Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:00
Coming up on Radiolab. All people
0:02
should have the right to acknowledge. The story of
0:04
one woman's attempt to make all knowledge
0:07
available to everyone, everywhere,
0:10
for free. That's a thing called PSY-HUB.
0:13
Listen wherever you get podcasts.
0:19
Listener supported WNYC
0:22
studios.
0:24
This is Science Friday. I'm Ira Flatow. Later
0:27
in the hour, the connection between warmer
0:29
temperatures and home run slugging,
0:32
yes, and a hopeful video game
0:34
about climate change. But first,
0:37
last week I was having this deja
0:39
vu. I was recalling a time way
0:41
back in 1975 when
0:44
scientists called a halt to their research
0:46
to discuss the possible consequences
0:48
of what they were doing. And back then
0:51
it was the shiny new tool of genetic
0:53
engineering recombinant
0:54
DNA that
0:56
caused Paul Berg and Maxine Singer
0:59
to organize a meeting of scientists to
1:01
draw voluntary guidelines to
1:03
ensure the safety of recombinant DNA
1:05
technology. It was called the Asilomar
1:08
Conference. Well, I was having those deja
1:10
vu thoughts last week when I learned of
1:12
another group of scientists releasing
1:14
an open letter warning of hazards
1:17
of a current tool called artificial
1:19
intelligence. It sounded all
1:22
too familiar and it stated, quote,
1:24
we call on all AI labs
1:26
to immediately pause for at least six
1:29
months the training of AI systems
1:32
more powerful than GPT-4
1:35
and called for researchers to jointly develop
1:37
and implement a set of shared safety
1:40
protocols governing the use of
1:42
AI. The letter was signed by a collection
1:44
of technologists and computer researchers,
1:47
including big names like Apple co-founders
1:49
Steve Wozniak and Tesla's Elon
1:52
Musk.
1:52
But others called the letter just another
1:54
round of hype over the AI field.
1:57
Joining me to talk about that is Dr. Stuart...
2:00
Russell. He's a professor of computer science
2:02
at Berkeley, director of the Kavli
2:04
Center for Ethics, Science, and the Public, and
2:06
co-author of one of the leading AI
2:08
textbooks. And he's a signatory
2:11
to the open letter I just mentioned. Welcome
2:13
to Science Friday.
2:15
Thank you, Ira. It's nice to be with you. Okay.
2:17
As I say, you're a signatory to this letter.
2:19
Why did you sign it? Why do you think a pause
2:22
is needed? In my view, the
2:24
AI systems that are currently being
2:27
developed and the ones that have been released recently
2:29
based on a technology called large
2:31
language models,
2:32
represent a type of technology that
2:35
is intrinsically very
2:38
difficult to understand
2:41
and very difficult to guarantee that
2:43
it's going to behave in a safe
2:45
way. So in a very
2:48
immediate sense, it presents risks,
2:51
not the sort of apocalyptic risks
2:54
of taking over the world and extinguishing
2:56
the human race, But real risks,
2:58
for example, last week in Belgium, a
3:01
man was reported to have committed
3:03
suicide directly as
3:05
a result of his relationship with
3:07
one of these chatbots, which was actually
3:10
advising him and as it were holding
3:12
his hand
3:13
while he was in the process
3:15
of committing suicide. The reason
3:18
why these systems are
3:20
very hard to provide any guarantees
3:22
for is that they are enormous black
3:25
boxes.
3:26
Can you sum up for me in 2500
3:28
words or less how these systems
3:30
work? So a large language model
3:33
is something that very simply predicts
3:35
the next word given the
3:37
sequence of preceding words in
3:40
a text or in a conversation.
3:42
And so you can use that for an interactive
3:44
conversation. If you put in a question,
3:46
then it will start generating words that
3:49
look like
3:49
an answer. And how do you make
3:52
them? You start with a blank slate of
3:54
about a trillion parameters in
3:56
an enormous what's called neural network.
3:59
you do about a billion trillion
4:02
random modifications of
4:04
those parameters to try
4:07
to get that network to become very good
4:09
at predicting the next word from a training
4:12
set
4:12
that is maybe 20 trillion
4:15
words, which is roughly comparable
4:18
to all the books that the human
4:20
race has ever written in the history
4:22
of of civilization. So that
4:24
system, when you interact
4:27
with it, displays remarkable
4:29
abilities. And I don't want to
4:32
disparage it in the sense that it can provide
4:35
lots of benefits for users
4:37
for companies. But it's
4:40
a black box. We do not understand
4:42
anything about how it works.
4:44
And the only way we have to
4:47
get it to behave itself, for example, not
4:49
to advise people on how
4:51
to commit suicide is to
4:53
essentially say bad dog or
4:56
good dog. And that's the process
4:58
that OpenAI, the creators of GPT-4
5:01
went through to
5:02
try to get it to behave itself.
5:06
They just hired a lot of people who would engage
5:08
in lots of conversations and every time it did
5:10
something they don't like, they would say bad
5:13
dog. And if it produced a good answer,
5:15
they would say good dog. And then
5:17
hopefully the system would adapt its parameters to
5:20
produce bad behavior less often. And they
5:22
proudly announced
5:23
that in terms of these forbidden
5:26
things like advising people to
5:28
commit suicide,
5:29
telling people how to make chemical weapons, giving
5:32
unlicensed medical advice, that
5:35
it was 29% better than the previous iteration of
5:39
their system. But 29%
5:41
better is still a very
5:43
long way from perfect because they have
5:45
actually no control over it. So
5:47
we're simply asking that
5:49
before you get to deploy a system
5:53
that's going to affect the lives of
5:55
millions or even billions of people. you
5:58
take sensible precautions.
6:00
to make sure that it doesn't present
6:02
undue risks and that it remains
6:05
within predictable guidelines and
6:08
so on. So that's the real reason behind
6:10
this request for a moratorium.
6:12
I think there are longer term issues
6:15
at stake here, not from the present
6:17
systems, but from future
6:19
generations of AI systems that may be
6:21
much more powerful still,
6:23
and they present correspondingly much
6:25
greater risks. Well, do they, these
6:28
future systems, present the
6:30
risk that Stephen Hawking was talking about
6:32
in 2014 when he said the development
6:34
of full artificial intelligence
6:36
could spell the end of the human race?
6:39
Theoretically, we don't
6:41
know when that type of
6:43
system, which we call sometimes artificial
6:46
super intelligence, we don't know when that's
6:48
going to arrive.
6:49
But if it does arrive within
6:52
our current approach to how we
6:54
build AI systems, in particularly these
6:56
black boxes, we would have no
6:58
way of ensuring that it's
7:00
safe in the sense that its behavior
7:03
is actually aligned with what the humans want
7:05
the future to be like. And then you're basically
7:08
setting up a chess match between us
7:11
and a system that's actually much
7:13
more intelligent than us and has
7:16
already thought of every possible countermeasure
7:18
we could try.
7:19
And so that's, in a real sense, the
7:22
loss of human control over the
7:24
future. So that's the
7:26
risk that Stephen Hawking is talking about. I
7:28
want to emphasize the current systems
7:31
do not present that risk, as far
7:33
as we know, to the extent that we understand
7:35
them at all, which is not very much. We
7:37
think they have
7:39
some fundamental limitations on their ability
7:42
to plan their future
7:44
activities. But at the rate
7:46
of progress we're seeing in AI, We
7:48
need actually to develop
7:51
methods
7:52
to ensure that when we build systems
7:54
that are more powerful than us, we
7:57
somehow retain power over
7:59
them. forever. If
8:02
that sounds like a difficult problem, it's
8:04
because it's a difficult problem. Well,
8:07
practically speaking then, what
8:09
do you expect people to do who
8:11
are in AI research? Is the horse
8:14
already out of the barn and are people willing
8:16
to listen to the signers of this letter
8:18
and pause? Or is it doesn't
8:21
take a lot of fancy lab equipment
8:23
like it did with genetic engineering to
8:25
move ahead? So I think this is a great
8:28
question. And your example of the of
8:30
the genetic engineers is
8:32
a really good one. And so Paul
8:34
Berg, who was one of the organizers
8:37
of that 1975 workshop in 2008, he wrote a retrospective. And
8:42
the last paragraph says there's
8:44
one lesson from Asilomar,
8:46
which is where they had the workshop, a lesson for
8:49
all of mankind.
8:50
And basically once commercial
8:53
interests start to dominate the conversation,
8:56
it will simply be too late. It's
8:58
all about the money. It's all about the money.
9:01
And often people's thinking
9:04
and decision making becomes very distorted
9:07
when we're in that situation. There's an old
9:09
saying, you can't get someone to understand
9:11
something if their livelihood
9:14
depends on not understanding it.
9:16
And I think there's a little bit of that going on here. In
9:20
the past, some of the principles such as
9:22
Sam Altman, the co-founder of
9:24
OpenAI, you You know, STEM has said
9:27
that there may come a point when governments
9:29
need to intervene and impose constraints
9:33
and basically
9:35
not release further systems until
9:37
they meet
9:38
certain kinds of safety properties. And
9:40
the petition is simply saying, well, maybe
9:42
this is that time. It's also
9:44
worth noting that the OECD, which is,
9:47
you know, an international
9:49
organization that all the advanced Western
9:52
economies are members of have issued
9:54
AI guidelines, called the OECD
9:56
AI
9:57
principles that have been ratified by
10:00
all the member states
10:01
that very explicitly say that AI
10:04
systems have to be robust and
10:06
predictable and you have to show that they don't
10:08
present an undue risk
10:10
before you can deploy them. So arguably
10:13
all the major governments have already supported
10:16
the petition that we are making. You
10:18
know, in genetic engineering, there are
10:20
all these ethical guidelines, but
10:22
there are still people who want to clone
10:25
a baby. Is there a way to protect
10:27
against a rogue AI researcher
10:29
who wants to ignore ethical
10:31
guidelines?
10:33
That's a tough, I mean, the rogue
10:35
thing, I think we have to work with
10:37
the hardware manufacturers because they're
10:39
the bottleneck, right? And there's only
10:41
a handful and they've already
10:43
agreed in the past, for example, with digital
10:46
rights management, that was a global
10:50
operation to get the hardware manufacturers
10:52
to implement digital rights.
10:54
So, I think it's not impossible
10:57
that we could get safety mechanisms built
10:59
into hardware where they just will refuse
11:02
to run programs that are not certifiably
11:04
safe.
11:05
So where does your mind take you from here? Are
11:07
you hopeful about the AI future
11:10
or more fearful than hopeful? I mean,
11:12
you gotta have a little bit of both there, right? So
11:14
I think I'm sort of naturally an optimist
11:16
and I've been working for about 10
11:18
years now. I want to try to understand how do
11:20
we retain power over systems more powerful
11:23
than ourselves, right? That's what
11:25
I call a control problem.
11:26
And I think there's a feasible path to solving that
11:29
problem. Then we've got to convince everyone
11:31
to adopt that approach so that unsafe
11:34
systems are not created.
11:37
And then we've got to make sure that somehow no
11:40
one, either deliberately or
11:42
accidentally creates an unsafe system and
11:45
unleashes it on the world. So there's a lot to do, but
11:48
I'm cautiously optimistic.
11:50
Am I thinking that
11:53
I had better hurry up? we had better hurry up in nailing
11:56
down these solutions and getting them into
11:59
the policy process. I
12:01
think, yes, I think my
12:03
estimate of when we'll have powerful
12:06
AI systems that could present a
12:09
major control risk has moved
12:11
closer to the present than
12:14
it was a few years ago. Well,
12:16
from your mouth to AI's ears, Dr.
12:18
Russell. Thanks a lot, Ira. It's been nice talking to
12:20
you. Dr. Stuart Russell, professor
12:22
of computer science at Berkeley. He's
12:24
director of the Kavli Center for Ethics Science
12:27
and the Public there. and he's the author of
12:29
the book, Human Compatible, Artificial
12:32
Intelligence and the Problem of
12:34
Control. Thanks again for joining us today. Thank
12:36
you. Bye-bye.
12:37
After the break, a look at the crew for the upcoming
12:40
Artemis mission to orbit the moon. Stay
12:42
with us.
12:44
Support for Science Friday also
12:46
comes from Schmidt Futures, a philanthropic
12:49
initiative founded by Eric and Wendy Schmidt.
12:52
As a Science
12:52
Friday listener, I think it's fair to assume that
12:54
you are a person who is curious about
12:57
pretty much everything. So I have a podcast
12:59
recommendation for you. I'm Steven
13:02
Dubner, the host of the podcast, Freakonomics
13:04
Radio. Every week we hear from
13:07
scholars, thinkers, and doers
13:09
as we tackle big questions like,
13:11
why is the banana in danger and
13:13
can it be saved? And has
13:16
disgust gone from keeping us safe
13:18
to getting in our way? So
13:21
give us a try. Listen to Free economics
13:23
radio wherever you get your podcasts.
13:26
This is Science Friday. I'm Ira Flato. This
13:29
week, NASA announced the four-person
13:32
crew of the Artemis 2 mission. That's
13:34
a mission to send astronauts to circle
13:36
the moon. There's three firsts
13:39
in the group. First woman, first
13:41
person of color, first Canadian
13:43
to travel to the moon. And while
13:45
these Artemis 2 astronauts will not
13:48
actually set foot on lunar soil, It's
13:50
an important milestone for NASA's first
13:53
moon mission since Apollo. Think
13:55
Apollo 8 v 2.0.
13:56
Joining
13:59
me now to
14:00
talk more about this week's announcement and the
14:02
future of the Artemis mission
14:04
is my guest, Swapna Krishna, host
14:06
of the PBS digital series, Far
14:08
Out. She's based in Philadelphia,
14:10
PA. Hey, welcome to Science Friday.
14:13
Thank you so much. Thanks for having me. You're
14:16
welcome. NASA really pulled
14:18
out all of the stops this week announcing
14:21
the crew for Artemis 2, even making a video
14:24
trailer, we might call it, of the
14:26
crew, which looks and sounds like a Hollywood
14:28
movie trailer. Let's listen to that. I'm
14:31
Christina Cook. I'm
14:33
the mission specialist. I'm Jeremy Hanson. I'm
14:37
a mission specialist. I'm Victor Glover.
14:41
I'm the pilot. I'm Reed
14:44
Wiseman. I'm the commander for the Artemis 2 mission
14:46
to the moon. To the moon. To
14:49
the moon. Wow,
14:52
that is to the moon swap. Is it not? It
14:56
is. It is. It's absolutely hilarious.
14:59
I love it. I love the pomp and circumstance
15:01
around these things. So like personally, I love
15:03
it, but it's also so earnest
15:05
and so cheesy and it just makes me, it
15:07
makes me laugh. Who is it aimed at? I
15:10
think it's aimed at kids. I think it really
15:13
is like it'll get anybody pumped up.
15:15
This is the Hollywood treatment, as you mentioned,
15:17
of
15:18
astronauts in the movies that we've
15:20
seen this bombastic music,
15:23
you know, the stepping out of the shadows. I
15:25
showed it to my
15:26
four-year-old and he said, Mama,
15:29
these are heroes. And I was like,
15:31
okay, yeah, I see that. Yeah, I mean, this
15:33
is a generation that's just seen robots
15:35
going to different places, Mars,
15:37
the moon, things like that. And
15:39
not people, like I did,
15:42
my generation. I grew up loving
15:44
space, but we didn't get this.
15:46
So it's about the younger generation,
15:48
sure, but for us,
15:51
it's always exciting to send people to space regardless
15:53
of where they're going. But it was
15:56
very much, you know, why aren't we doing
15:58
this? Why aren't we going further out? been
16:00
to the moon. Why are we not
16:03
there now? And so it's exciting for
16:05
me to see this as well.
16:06
Let's talk more about the makeup of
16:08
this four person crew.
16:11
Who are they? We've got Jeremy Hansen,
16:13
who's a rookie. He's our Canadian astronaut
16:15
on this mission. Christina Koch,
16:18
who I believe she has the record
16:20
for the longest duration
16:23
single space flight by a woman because
16:26
of her stint on the ISS. She's a mission specialist,
16:28
as is Jeremy Hansen. We have Victor
16:30
Glover, who's going to be the first person
16:32
of color. And he was
16:34
the pilot for Crew 1, so very
16:37
experienced. And then Reid Weissman is
16:39
the commander of the mission, also
16:41
very
16:42
experienced. And he was the chief of the astronaut office
16:44
before my speculation
16:47
is he resigned because he wanted
16:49
this flight. So there
16:52
were a lot of questions. Why do you resigned.
16:54
And I think now we know.
16:56
You know, I look at the diversity,
16:58
and I think that's what NASA is aiming for,
17:01
right? A diverse crew
17:03
to land or at least be the first
17:06
people to circle the moon.
17:09
And I'm thinking that, wait,
17:11
you know, to be clear, NASA deployed
17:13
diverse crews to the space shuttle for what, 30
17:16
years? And now it's,
17:18
you know, isn't it almost an everyday
17:21
thing, an everyday occurrence?
17:23
I think it absolutely is for,
17:26
you know, going to the ISS, going to low
17:28
Earth orbit, but we've only ever had white men
17:31
go to the moon. I think it's good that we're paying
17:33
attention to it. I think a lot of people are a little
17:35
cynical about like, well, this should have already happened. But
17:38
yeah, we haven't been in the moon in 50 years. I am glad they're making
17:40
explicit that
17:41
this will be the
17:44
first woman, the first person of color. And
17:46
that's been a part of the Artemis program ever since
17:48
the former NASA Administrator Jim Bridenstine kind of
17:51
pulled together this
17:52
program from, you know, what
17:55
was left after it had
17:57
been gutted over and over again.
17:58
Let's talk about the mi-
18:00
as I say, it's not actually
18:02
landing on the moon. What is
18:04
the purpose? The
18:05
purpose is basically to test the hardware
18:07
and the software. We've been to the moon,
18:10
of course, but as
18:12
much as I wish, you know, the bombastic trailer
18:14
makes it feel like a movie, but it's not a movie.
18:17
You can't just like boost Apollo 11 from the National Air
18:19
and Space Museum, stick it on
18:21
top of a rocket and go back to the moon. Like
18:23
it's all new hardware, all new software. So
18:25
we need to test it. And before we can land on the
18:27
moon, we need to test how it does
18:29
in orbit. We did that a little bit with the first
18:31
mission, Artemis 1, which
18:33
took off in November, but now
18:35
this is the first crewed mission of the program. So what they're
18:38
going to do is take off and do a
18:40
lot of testing of the hardware and software
18:42
in low Earth orbit and high Earth orbit,
18:44
and then around the moon and back. And you
18:46
know, we're always testing how space
18:48
affects people, right, with the radiation,
18:51
the time in space.
18:53
Basically, because of the radiation
18:55
trapped around the Earth, thanks to the Van
18:58
Allen belts. We don't
19:00
have a ton of data on what
19:02
happens on longer duration space
19:05
flights to the moon and back. All we have
19:07
are those Apollo flights. So what this
19:09
will do is it's going to test what kind
19:11
of radiation these astronauts are exposed
19:14
to outside of, you know, the
19:17
Earth's magnetic field, what
19:19
will happen to them when they pass through the Van Allen belts,
19:21
and how safe is it to send astronauts
19:23
on these long distance and duration space missions.
19:26
We don't know.
19:26
Yeah. And we're also recreating
19:29
the iconic Apollo 8
19:31
mission, which went around the moon. And Bill
19:33
Anders took that
19:35
iconic Earthrise photo of
19:37
the Earth, right, peeking up behind
19:39
the moon. Are we going to have a moment like that, you think,
19:42
on this mission?
19:43
100%. We are definitely going to have a
19:45
moment like that. I think there wasn't as much
19:47
awareness, especially the early days of Apollo,
19:50
there wasn't as much awareness of how important
19:52
photography was Earthrise was actually
19:54
an afterthought by Bill Anders. They saw
19:56
this view and were like, Oh, we probably should take
19:59
a picture. I'm sure some somebody
20:00
will be assigned
20:01
to take a picture and to take many pictures
20:04
as we come back around from the far side
20:06
of the moon as this mission does. The images
20:09
are so important. The science is important. The
20:11
text is important. But in terms of communication,
20:13
you cannot beat someone who has a busy
20:15
day, who doesn't feel like they have time to sit down and read an
20:18
article about the mission, but they can get captured
20:20
by a photo. Not that we have all this anticipation
20:23
that we know what day we're
20:25
talking about for this launch.
20:27
We don't even know what year we're talking
20:30
about yet. The mission
20:32
is currently scheduled for no earlier than November 2024.
20:36
That is probably unlikely. You
20:38
know, 2025 is likely, I think, for this
20:41
mission. Delays are part of the game
20:43
when you're talking about space and, you
20:45
know, SLS did have a lot of problems
20:48
for Artemis 1's launch. There's
20:51
a lot of stuff to work through and because this is going
20:53
to be the first mission that humans are
20:55
on, there's going to be even more to work through.
20:57
But hopefully,
20:57
they'll have figured out a lot of these propulsion issues
21:00
before the rocket is even rolled out. And
21:02
that's a big difference also between the Apollo
21:04
mission. I mean, it took only seven months
21:07
between circling the moon in what, December,
21:09
and
21:10
then they landed on the moon
21:12
back in July. And that bingo,
21:15
it wasn't years after that. Why such a
21:17
big delay here?
21:19
It's really simple, and it's money.
21:22
It's basically, there's just not enough funding
21:24
to get done what they want to get done to To save
21:26
money, SLS's engines were literally
21:29
ripped out of the space shuttles. Those RS-25
21:31
engines
21:32
were taken from the space shuttles and put onto SLS
21:35
in an effort to save money, but it
21:37
has actually
21:38
been the biggest
21:41
cost, one of the biggest cost overruns in NASA
21:43
history. It's always about the money,
21:45
isn't it? Yeah. Would any of the
21:47
astronauts in Artemis 2, the folks who
21:50
are going on this mission, be
21:51
beyond the mission that follows and actually
21:54
step foot
21:55
on the lunar surface?
21:57
Not in the mission that follows, just
22:00
because NASA doesn't assign astronauts
22:02
to back-to-back missions. It's likely that the Artemis
22:05
III astronauts will be assigned in training
22:07
by the time that Artemis II takes off. But
22:10
so no, not the next mission, but it could
22:13
absolutely be a subsequent mission. I would not surprise
22:15
to see somebody like Christina Koch or Victor Glover
22:18
assigned as a commander of a future mission.
22:20
Yeah, yeah.
22:21
We always hear that the eventual goal
22:24
here in these moon missions is a
22:26
stepping stone to go to Mars,
22:29
right? How close are we
22:31
really to that space mission
22:33
to Mars?
22:34
The short answer is we are not close. As
22:37
much as we would like for it to happen fast, we need
22:39
to get used to working and living in lunar orbit
22:42
and on the lunar surface
22:44
before we can move on to Mars. Artemis
22:46
IV, which is 2030 at this point, is
22:49
around putting lunar gateway
22:52
in orbit of the moon, which is
22:54
a lunar space station. It's gonna be like
22:56
the ISS, but in lunar orbit.
22:59
That's gonna be the hub for the Mars missions, but that has to be
23:01
assembled first. So if we're not talking about
23:03
assembling Lunar Gateway until 2030
23:06
and then completing, you know,
23:09
five to 10 years to complete that,
23:12
then we can go to Mars. And you know, so
23:14
it's gonna be a while is
23:16
the short answer. Well, we'll all look forward
23:19
to this swap down and you'll keep
23:21
coming back and talk about it. Will you please? I
23:23
will, yes. A human space flight,
23:25
it makes my heart flutter. So I'm so excited
23:28
about this. Very exciting.
23:30
And I hope everybody who's watching it, this generation
23:33
will be as excited as I was and all the
23:35
other folks when
23:38
we finally got somebody to land on the moon in 1969. Swapta
23:41
Krishna, host of the PBS digital
23:43
series, Far Out. Thank you for taking
23:45
time to be with us today.
23:47
Thank you.
23:51
Thank you. Yes,
24:05
the baseball season is in full swing
24:08
and you know I love baseball. And
24:10
the new rules imposed this year are aimed
24:13
at making the game even more enjoyable.
24:16
But Mother Nature may be imparting
24:18
a bit of her own rulemaking. Research
24:20
published this week in the Bulletin of the
24:22
American Meteorological Society says
24:25
that climate change, specifically
24:27
global warming, may make the
24:29
weather more favorable for home
24:31
runs
24:33
and it may favor some teams
24:35
over others. Joining me is Christopher
24:38
Callahan. He's a PhD candidate in geography
24:40
at Dartmouth College in Hanover, New Hampshire
24:43
and lead author on that report. Welcome
24:45
to Science Friday. Hi,
24:46
thank you very much for having me. You're
24:48
welcome. Okay, so how does warming make for
24:51
more home runs? Well, when
24:53
temperatures go up, the air is
24:55
less dense. This is a basic physical
24:57
mechanism that we know about. And we know
24:59
that when the air is less dense, there's
25:02
less air resistance. So it's easier for a ball
25:04
to fly through it. And so a batted ball
25:06
will simply carry farther and is more likely
25:08
to be a home run.
25:09
And how big of an effect are we talking about
25:11
here? At present, the effect is relatively
25:14
small. We can link climate change
25:16
to about 500 home runs over
25:19
the last 10 years, which is only about 1%
25:21
of the number of home runs in
25:23
total that were hit in Major League Baseball. That
25:26
being said, if we move into
25:28
the future and keep emitting greenhouse gases
25:30
substantially, this effect could get much
25:32
larger. And we could be talking about hundreds
25:34
more home runs per year later in this
25:37
century.
25:38
Wow, we need an asterisk, you know, in
25:40
the pre-climate change days?
25:43
Yeah, maybe. A new type
25:45
of asterisk. Can
25:47
you actually compare a hit in two different
25:50
ballparks? Well,
25:51
there's so many things that are different between ballparks.
25:53
We know that they have different dimensions, they're at
25:56
different elevations, the teams
25:58
are different, if it's two different teams.
26:00
And so it is difficult for us
26:02
to say that any two individual
26:04
hits are comparable, but we have enough
26:06
data over the course of the last 60 or 70
26:09
years on nearly every baseball
26:11
game, including the home runs and the temperature
26:13
in that game. And so we can make these sort of general
26:16
claims about average changes
26:18
in home runs due to climate change independent
26:21
of any particular hit in any particular ballpark.
26:24
Can you actually put a number on it? I mean, how many
26:26
hits per every degree of temperature? Yeah,
26:29
so we found that for
26:31
every one degree Celsius increase in
26:33
temperature, there's about a 1.9%
26:36
increase in the number of home runs in that
26:38
game. Wow. Now, there's only two or three
26:41
home runs in any given game on average. And
26:43
so that effect is not all that large
26:45
from the perspective of any one game. But
26:48
once you start thinking about climate change over the last
26:50
several decades across many
26:52
ballparks and many years, and then moving into
26:54
the future, the numbers can start accumulating.
26:57
This is Science Friday from WNYC
26:59
Studios.
27:01
You know, a lot of research papers are
27:03
statistical. They look at the statistics
27:05
of stuff on the extrapolate.
27:07
Are you looking at actual hits by
27:09
actual batters in your work?
27:11
We are. I mean, we are using the statistics
27:14
of those actual hits. So we have data
27:16
on the number of home runs in each game,
27:19
the temperature of that game. And then we also
27:21
use data on individual batted balls
27:23
from the more recent era when we have this data
27:26
from the stack cast system of high resolution,
27:28
high speed cameras in each ballpark.
27:30
And so we can use data
27:33
on actual hits by actual batters
27:35
and say, when a temperature is warmer, those
27:37
hits go farther. And we can observe that using real
27:39
data. Can you factor in any other changes,
27:41
perhaps differences in the ball, you
27:44
hear people talking about some years
27:46
the ball is juiced, or they think there's
27:48
something wrong with it. Can you factor that in?
27:50
Absolutely. So we do our best to control
27:53
for those other factors. We know that there are
27:55
changes in the construction of the ball over time.
27:57
We also know that different ballparks have different
27:59
dimensions. dimensions and different climates and different
28:01
elevations. And so we can factor
28:04
out those other things and say, what
28:06
is the influence of temperature independent
28:08
of those other things? Now, to be clear,
28:11
because we find that temperature has not
28:13
been the single thing driving increases
28:16
in home runs recently, it is still very
28:18
valid to say that much of the home run
28:20
surge might be due to, for example,
28:23
a juice ball. And so our results should not
28:25
be taken to disprove that theory.
28:27
It's just an additional thing on top of that.
28:29
So, are there teams or ballparks
28:32
where this would be most significant? So
28:34
we find moving into the future, there's
28:37
going to be an increase in home runs due to climate
28:39
change. But as you said, it's going to be different in different
28:41
places. And the things that shape
28:43
how different that effect is, aside
28:46
from simply how much global warming there is
28:48
in each individual place, is do
28:51
these ballparks have domes on them that insulate
28:53
them from ambient conditions? And
28:55
are they playing games in the middle
28:57
of the day or in the evening?
28:59
So for example, Wrigley Field, which is
29:01
walking distance from where I grew up in
29:03
Chicago,
29:05
is going to see the highest
29:07
increase in home runs in the future. The reason
29:09
for that is because it's open air, so it
29:11
doesn't have a dome that insulates it from the
29:13
ambient weather conditions. And most
29:15
of its games are played during the day when it's hottest
29:18
rather than in the milder evening conditions.
29:20
Wow.
29:21
Are you ready to go make book
29:23
on that? I am not.
29:25
So
29:28
luckily Red Sox versus Yankees should have similar
29:30
effects. We'll see, potentially.
29:35
This is great. It looks to me like you're obviously
29:37
a Chicago Cubs fan,
29:40
correct? Absolutely.
29:41
Did you get into this line
29:44
of reasoning and line of research because
29:46
you're a baseball fan? Yeah, I did.
29:48
I definitely would not have been motivated to think about
29:50
it if I was not already thinking about
29:53
baseball on my spare time. I
29:55
knew that people had hypothesized
29:58
about this sort of link between. temperature
30:00
and air density in home runs. And
30:02
I'm interested enough in baseball that I decided, well,
30:04
I'd like to go see if that effect is actually there
30:06
in the large scale data that we can use. And
30:09
so that's what motivated me to go do it.
30:11
Well, in a season where they're looking for more
30:13
base hits than home runs, we'll
30:15
take your advice and keep an eye out on
30:17
it. Thank you for taking time to be with us today.
30:20
Thank you very much for having me.
30:22
Christopher Callahan, PhD candidate
30:24
in geography at Dartmouth College in Hanover,
30:27
New Hampshire.
30:28
After the break, speaking of climate, a
30:30
new video game that lets you take on climate
30:33
change with a hopeful outlook.
30:35
We wanted to try and make something that
30:37
was inspired by the
30:39
genre of building games, but instead
30:42
of building a city, we're building
30:44
nature instead.
30:45
Stay with us.
30:47
Support for Science Friday also comes
30:50
from Schmidt Futures, a philanthropic
30:52
initiative founded by Eric and Wendy Schmidt.
30:56
This is Science Friday. I'm Ira Flatow.
30:58
Now if you follow our video games, you'll recall
31:01
we've been covering how game developers
31:03
are responding to climate change in their
31:05
games. Some are a little dystopian,
31:08
you could say. Others are a
31:10
bit more hopeful. Producer
31:13
Dee Pidderschmidt is here with me to play one
31:15
that I hear falls into the latter
31:17
camp, hopefully.
31:18
Hi Dee. Hey Ira. Tell
31:20
me about this game. Yeah, so it's
31:23
called Terranil. Have you ever played games
31:25
like Roller Coaster Tycoon or city skylines
31:28
or civilization, anything like that? Yes, where
31:30
you build stuff up, right? You start with
31:32
zero and you build like a city or something. Exactly,
31:35
you kind of have this like God-like view of
31:37
a huge area of land and like, maybe we're
31:39
growing crops to start a town, which
31:41
leads to chopping down trees to build buildings, which
31:43
could lead to like building factories or something like that.
31:46
And eventually you have this kind of like satisfying machine
31:48
loop kind of running. I wish I were that
31:50
good at it. Yeah, but yes, I know what you
31:53
mean, yes. So Terra-Nil is
31:55
kind of the reverse of that. So you start with a barren wasteland
31:57
that you assume has kind of been like ravaged by climate change.
32:00
and it's just like dirt and some rocks and polluted
32:02
stuff. But your whole goal is to restore
32:05
it to a thriving natural ecosystem. And
32:07
then the other part of your goal is just to
32:09
straight up leave when you're done. Remove any
32:12
trace that you were ever there, no human presence. Wow,
32:15
it's like camping in the forest, right? Yeah,
32:17
leave it better than you found it. Yeah, exactly.
32:20
So you have the game pulled up on your end, right?
32:22
I have it up. Okay, so I'm gonna press the
32:24
button, says new restoration. It's
32:28
just a barren land with a tree
32:30
and one leaf that just blew
32:32
off. Wow, that's very sad
32:34
looking, like the face of Mars here.
32:36
Yeah, that's not a good sign.
32:38
Oh, it's now starting. It's providing
32:40
me electricity. Tap to select
32:42
the turbine. And it's blowing. I
32:44
see the wind is turning the blades. Yay. A
32:48
toxin scrubber. It says it's cleaning
32:50
the soil. Wow, I see there are green
32:53
patches there now. Here we go. So
32:55
that's kind of the vibe. I love it. Yeah. So
32:57
I talked to the lead designer of Terrenel.
32:59
His name is Sam Alfred to just talk about
33:01
how the game came together, what sort
33:04
of scientific research they had to do for it. And
33:06
I started by asking him how the idea
33:08
for it came about. So we wanted
33:10
to try and make something that was inspired
33:13
by the genre of building games,
33:15
but instead of building a city, we're
33:18
building nature instead. So starting
33:20
with this barren wasteland and you're bringing life
33:22
back to it. That's where it all began. Yeah,
33:25
so it's working within the city builder genre.
33:28
Can
33:28
you describe how those games
33:30
have traditionally been designed and played and
33:32
maybe what was like frustrating to you about those games?
33:35
So typically everything, particularly
33:37
the natural world in a building game is
33:39
important for what value
33:41
can be extracted from it. Like forests
33:44
are valuable insofar as you can
33:46
chop down the trees, so you get the lumber,
33:48
so you can build industry or rocks
33:50
are valuable, so you can build castles or
33:53
is valuable so you can run your factory.
33:56
And Terinul makes the argument,
33:59
not overtly, but through its design
34:02
that nature and the natural world
34:04
is intrinsically valuable, that it
34:07
has its own value rather
34:09
than value from what you
34:11
can gain from it. So there's not a lot of that feeling
34:14
in the builder genre. It's very much
34:16
a kind of numbers go up style
34:19
of game, where you start with
34:21
a village and you want to turn it into a town, and
34:24
then you want to turn it into a city, and
34:26
in order to do that, the environment is
34:28
just a tool. It's just a tool. And we wanted
34:30
to instead try and make a game where the environment
34:32
was not the tool, the environment was at the forefront
34:35
of the player's mind when they were playing the game.
34:37
In the moment of my gameplay, you are constructing
34:40
wind turbines or toxin scrubbers or irrigators,
34:43
water pumps. And it's like this very relaxed,
34:47
slow paced, meditative and restorative.
34:50
Yeah. What kind of scientific research
34:52
did you and the team do for
34:54
the game? So our studio
34:57
is based in Cape Town in South Africa.
35:00
And I've always had a deep love of
35:02
the environment and of nature. I spend my childhood
35:04
hiking all over the country. And in
35:06
Cape Town, there is this incredible
35:09
biome of wildflowers called
35:11
fainbos that only grows in the
35:13
Western Cape around Cape Town. And
35:16
it also has this incredible property that
35:18
require fire to germinate. They
35:20
need to be burned in order to release
35:23
their seeds. And what this does
35:25
is it means
35:25
that all the grass and the scrub in the bushes
35:28
that might have been competing for resources
35:30
are no longer there. And the seeds can grow
35:32
in the nutritious apps that's left behind after
35:35
the fire. So the first major
35:38
update we did was taking this inspiration
35:40
from Fain Boss and putting this controlled
35:43
burn fire mechanic into the game.
35:46
That was the first real
35:49
idea that, hey, maybe we can look
35:51
at the natural world and find some
35:53
incredible examples of ideas
35:55
that people don't usually put in video games.
35:58
Yeah, actually when I first used the controlled
36:00
burn mechanic. I just love that controlled
36:02
burn as a mechanic exists at all in
36:05
a game, but I ended up
36:07
almost burning my entire pot
36:10
of land. That happens more often than you
36:12
think. Yeah, and
36:14
I definitely had to hit the undo button after
36:16
that. I was like, oh my god, no. But when
36:18
you and the team were trying to think of different methods of restoring
36:21
the land, were you ever like, oh, we need some
36:23
way to do X to the land and
36:25
And actually this real world method would
36:27
be like a perfect verb for the player
36:30
to take. Yeah, there are a couple
36:32
of other really good examples from the game.
36:34
We looked to five broad categories
36:37
and I think 16 subcategories
36:39
of biome classification on Earth. And
36:41
we thought, wouldn't it be really cool if we could
36:43
make one level for each of the overarching
36:46
classifications? And so the
36:48
approach to the game's development became, okay,
36:51
we're going to make our tropical region now. Let's
36:53
do some research about what kind of
36:55
plants
36:55
grow there, what kind of biomes
36:58
could you find. And so one of the elements
37:00
you have to restore in the tropical area is coral reefs.
37:02
Coral reef destruction and coral
37:04
bleaching is a real big problem in
37:07
our oceans today with sea temperatures
37:09
rising and shipping, things like
37:11
that. Coral reefs are getting destroyed all
37:13
over the place. So we did
37:15
some research into how coral
37:17
restoration projects work in the real world and
37:20
found some really interesting stuff there. So
37:23
coral restoration projects often happen
37:25
on land first in coral nurseries
37:28
where there are these pools of
37:30
ocean water that have steel
37:33
frames in them and coral
37:35
apparently grows really well on steel.
37:38
In the case of a coral nursery, the steel
37:40
frame is used as like a skeleton
37:42
for a new coral reef and existing coral
37:45
is taken and then through a process
37:47
called micro-fracturing is spread
37:49
out over the frame and adhered to
37:52
the frame all over the place and that increased
37:54
surface area of the the coral means
37:56
that the coral growth is sped up significantly.
37:59
And then once that.
38:00
coral is mature, it can be dropped
38:02
into the ocean just as is.
38:05
And over time, the coral will come to completely
38:07
overgrow the frame and just use it as
38:09
a backbone. And natural coral reefs do
38:11
this too with rock. But in this way,
38:13
we can sort of speed up the process of coral
38:16
reef growth. So in Terra Mill, we have
38:18
a coral laboratory that you construct
38:20
on the land in the tropical region. And
38:23
then you use a monorail network
38:25
that's an important part of the
38:27
challenge of that particular level to
38:30
move that coral core
38:32
into the ocean to then grow a reef.
38:35
Yeah, you know, before I started
38:38
playing this game, you know, I was kind of like, okay,
38:40
it's gonna be about climate change in some ways. And
38:43
I think I was kind of subconsciously
38:46
preparing myself for it to be like kind of heavy,
38:49
but instead the overwhelming emotion I got
38:51
when I was playing it was like, oh, wow, like all
38:54
of this restoring our ecosystems, focusing
38:56
on balance, actively making the
38:58
environment better around you. It
39:00
didn't only feel really good, like in the short
39:03
term kind of like, rewarding dopamine
39:05
hit that strategy builder games are like
39:07
really good at, but also big picture, like feeling
39:09
optimistic about action we can positively
39:12
take about our future with climate change. Like
39:14
there is this massive task ahead of us, but
39:16
this game really gave me a firm sense of like, oh
39:19
yeah, like we can do this, it's gonna be a lot of work,
39:21
but we can change our perspective about
39:23
the work that'll be involved with combating the climate
39:25
crisis And how much of that
39:27
did you and the team have in mind while you were
39:29
making the game? Well, I mean, it's
39:33
really nice to hear that because this idea
39:35
of climate positivity to combat
39:37
climate apathy is very core
39:40
to what we were trying to do with the game. We're very
39:42
much of the belief that the things we
39:44
consume as a culture, the media,
39:46
but also the cultural preconceptions
39:49
and cultural beliefs are kind of like a self-fulfilling
39:52
prophecy where if Every game
39:54
ever made is just about dystopian
39:57
futures and cyberpunk cities. then
39:59
it
39:59
feels
40:00
almost inevitable that we'll end up
40:02
there. At one point in the development of the
40:04
game, we did have many discussions
40:06
about whether or not we should include what
40:08
happened to cause us to get to
40:10
this point. But in the end, we
40:12
decided to include none of that because it's not a game
40:15
about browbeating the player with these
40:17
all the things that went wrong. It's a game about imagining
40:19
a better future. So yeah, that's
40:21
really nice to hear. That's exactly what we're going for. You
40:25
talked before about like, you and the team
40:27
didn't want to make Terranil be like infinitely
40:30
replayable, which is something a lot of other developers
40:32
try to do with their games. Can you talk
40:34
about the impulse behind that? Yeah, so
40:37
quite early on in the game's design, we realized
40:39
that if this was going to be a game about balance
40:42
of ecosystems in a level of Terranil,
40:44
you have to have wetlands and forests and
40:46
fainbos all in the same landscape.
40:49
Creating too much of one means that
40:51
you can't use that space for another. And
40:53
so it's very much part of the game's identity
40:56
to be about balance. Being finite
40:59
is kind of a natural extension
41:01
of that idea. If we
41:04
just gave you the option to just, I
41:06
don't know, buy more land, because
41:08
you grew too many trees, now you
41:10
need some wetlands, it would kind of undercut
41:13
the philosophical idea we're going
41:15
for that more important than
41:18
growth is balance. It's a game where
41:20
at most the level will take you two
41:22
hours or something like that, as opposed
41:24
to being able to build the same
41:27
city for 100 hours. And the fact
41:29
that you can't just keep playing it forever, that
41:31
balance is the ultimate goal. Infinite
41:34
growth is not the goal. And this is, again,
41:36
about the environment, but it's also a little bit
41:38
of a statement about society and
41:41
some of the problems inherent in the way
41:43
we view economic growth and the value
41:45
we put on your quarterly returns. They've got
41:47
to be bigger this quarter than they were last quarter.
41:50
Sustainability doesn't come into it. Yeah. What
41:53
role do you think games serve
41:55
in the context of climate change and climate
41:57
action?
41:58
I think games can be this
42:01
incredible window into what
42:03
is possible.
42:04
Games are things people do for
42:06
fun, generally speaking. They're a recreational
42:09
activity, and therefore, they're a really great
42:11
vehicle for getting people to think about
42:13
things differently because they're relaxing.
42:16
They're not being forced to engage
42:18
with something. They're choosing to engage with it, right?
42:21
So you can use your video game to
42:23
let people know that there
42:25
are things that exist that maybe they
42:27
hadn't heard of before and makes themselves,
42:30
huh,
42:31
I'm going to look this up. I'm going to find out more about
42:33
it, and maybe even go a step further
42:35
and be inspired to do something else in the
42:37
real world. And I think games have a unique
42:39
opportunity in that respect, because unlike
42:42
a film, when you're playing a game, you're
42:44
playing the game. You're
42:46
not just receiving it. You are taking
42:48
the actions. And so if we
42:50
as game developers can help players take
42:52
interesting actions, let them
42:55
think and learn in different ways. It's
42:57
a hugely powerful medium. Mm-hmm. I
42:59
think that's a great place to end it. Thank you for taking
43:02
the time, Sam, and thanks for the game. Cool.
43:04
Thanks so much for having me.
43:07
Sam Alfred is the lead designer of Terranil,
43:10
developed by the Studio FreeLives. Terranil
43:12
is playable through Steam on PC and
43:15
Netflix games on iOS and Android. And
43:17
if you want to check out the trailer for the game,
43:20
yes, head over to our website, sciencefriday.com
43:24
Next,
43:30
an exercise pool for tiny
43:32
worms. Yes, all in the name
43:34
of science. Here's the backstory. A
43:37
team of researchers at CU Boulder
43:39
are trying to help treat people with Parkinson's
43:42
and other neurodegenerative diseases.
43:44
That is no small feat. And
43:47
to do that, they're turning to trusted collaborators,
43:50
C. elegans. These are tiny worms
43:52
just one millimeter long and they
43:54
are often used to study human health. You've
43:57
heard us talk about them before. question
44:00
is, does exercise affect
44:02
brain health? The method is
44:04
putting a bunch of these worms in an exercise
44:07
class. And I'm not talking about weights,
44:10
but in the water in a tiny pool
44:12
nicknamed Jim on a Chip.
44:15
Here to coach us through the new study is
44:17
Dr. Jowita Badra, researcher
44:20
at the University of Colorado Boulder.
44:22
Welcome to Science Friday. Thanks
44:24
for having me. That
44:26
was nice to have you. Okay, before
44:29
we take a swim in the worm gym, what
44:31
can C. elegans tell us about the human
44:34
brain? What do you hope to see here?
44:36
So, C. elegans is a simple organism.
44:39
However, one third of its body
44:41
cells are neurons. So we see about 300
44:44
neurons in C. elegans.
44:47
And it's very easy to visualize under
44:49
microscope. Another reason we use it is
44:51
because it's very fast. it actually reaches
44:54
its adulthood within three days.
44:56
So it gives us an excellent tool to
44:59
study any age-related diseases
45:02
because you know you don't have to wait for
45:05
months to look at them.
45:06
So how does a gym on a chip
45:08
work? What's the concept here? Okay,
45:11
so you can think of these as small
45:14
swimming pools for the worm. We
45:16
use electrodes to generate
45:19
wave on the surface of the
45:21
chamber, which creates
45:24
this whirlpool. So the worms
45:27
needs to constantly fight
45:29
against this whirlpool. In this way,
45:31
you can actually control the intensity
45:34
of exercise.
45:34
Wow. You know, years ago, we talked
45:36
about a shrimp on a treadmill, but never a worm
45:39
in a pool. And
45:41
so as they're swimming, as they're exercising
45:44
against the tide, so to speak, against the
45:46
whirlpool, what are you looking for?
45:48
The worms that we used for this
45:51
particular study are
45:53
a good model for Parkinson's disease. So
45:55
these worms actually have human
45:58
alpha-synuclein. expression. So human
46:01
alpha-cyanuclein forms these aggregates
46:03
that are called Lewy bodies, and
46:06
that can cause neuronal loss
46:09
in humans. And we can see the
46:11
same phenotype in worms
46:13
as well. Well,
46:14
can you tell if the exercise
46:16
does prevent the, you know, the
46:19
neurodegeneration?
46:20
Yes, actually it does. So with
46:23
only five minutes of exercise for
46:26
two days, we have seen a significant
46:29
loss of neurons, specifically
46:32
the dopaminergic neurons in these worms.
46:34
We know this in humans, that exercise
46:36
is good for us, right? You're discovering
46:39
that exercise is good for neurotransmitters
46:41
and good for keeping your brain healthy.
46:44
At least in the worms, can we apply this to people?
46:47
The reason for our study is this
46:49
video that came up in, I think, CNN on some
46:52
other news channel from Cleveland Clinic,
46:54
Dr. Alberts did some experiments.
46:57
So one of his patients is a Parkinson's
46:59
patient, and he went for
47:01
tandem bike riding with
47:04
him. And after that, the
47:06
tremors in his hands almost
47:08
disappeared. So this kind of told
47:11
us, like, okay, so exercise
47:13
intensity is important because tandem
47:16
biking is an assisted exercise. Our
47:19
main goal was to find
47:21
out what is the optimum intensity
47:24
and duration of exercise. to be
47:26
found is that there
47:28
is a sweet spot.
47:30
Too little exercise does not help. And
47:32
too much exercise also does not
47:35
help.
47:35
It's just this optimum intensity
47:38
that causes the maximum benefits
47:41
in response to neuronal loss
47:43
that be observed in these worms.
47:45
Wow, can you translate that to people
47:48
where that sweet spot is? Now,
47:50
it's hard to translate that in
47:53
people. And that's okay.
47:55
Parkinson's disease is an age-dependent
47:58
disease. mostly it affects people. who
48:00
are above 60 years old. And
48:03
it causes these tremors, and then
48:05
there are a problem in movements
48:07
which makes it more difficult to
48:10
exercise. Our goal
48:12
is actually to use this device for
48:14
drug screening. We are looking
48:17
at the markers that gets
48:19
changed due to this perfect
48:22
optimum intensity of exercise. Once
48:24
we get those, we are gonna do a drug screen
48:26
to look, identify the targets.
48:29
and see if these drugs
48:32
can maybe completely replace
48:34
exercise altogether or maybe
48:37
enhance the benefits
48:39
of exercise. Hope to have you back
48:41
when you've gotten more results. I
48:43
hope so too. Thank you very much. You're
48:46
welcome. Dr. Jowita Bhadra is a researcher
48:48
at the University of Colorado in Boulder.
48:51
We'll see you next week.
48:52
I'm Ira Flatow.
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More