We built computers to expand our brains. Originally
scientists build computers to solve arithmetic, but they turned out to be incredibly useful
for many other things as well: running the entire internet, lifelike graphics, artificial
brains or simulating the Universe, but amazingly all of it boils down to just flipping zeros
and ones. Computers have become smaller and more powerful
at an incredible rate. There is more computing power in your cell phone then there was in
the entire world in the mid 60s. And the entire Apollo moon landing could have been run on
a couple of Nintendos. Computer science is the subject that studies
what computers can do. It is diverse and overlapping field but I’m going to split it into three
parts. The fundamental theory of computer science, computer engineering, and Applications. We’ll start with the father of theoretical
computer science: Alan Turing, who formalised the concept of a Turing machine which is a
simple description of a general purpose computer. People came up with other designs for computing
machines but they are all equivalent to a Turing machine which makes it the foundation
of computer science. A Turing machine contains several parts: An
infinitely long tape that is split into cells containing symbols. There is also a head that can read and write
symbols to the tape, a state register that stores the state of the head and a list of
possible instructions. In todays computers the tape is like the working
memory or RAM, the head is the central processing unit and the list of instructions is held
in the computer’s memory. Even though a Turing machine is a simple set
of rules, it’s incredibly powerful, and this is essentially what all computers do
nowadays. Although our computers obviously have a few more parts like permanent storage
and all the other components. Every problem that is computable by a Turing
machine is computable using Lambda calculus which is the basis of research in programming
languages. Computability Theory attempts to classify
what is and isn’t computable. There are some problems that due to their very nature,
can never be solved by a computer, a famous example is the halting problem where you try
to predict whether a program will stop running or carry on forever. There are programs where
this is impossible to answer by a computer or a human. Many problems are theoretically solvable but
in practice take too much memory or more steps than lifetime of the Universe to solve, and
computational complexity attempts to categorise these problems according to how they scale.
There are many different classes of complexity and many classes of problem that fall into
each type. There are a lot of real world problems that
fall into these impossible to solve categories, but fortunately computer scientists have a
bunch of sneaky tricks where you can fudge things and get pretty good answers but you’ll
never know if they are the best answer. An algorithm is a set of instructions independent
of the hardware or programming language designed to solve a particular problem. It is kind
of like a recipe of how to build a program and a lot of work is put into developing algorithms
to get the best out of computers. Different algorithms can get to the same final result,
like sorting a random set of numbers into order, but some algorithms are much more efficient
than others, this is studied in O(n) complexity. Information theory studies the properties
of information and how it can be measured, stored and communicated. One application of
this is how well you can compress data, making it take up less memory while preserving all
or most of the information but there are lots of other applications. Related to information
theory is coding theory. And Cryptography is obviously very important
for keeping information sent over the internet secret. There are many different encryption
schemes which scramble data and usually rely on some very complex mathematical problem
to keep the information locked up. These are the main branches of theoretical
computer science although there are many, more I don’t have time to go into Logic,
Graph Theory, Computational Geometry, Automata Theory, Quantum Computation, Parallel Programming,
Formal Methods and Datastructures, but lets move on to computer engineering. Designing computers is difficult because they
have to do so many different things. Designers need to try and make sure they are capable
of solving many different kinds of problem as optimally as possible. Every single task that is run on the computer
goes through the core of the computer: the CPU. When you are doing lots of different
things at the same time, the CPU needs to switch back and forth between these jobs to
make sure everything gets done in a reasonable time. This is controlled by a scheduler, which
chooses what to do when and tries to get through the tasks in the most efficient way, which
can be very difficult problem. Multiprocessing helps speed things up because
the CPU has several cores that can execute multiple jobs in parallel. But this makes
the job of the scheduler even more complex. Computer architecture is how a processor is
designed to perform tasks and different architectures are good at different things. CPUs are general
purpose, GPUs are optimised for graphics and FPGAs can be programmed to be very fast at
a very narrow range of task. On top of the raw hardware there are many
layers of software, written by programmers using many different programming languages.
A programming language is how humans tell a computer what to do and they vary greatly
depending on the job at hand from low level languages like assembly through to high level
languages like python or javascript for coding websites and apps. In general, the closer
a language is to the hardware, the more difficult it is for humans to use. At all stages of this hierarchy the code that
programmers write needs to be turned into raw CPU instructions and this is done by one
or several programs called compilers. Designing programming languages and compilers
is a big deal, because they are the tool that software engineers use to make everything
and so they need to be as easy to use as possible but also be versatile enough to allow the
programmers to build their crazy ideas. The operating system is the most important
piece of software on the computer as it is what we interact with and it controls how
all of the other programs are run on the hardware, and engineering a good operating system is
a huge challenge. This brings us to software engineering: writing
bundles of instructions telling the computer what to do. Building good software is an art
form because you have to translate your creative ideas into logical instructions in a specific
language, make it as efficient as possible to run and as free of errors as you can. So
there are many best practices and design philosophies that people follow. Some other important areas are getting many
computers to communicate and work together together to solve problems. Storing and retrieving
large amounts of data. Determining how well computer systems are performing at specific
tasks, and creating highly detailed and realistic graphics. Now we get to a really cool part of computer
science, getting computers to solve real world problems. These technologies underlie a lot
of the programs, apps and websites we use. When you are going on vacation and you want
to get the best trip for the money you are solving an optimisation problem. Optimisation
problems appear everywhere and finding the best path or most efficient combination of
parts can save businesses millions of dollars. This is related to Boolean Satisfiability
where you attempt to work out if a logic formula can be satisfied or not. This was the first
problem proved to be NP-complete and and so widely considered to be impossible to solve,
but amazing development of new SAT solvers means that huge SAT problems are solved routinely
today especially in artificial intelligence. Computers extend our brains multiply our cognitive
abilities. The forefront of computer science research is developing computer systems that
can think for themselves: Artificial Intelligence. There are many avenues that AI research takes,
the most prominent of which is machine learning which aims to develop algorithms and techniques
to enable computers to learn from large amounts of data and then use what they’ve learned
to do something useful like make decisions or classify things. And there are many different types of machine
learning. Closely related are fields like computer vision,
trying to make computers able to see objects in images like we do, which uses image processing
techniques. Natural language processing aims to get computers
to understand and communicate using human language, or to process large amounts of data
in the form of words for analysis. This commonly uses another field called knowledge
representation where data is organised according to their relationships, like words with similar
meanings are clustered together. Machine learning algorithms have improved
because of the large amount of data we give them. Big data looks at how to manage and
analyse large amounts of data to get value from it. And will get even more data from
the Internet of Things, adding data collection and communications to everyday objects. Hacking is not a traditional academic discipline
but definitely worth mentioning. Trying to find weaknesses in computer systems, and take
advantage of them without being noticed. Computational Science uses computers to help
answer scientific questions from fundamental physics to neuroscience, and often makes use
of Super Computing which throws the weight of the worlds most powerful computers at very
large problems, often in the area of Simulation. Then there is Human Computer Interaction which
looks at how to design computer systems to be easy and intuitive to use. Virtual reality,
Augmented Reality and Teleprescence enhancing or replacing our experience of reality. And
finally Robotics which gives computers a physical embodiment, from a roomba to trying to make
intelligent human like machines. So that is my Map of Computer Science, a field
that is still developing as fast as it ever has despite that fact that the underlying
hardware is hitting some hard limits as we struggle to miniaturise transistors any more,
so lots of people are working on other kinds of computers to try and overcome this problem.
Computers have had an absolutely huge impact on human society and so it is going to be
interesting to see where this technology goes in the next one hundred years. Who knows,
perhaps one day, we’ll all be computers. As per usual if you want to get hold of this
map as a poster I have made it available so check in the description below for some links,
and also if you want to find out more about computer science I recommend you check out
the sponsor for this video brilliant dot org. People often ask me how to go about learning
more about the kinds of subjects I cover in these videos, and as well as watching videos,
a really great way is to get down and solve some real problems. And brilliant does and
excellent job at this. It’s a really cool website and also an app
which helps you learn by getting you to solve interesting problems in science, mathematics
and computer science. And each of the courses starts off kind of easy and fun and then gets
more and more challenging as you master the concepts. If you want to learn specifically
about computers science they have got whole courses built around topics in this video
like logic, algorithms, machine learning, artificial intelligence, so if you want to
check that out brilliant dot org slash dos, or even better click the link in the description
below because that lets them know that you have come from here. So thanks again for watching,
and I’ll be back soon with a new video.