2001: A Space Odyssey: From Science Fiction to Science Fact


>>Welcome to the tech takeover. Is everone having a good time? I’m joined here with Professor Rada Mihalcea and Professor Ben Kuipers and we’re gonna talk a little bit about how the
technology you see in 2001: A Space Odyssey relates to kind of the work that we’re
doing in Michigan engineering. We’re gonna focus on
spacecraft propulsion, robotics and artificial intelligence. So my name is
Alec Gallimore. I’m a professor of aerospace engineering and the dean of
the College of Engineering and I’m gonna focus my talk on advanced spacecraft
propulsion systems. So as a starting point what I’m showing in this image is
a picture of the Discovery One spacecraft which is a spacecraft in the
movie, 2001, that sends the astronauts to Jupiter. This spacecraft is
nuclear-powered, about 700 feet long, and can travel at over a hundred thousand
miles an hour. So to give you context to that speed it covers the distance
between the earth and the moon in about two hours or so. So the question is how
close are we to developing that kind of spacecraft propulsion technology and in
fact you’re seeing an example of that in that display back there. The same plasma
drive that is used on this spacecraft in the movies and in the book is
being demonstrated in that chamber over there behind you. But let me start with
this question does science fiction drive technology or just technology drive
science fiction? So let me give you some examples. I don’t know what do you think? So this is to the left is from 2001: A
Space Odyssey. What does that look like to the left? iPad. 1968, an iPad. Wow not bad. Looks familiar, alright, how about
that. So let me go back, does science technology drive science fiction or does
science fiction drive technology? My answer: yes, it goes both ways in my
humble opinion. So what I want to focus on is why do we have a need for speed? We
just launched with the Solar probe the Parker Solar Probe – the fastest human-made object. When that probe passes by the Sun it’s going to be traveling at over
400,000 miles an hour. Earth to the moon in about 30 minutes, a
little bit over 30 minutes. So what’s the need for speed? One is to get around space it’s all about trajectories and
what is a trajectory? Well space travel also is all about getting on and off the
right conveyor belt. In this case a conveyor belt around the Sun, around the
earth, around the moon at the right time, and that’s what we call a trajectory. And
what it is: it’s a balance between gravity from the Sun and from the earth
and movement so let me give you a concrete example imagine you have a
student here and she has a string with a mass at the end and she twirls it around
and clearly what’s going on is that the Rope is pulling the mass so it doesn’t
fling off and her speed is keeping the the rope rope taut you’re saying well
how does that apply to the space travel well actually that flies a lot because
imagine now in space travel she would be the Sun the string would actually be
gravity the distance though is 93 million miles so every time you look at
the Sun think that Sun is 93 million miles and the earth is moving around the
Sun to balance his gravitational pull at 67,000 miles per hour a jet airliner
moves at five or six hundred miles an hour so things in space tend to move
very very quickly so how do we get things moving fast in space propulsion
and the one that’s the most common is what’s called rocket propulsion and how
it works is that we take a fuel like hydrogen or kerosene and we mix it
with an oxidizer like oxygen and then we have it burn in a chamber call a rocket
chamber and expands to a nozzle about ten thousand miles an hour and that gas
moving at ten thousand miles an hour provides a force that accelerates the
spacecraft so how fast can rocket propelled objects travel well any time
you send an object in orbit like we used to do in the Space Shuttle you have to
be at almost 18,000 miles an hour remember a jet airliner travels that
around 600 miles an hour 18,000 miles an hour but wait there’s more if you want
to go to deep space like the moon or beyond you have to go at 25,000 miles an
hour because of that speed you break the bonds of Earth’s gravity and you can be
in deep space the fastest chemically propelled system that we have so far is
the Pluto probe that was launched a few years ago called new horizons and that
rocket sent a large rocket about the size of the Lurie Tower or even bell
tower here sent a spacecraft the size of a piano up to is to be about 43,000
miles an hour so that’s pretty good the question is why do we need to go fast
and partly because space turns out to be really big so let me give you do an
exercise about how big space is in terms of interplanetary space imagine we have
a scale where the distance between the earth and the moon is that is across
your finger right so earth is over here the moon is over here on that scale how
far are some heavenly bodies from us alright so for example earth to the moon
is with a my finger Mars would be about eight feet Jupiter
65 feet Saturn about 110 feet Neptune longer than a football field
interesting enough the nearest star besides the Sun how far is that on that
scale earth the moon is a width of my finger well the nearest star called
Alpha Centauri other than our Sun is 600 miles and that is 4.2 light-years away
the galaxy we live in is a hundred thousand light-years wide and the
nearest gallic see Andromeda is 2 million light years
and we expect the universe to be about 13 billion light years wide so it’s just
mind-boggling sighs let’s do it in time now
so we take that rocket that we launched to Pluto at forty three thousand miles
an hour which is only about six hours to the moon or so how long does it take if
he was straight line of that speed to Mars at the closest approach sixty days
would take over a year to get to Jupiter that speed two years to Saturn several
years seven years to Neptune six hundred centuries to the nearest star other than
our Sun so one question is let’s talk about sending people to Mars nASA has
plans to do that in the 2030s or so and if we did it with a chemical system we
can get there but it would take nine months and believe or not one of the
challenges with having the trip take nine months is this safety and the
healthy astronauts once they got there there are some studies that show because
of radiation that comes from the Sun that the astronauts would have on
average a life time 15 years less than if they had not gone because of
radiation dosage because a high probability of getting cancer 15 years
and so we want to do is see if there’s a way of getting them there quicker and so
that speaks to this concept of electric propulsion instead of burning things
like a fuel and oxidizer we take the power that’s onboard the spacecraft and
we heat a gas to tremendous temperatures in a chemical system there might be five
thousand degrees in these plasma drives similar to ones you see over there they
can be up to five hundred thousand degrees or so and you need that energy
to move these particles at tremendous speeds so to my right the image to the
right is a picture of one of the plasma drives we developed with NASA as a
prototype a few years ago so this is what spacecraft that will say send
humans to Mars in the 30s will look like if they use electric propulsion to the
left is a solar electrically propelled system it uses solar rays to power the
plasma drives and it may have a chemical rocket for kicks when it needs it and to
the to the student to the left and to the right is a nuclear propelled
spacecraft and the reason we’re interested in this electric repulsion is
that weak cut the trip times down to six or maybe
even three months so you think about that chamber over there and it turns out
we have a much larger one in my lab my lab is called the plasma dynamics and
electrical propulsion laboratory and our biggest chamber is twenty by thirty feet
long and what we do is we do work for NASA primarily in government agencies
like the Air Force to prototype and design work with them to develop
advanced plasma dyes including a prototype system that may one day seeing
people to Mars it’s very interesting we lose a lot of robots too you do this
work some of the robot systems you’ve seen here and what we do is we’re able
to provide a pressure that’s about 10 billion times lower than the pressure
that you have right here at sea level in the chamber when we’re operating these
thrusters and just to give you an idea the scale of it this is how big it is
it’s actually one of the 10 most capable vacuum chambers on the planet in terms
of size and its ability to maintain a low pressure so it’s pretty amazing and
we needed a chamber that big because the thruster that we’re developing is also
quite large so this is a picture of a the first prototype Mars engine the
first prototype propulsion system that may send humans to Mars and cut the trip
times from nine months down to about four to five months or so to the left
you see it next to one of my graduate students who’s now an engineer at NASA
dr. Scott Hall and to the right is the plasma drive operating in our vacuum
chamber and we received a lot of notoriety for that because we ended up
breaking a number of World Records in terms of power and thrust produced by a
plasma drive so the question though is is this the system that’s used on the
discovery one spacecraft in the movie and actually know what they were gonna
do and again this was made in the 60s so they were even more flamboyant in the
60s in terms of their thoughts was they were gonna use a nuclear propulsion
system where they’re gonna take gas uranium that’s highly radioactive and
mix it with a propellant and have it expand through a nozzle very elaborate
but it requires a tremendous amount of power to do that realistically speaking
if we move fast forward 50 years in order to achieve this kind of capability
we could do it probably in the middle of this century or so but we probably
wouldn’t use is kind of gas-phase uranium type of
drive we would absolutely use one of the plasma drives like the kinds that I
showed you before that we’re developing so with that thank you for this segment
we’re gonna now pivot to the other elements we’re gonna talk about
artificial intelligence and just to tee it up again like I showed you images
from the movie the iPads and things like that in the Space Station’s and the
space shuttles just wanted to show you a couple of images so for example the top
left is how from the movie you all recognize on the bottom left what’s
going on the Jeopardy the Jeopardy competition that was actually won by an
artificial agent as well as as well as the chip chest championship
being being handled by a robotic system so with that I’m going to pivot over and
to invite my colleagues Radha to give the first presentation thank you good evening I am very pleased to have
the opportunity to share with you how I think that the past present and future
of artificial intelligence realized to the movie that you are about to enjoy
and I direct the artificial intelligence laboratory here at Michigan this is a
group of amazing faculty nearly 20 faculty working on the core areas in AI
covering aspects such as learning vision cognition perception language robotics
and so forth and I would like to start by going back in time and I wouldn’t
even stop to the movie in 1968 but rather go even further back to the 40s
and 50s when Europe was torn by the Second World War
and when a brilliant mathematician by the name of Alan Turing he’ll develop an
algorithm that was helping people to break codes he is probably best known
for breaking the Enigma code but another thing that is well known for her is this
paper which he published in 1950 in which he addresses the question can
machines sink and what he suggested is that we could test whether machines can
sink through again he called it the imitation game and let’s assume you’re a
participant in the game what would happen you will be invited to be a judge
you would come into a room and then there will be a number of consoles and
what you have to do is to see that each of these consoles in turn and simply
have a conversation the catch here however is that you don’t know who’s on
the other side so you have this conversation at the first console and
after that you have to determine whether on the other side was a human or a
machine and doing this repeatedly eventually
there might be a machine that say if you do confuse it and think that there was a
human then we would say that that particular machine is actually sinking
and so that would have passed this imitation game the way to input it now
over time that game changed name and we now refer to it as the Turing test and
it’s one of the tests that we sort of live by in artificial intelligence is
still one of our main goals so the movie came later in 1968 and I would like to
play just a fragment from this movie to see how a conversation can happen
between a human and a machine so as you can see you could possibly
without help me having told you that that was a machine you could have
possibly confused and machine for a human it’s actually sounding very
naturally including inflections the content and so forth I would like to
also make a speculation like I said during paper came out in 1950 and what
he also said in that paper was that we would pass this imitation game or as we
will say today the Turing test would have been passed by the end of the 20th
century the movie is called 2001 so maybe the director was actually informed
by this projection that Turing made by calling 2001 and maybe wasn’t
necessarily intended as science fiction but rather as a vision as to where we
would be in AI now let’s look also what was actually the state of affairs at
that time so in 1960s there was a dialogue system so there were systems
that people build however they were nothing like what you just saw that they
were in fact rule-based and using canned responses and here is an example from
Eliza the therapist it was billed in the 60s but Eliza is still alive and well
you can actually test it out so if you go on the internet and just look for
Eliza you can have a conversation and this is the conversation that I had with
Eliza just a few days ago and it’s fully unedited so here is how it went
Eliza hello I am Eliza me hi Eliza how are you today
what would you like to discuss me I’m giving a presentation on Friday Eliza do
you believe it is normal to be giving a presentation on Friday me it is about
the past present and future of AI Eliza come come elucidate your thoughts me well I was hoping you could help me
here should I talk about you Eliza you are not really talking about me are you
me you tell me should I you are a good example of a rule-based dialogue system
Eliza do you sometimes wish you were a good example of a rule a dialogue system so you can sort of see how this was
going how you could even predict what is going to say next and it’s all fully
rule-based now fortunately there is a very interesting phenomenon that is
happening when it comes to progress progress is not linear progress in fact
has this exponential shape and one of the biggest proponents behind this
acceleration of progress is Ray Kurzweil and really the effect of these that what
took a hundred years in the past to happen is going to take ten years next
and one year after and so forth and so with that we’ve been really making
tremendous progress in the 60s and I want to highlight just a few of the
things where we made progress one of them is hardware this is the very first
computer it was an even color computer it was called the ENIAC that was the
proposed word for this thing eventually I didn’t cut up we do proudly still have
a piece of DNA at Michigan if you go in the College of Engineering the computer
science and engineering building has one piece of this computer now this you see
it was a full room now instead we have the desktops of so these are very tiny
much more powerful than that one full room was before we also have the data
centers which somehow ironically they look very much like those computers at
46 you still see the wires and all that still the room full of equipment however
in terms of what they can store what they can do there is not even room for
comparison they are tremendously powerful another thing that happened
is data if we look back in 2000 and an estimate of how much data there was on
the Internet we were talking about terabytes
so terabytes would be thousands of gigabytes then later we started talking
about petabytes and that would be millions of gigabytes then we had
exabytes which would be billions of gigabytes and you see how we are even
adjusting our language because we don’t have enough words to talk about how much
data that is there and who know how we call what’s out there right now we have
the data bytes so data bytes is how much data there is there and this really
means there is a lot of data that algorithms can learn from so we can
build all these amazing tools another thing that happened is like I said
algorithms so in the 40s and 60s 1450 60s we had surprisingly neural networks
so neural networks which are so popular today were actually introduced many
years ago they didn’t eventually catch up because we didn’t have enough
hardware or the right hardware or the right data then later we had so they’re
starting year 2000 various machine learning algorithms decision tree
support vector machines random forests and so forth and then for the past few
years we started having a lot of algorithms that are relying again or
neural networks or deep neural networks but this time with the support of the
right hardware and the right amount of of data now guess what was in between
there was winter so that was the so-called
AI winter there are still progress happening but the trust of society was
maybe a little less than what is there right now and I’ll mention one more
thing that I think changed significantly since the 60s and that’s a very reason
so I cannot even put it on the curve it’s crowdsourcing the fact that people
like you and I are actually contributing to the algorithm contributing to various
bits of data if I know for instance how to translate into French I would
contribute one French translation you would contribute one and then two
we create enough data that we can feed a machine translation system so what was
the result of all these a lot we have human level performance speech
processing so we get tools that can go from speech to text in a way that it
almost looked like a professional transcribers we have ways to process to
text and understand what’s being said whether it’s about a person or a
location whether there is an action and so forth we have question answering
systems that are able to answer the question that we asked like Siri or
Alexa and we have machine translation that can translate between many
languages of the world and so many other things now remember the Eliza
system that was asking me whether I wish I were a rule-based system myself let’s
see where we are in terms of dialogue today and this is from a demo this year
the duplex system that was introduced by AI researchers at Google this is a system that is targeted toward
a certain task which is that of making appointments but as you can see it
sounds pretty natural so the question going back to where I
started have we passed the Turing test yet I would say not quite and we have
made impressive achievements and there was a lot that happened for the past 50
years there is still a lot to do there are still a lot of languages to
translate there is still a lot to develop in terms of doing question
answering in terms of building systems that can have natural conversation and
so forth and this is just talking about how we process language and there are so
many other areas so I will leave you with my favorite curve shape progress
it’s accelerating and so the future is promising even more exciting findings
thank you so thank you very much for inviting me
here I wanted to talk about AI and 2001 one of the things 2001 came out in 1968
as we’ve already seen and it was just an amazing experience for those of us who
saw it at the time it was the major art of the movie as you’ll see is the
transformation of humanity from prehistory to post history and that was
just a fantastic thing to look at in the psychedelic 60s so a few years later I
started graduate school in pure math and I ran across the MIT artificial
intelligence lab and that grabbed my attention took me away from pure math
and I decided that this would be the focus of my life’s work and I ended up
spending my time on the problem of common sense knowledge which comes up in
a lot of different ways space objects actions and so forth and I’ll talk a
little more about that in just a moment one day at the artificial intelligence
lab with no particular fanfare who should show up but arthur c clarke who
was one of the world’s great science-fiction writers and the author
of the novel 2001 a space odyssey my adviser marvin minsky was a great fan of
science fiction and he had been a consultant on this movie he had helped
giving advice on the nature of artificial intelligence and the design
for how who is the artificial intelligence system that we’ll see in
the movie now how ended up being a huge inspiration for many people who went on
into artificial intelligence and we see how carrying on there
sophisticated conversations with people at a level that cannot yet be matched
really howls conversations are beyond the current state of the art in terms of
coherence and just substantive content even 50 years later now he’ll interacts
with people through its camera and through its voice but actually the whole
spacecraft is essentially Howells body its nervous system and its activities
are taking place all through the spacecraft as we’ll see now I’m sure
it’s not a spoiler to tell you that in the movie Hal ends up being a bad guy so
he’s a villain and but what I want to do is I want to draw your attention to a
particular theme that comes up as we go through the movie in the part of this
arc that that deals with the space voyage so during the voyage we’re going
to start out and we’re gonna see how with what I hope okay so what he’ll and
one of the crew members is Frank is playing chess so this was fun and you can go out and
buy something that would that would let you have that much fun easily but in the
movie one of the thing a much more serious game of chess is going to take
place and I just want to draw your attention to it so here we’re going to
get the opening move in that game after some things happen we get the
endgame here but this is not the end of the game this is the beginning of the
endgame so if artificially intelligent systems
start doing this this could be a considerable concern however I don’t
think that we’re in immediate danger of this we have made dramatic progress over
the last 50 years and the AIS that we’re building are really idiot savants that
is that they’re extremely good at some specialized thing but they don’t have
anything like the kind of general intelligence that we’re seeing in how
there is a great deal of promise but the risks are things that are manageable at
the moment rather than catastrophic the way it is as portrayed in here now
nonetheless the kind of progress that we’ve seen has it has encouraged a
number of researchers in AI including me to think that there’s another kind of
common sense knowledge that we need to look at which is ethics and ethics
includes knowledge of what is appropriate to do it
ethics involves knowledge and skill AI researchers can study how knowledge like
that is acquired its how its represented and how it’s used and we would invent we
investigate how how robots like people will evaluate the things that they’re
considering doing so that they can treat people and treat everyone around them
appropriately instead of treating them like chess pieces thank you thank you we can do Q&A we have a couple
of people running the mics if you’re interested in asking a question I asked
you to stand up and wait for the microphone and then far away please
Martha Mars rover the the let the that spacecraft just landed on Mars and
they’re there hiding in this sandstorm right now how much longer and how more
things do we have like that that are in process oh okay so there are a number of
Rovers being prepared there’s a rover called Mars 2020 that’s being launched
and one of the goals of that Rover is actually to collect samples and hold the
samples for a follow-on mission called a Mars sample return mission we shall have
a robotic system that will bring those rocks back to earth and that’s a
precursor for starting to send people to Mars in the early 2030s or so yes if you were to design a mission like in
2001 but try to mitigate the risk that you said came from having such a general
intelligence would one solution be to have just a whole collection of the
idiots of idiot savant a eyes that can’t talk to each other probably not having
them that cannot talk to each other but I think we need to have a notion that an
individual agent an individual artificially intelligent agent needs to
know that it has a responsibility to the society
now in the case of a spaceship like that it would be the society which is the
crew here he’ll believe that it had a responsibility towards the mission
towards accomplishing this mission and it was prepared to sacrifice make many
sacrifices in order to accomplish the mission
whereas we expect people to relate well to their society and to the other people
in them and to treat them well oh okay okay one question so I know
there’s a lot of excitement over deep learning neural networks and machine
learning systems and they’re used pretty much everywhere in all of our devices in
our software but I was wondering in terms of developing true AI is that do
you think this is the right approach how were you thinking too much into one
method to develop these systems are there other ways that we can get to
developing AI I would say recent findings and it there is a lot of
excitement around deep learning and so there is good reason for trying to apply
that approach to a number of problems I don’t think that solves it all I do
think that eventually there will be other methods that will have to come
into play and even now there are people who realize that deep learning is not
the solution for everything and so we are looking at alternative approaches as
well for instance when you have a problem that does not have a lot of data
neural networks are limited there so we have to think of other ways to tackle
that one of the things that is an important thread of artificial
intelligence research is asking how does the knowledge in a particular domain get
represented how is it structured how is it acquired how is it used and the the
methods of deep learning are very powerful interesting methods but they
basically jump around all of those questions and they say let’s take pixels
in and we’ll get raw bits out and then we’ll grade those and we’ll train the
things in between but it never really asks those questions about how the in
the intermediate stages need to be structured and almost certainly we will
need to develop those intermediate stages kind of a question for both of you
are there any milestones or applications for AI that you’re particularly excited
for us to kind of achieve or CAI in a particular field or anything like that
well I would argue and of course it’s biased by my own expertise and interest
in favor of natural language understanding so actually being able to
determine what’s being said not in the shallow way of just being able to
classify a document for instance but in the deeper way of figuring out for
instance what are the relations between the people that are mentioned in that
document and what is their emotion or what are their actions what are their
planning to do what our intention and so forth so I think that is one major goal
within AI in particularly within my field of natural language processing and
and I would approach that by saying one of the things we know about human
knowledge and human intelligence is that it starts out at a very primitive level
and children learn it through a huge variety of states of knowledge and
different ways of conceptualizing the world and it may very well be that
looking carefully at that will give us more insights into the structure of the
knowledge and will tell us stages that that we will have to go through in order
to create artificially intelligent creatures that actually can handle the
world in a sophisticated way as we could as we do on a political note how much of
this information is kept is shared among nations how much does the USA keep it
quiet this is ours I would say the stuff that
we work on is shared publicly that that the function of science is to share the
result and build on other people’s result
so science is a is a fundamentally idealistic process where we build on
build on each other’s methods and so I think a situation that confines this to
individual countries or individual corporations that want relative
advantage I think is fragmenting something that works best as a unified
international worldwide process well with that thank you for joining us thank
you

Leave a Reply

Your email address will not be published. Required fields are marked *