone in your pocket right now is faster than that Cray-2 but it's awesome to see Cray still being used. I'm not sure─

E: Yeah the Cray brand has legs.

B: Yeah I remember, do you remember what the Cray used to look like? It was basically like one tower with a like a seating, a circular seating all around it. It was very compact and very small compared to what we got today. Okay enough of memory lane. So the HP's Cray EX platform. There's 74 cabinets. It's in 74 purpose-built cabinets. If you look inside those cabinets you're gonna find 9 400 CPUs, Central Processing Units, and 37 000 GPUs or Graphical Processing Units. The total core count is 8 730 112. That's a lot of cores.

J: Oh my god.

B: Also you got cables.

J: You know just the idea of them being able to chain that many─

B: Oh my god it's nuts.

J: ─close together. Like there has to be specific hardware and software that goes with that.

B: It's amazing. And don't forget Jay there's 90 miles of networking cables. Oh my god. That's 90 miles. That's crazy. All right so how do you think you'd even cool that. One of the things I remember reading about exascale supercomputers years ago it's like it's going to run so hot it's gonna consume so much power. How are you going to cool this thing. It's gonna be crazy. Well it uses four 350 horsepower pumps.

C: Wow.

B: Four 350 horsepower pumps moving six thousand gallons of water through the cabinets every minute. Every 60 seconds oh man.

E: The water it's still what you use?

B: It's great for carrying away heat man. And I remember it's also works against you Evan. Because I remember when I was scuba diving and you're in the water and that water takes away your heat. You hear your heat the heat being sucked away. It just cools you down so fast. It's too efficient. Enough of that. All right (laughter) also there's other milestones. And this is really cool. Frontier is number one on the GREEN500 list which scores commercial supercomputers for energy use and efficiency. Number one on that list which is really cool. And there was a new category this year called mixed precision computing. This is a brand new category. And that rates performances based on formats commonly used for artificial intelligence. And here the score was 6.88 exaflops. So you see this is even faster it seems. 6.88 exaflops. But it also it depends on the test, right? Because there's so many ways that you could, how do you encapsulate performance overall performance. You can't. You can't get one number for overall performance. There's lots of different ways to do it. So this is great because it's AI. It's based on these AI formats which of course are becoming increasingly common and all over the news. So the next inevitable question you should be asking is well what can it do that's so special? How do you even?

E: Solve Worlde. (Bob laughs)

C: I got Wordle in two today.

B: It would solve Wordle in one attosecond.

S: I almost got it in two but it took me three tries to get the last two letters.

E: This computer will solve Wordle─

B: Boring.

E: ─before the word is even chosen.

B: Exactly. All right now of course the specific impact like this system and similar systems will have because there's gonna be a lot more of these exaflop super computers as in over the next five to ten years. So that the impact is you can't really predict it but you can say things that you can say that exascale supercomputers without question they're going to impact your life in potentially major ways. Precision medicine, climate change models, water use, material science, nuclear physics, national security, pornography (Cara laughs) you name it. It's going to have an impact. How about this: forecasting viruses that leap species. Zoonotic, right? To avoid the next pandemic. These computers can help with that. Here's another way to look at it. I like this one. Now imagine you're at work and you're investigating a promising new pharmaceutical compound. As one does at work I suppose. People do that all over the place and typically you're going to wait weeks weeks for the results. With Frontier you'll be able to know the answer in a few hours. So instead of waiting for almost a month you're going to get it in just a few hours. In between eating your breakfast bagel and your cafeteria lunch you're going to get the answer. Even better here's another way to look at it. Instead of waiting a half hour or more for the number crunching that you do on your project you can get the answer in seconds. In seconds. So you could stay at your desk. You don't have to get up, get a coffee, chat with your friends, do some other weird stuff whatever you do at work. Then you go back and get in seconds you could stay at your desk and then you can go to the next task. So I think it could really speed things up and really help with creativity. Because you're getting the answer right then and there. As you know in our society there's nothing like having immediate feedback.

C: This is starting to sound really like dystopian to me Bob.

B: Oh my god.

C: You can be tethered to your desk all day every day and constantly accomplish things. (laughs)

B: Oh my god. Pharmaceutical compounds!

E: Don't tease Bob.

B: Imagine! Oh look I just found a cure for cancer my day is so boring. (laughter)

J: Cara, Cara─

E: Wow this thing cures cancer.

J: Cara is this the perfect example of don't get Bob started?

C: I was gonna say that.

B: It is. It's weird though because it's like. It's just a tool. It's just a tool. You sure you could use this to do really nasty stuff. But hey man if we follow that down to its logical conclusion you'd say let's ban all cars because so many people died. This is just a tool.

C: To who are you straw manning right now? (laughs)

B: I'm totally straw manning. So if this pattern holds. Which it won't. I will announce on May 23^rd 2036 on SGU episode 1611 that some country, or company, or artificial intelligence perhaps has created a zeta-scale supercomputer with 10²¹ flops. That's a sextillion calculations per second. And you know the jokes are going to be flying when we have this zetaflop supercomputer sometime in the 2030s. Hopefully. Just real quick when we reach that milestone I'm going to be on the show freaking out as I am this show. And it could, people predict that it will be able to do things like for example accurately forecast the global weather for two weeks in the future. Pretty good. That's really good actually. Two weeks? That's pretty amazing because you're up to that hard limit of chaos and so I'm not sure how much how much deeper in the future we're ever going to be able to get in predicting something chaotic like weather. But two weeks and having a fairly accurate forecast using a zeta scale super computer. Pretty sweet. One more thing. They predict they predict that zetascale or yattascale systems─

E: Yes here we go.

B: ─might be able to accurately model the whole human brain. At least according to Kay Kirkpatrick who published as much in 2019. I think we're gonna see some amazing stuff with the zetascale and yattascale super computers but we now have exascale supercomputers and look forward to it. It's gonna do some amazing things. Good and bad Cara.

C: Are your-- (laughs) I didn't say it was going to do bad things. I don't want to not be able to get up from my desk. That's all I said.

B: You just harshing oh I get you, all right.

S: You can get up from your desk you just don't have to to waste your time.

C: Exactly. Like those breaks are actually healthy. No what I was wondering Bob based on Steve's news item last week. Your estimation of when we're gonna hit this next milestone. Is that utilizing a linearity bias?

B: Oh yeah absolutely that. And that's specifically why I said─

S: If the trend continues which it probably won't.

B: ─no I didn't say trend. I said if this pattern holds. Which it won't. So I just absolutely this may take there are some predictions that say this could happen you know even before 2030. I doubt that. A lot some people are predicting in the 2030s. They could be, they could hit some fundamental problems that are going to prevent this and delay it for even more. Maybe 2040s. But I think it's you're pretty safe to say sometime in the 2030s. That's a pretty big net. 20-2030s

E: Bob when did China come up with their exascale computer?

B: It's I think it was─

E: Years ago?

B: ─like--oh no no. Earlier this year, late last year.

E: Okay so it's been months.

B: Relatively recently but it's just. And kudos to them if it just I know so little about it because they just didn't publish it. And they didn't submit it for this.

S: They're not transparent so it doesn't count.

E: Well. Yeah, that part of it.

B: Yeah. In a sense. In a real sense it doesn't count. But I mean it if it truly existed─

S: That's the thing.

B: ─and it seemed like it really did. Some reliable sources say that this thing does exist. And it did pass the landmark with over one exaflop. So but if it did exist it does and there's no taking that away from them but I mean sorry. You're not in the limelight because you chose not to be in the limelight and we got to give the Oak Ridge National Laboratory all the kudos it deserves for this.

S: All right thanks Bob.

Dinosaurs Warm or Cold Blooded (16:17)

• Were Dinosaurs Warm or Cold-Blooded?

Preventing Violent Crime (30:17)

• A study gave cash and therapy to men at risk of criminal behavior. 10 years later, the results are in.

Revising Evolutionary Trees (46:29)

• Molecular phylogenies map to biogeography better than morphological ones

New Optical Illusion (1:02:39)

• This illusion, new to science, is strong enough to trick our reflexes

Quickie with Bob (1:11:44)

• Gato AI

Who's That Noisy? (1:14:18)

Answer to previous Noisy:

All vocalizations from Smooth Criminal by Michael Jackson

_hearme_put_text_about_answer_here_

New Noisy (1:17:04)

[high pitched whoops and grunting, squeaking calls]

J: The more information you give me, the better. If you think you have an answer, or you want to send me in something cool, just like that Noisy--somebody just sent that in to me--you can email me at WTN@theskepticsguide.org

Questions/Emails/Corrections/Follow-ups (1:18:59)

_consider_using_block_quotes_for_emails_read_aloud_in_this_segment_
with_reduced_spacing_for_long_chunks –

Science or Fiction (1:28:40)

Item #1: A new analysis finds that the Cueva de Ardales in Spain, famous for its many cave paintings, was likely occupied by human ancestors for as long as 500,000 years. Item #2: In a prospective study, dogs were able to detect asymptomatic SARS-CoV-2 infection with higher sensitivity than nasopharyngeal antigen testing, 97% vs 84%. Item #3: Scientists have developed a molecular drill that is activated by visible light and rotates at 2-3 million times per second, which can drill through bacterial membranes and can be used as a broad-spectrum, rapid-acting antibiotic for skin infections.

Answer	Item
Fiction	500,000-year occupation
Science	Dogs better than antigen test
Science	Molecular drill

Host	Result
Steve	sweep

Rogue	Guess
Bob	Molecular drill
Cara	Molecular drill
Evan	Molecular drill
Jay	Dogs better than antigen test

Voice-over: It's time for Science or Fiction.

Cara's Response

... [in Cockney accent, as in The Rain in Spain] "the kive in Spine falls minely..."

Skeptical Quote of the Week (1:46:35)

When I was a kid, people wanted to be an astronaut. Today, kids want to be famous, and that's totally the wrong approach. You have to have authenticity in what you're doing. You have to really care about the core message of what you're saying, and then everything else will fall into place.
– David Copperfield, American magician

Signoff/Announcements (1:51:16)

S: —and until next week, this is your Skeptics' Guide to the Universe.

S: Skeptics' Guide to the Universe is produced by SGU Productions, dedicated to promoting science and critical thinking. For more information, visit us at theskepticsguide.org. Send your questions to info@theskepticsguide.org. And, if you would like to support the show and all the work that we do, go to patreon.com/SkepticsGuide and consider becoming a patron and becoming part of the SGU community. Our listeners and supporters are what make SGU possible.

[top]                        

Today I Learned

• Fact/Description, possibly with an article reference
• Fact/Description
• Fact/Description

References

Vocabulary