CM10228 / Programming Ib:
Applications of Search IV: Robots, Humans &
-I Shape of unit
- No more lectures likely to enter into CW.
- Old intro: these lectures for a) personal enrichment, b)
- Education about way more than marks!
- But then – don't want weak programmers to fail programming
- So now these lectures are all on the exam...
- Remember CW3!
- Have a lot of CW due next week, so try to finish some this
- Revision lecture
- Week 11 or Tuesday of Revision week?
I. Robots vs. Agents
- Definition of a robot: something that generates actions
in the real world based on perception of the real world
- This actually includes internet agents that book plane
- This also includes thermostats.
- Intelligent behaviour is defined by interactions with the real
- A robot's morphology
can be as important as its control.
- The Chinese
Room vs the Turing Test.
- Should robots be given ethical consideration?
- Moral agents are considered responsible for their own
behaviour, like adults, unlike children or dogs (mostly).
- Moral patients are things we owe obligation to, may
not be moral agents or even agents. e.g. paintings, the
- Ethical systems are not fixed, so there's no one answer.
- They vary by culture.
- They change over time
- shifts in norms
opinion on robot ethics is that since AI is something we
build, we are morally obliged to maintain our own moral agency
(not pass it to robots), and to try to limit our obligations
to them as well. But this is an open
- Can robots think like humans?
- Face the same laws of computation: tractability.
- But very different processors, algorithms, sensors,
- Humans are smart partly because we share computation.
- Can AI & robots exploit our computation too?
II. Types of Search III: Parallel Search
- One way to solve the problem of combinatorics is to have the
search run in parallel.
- Every time you have multiple options, have one thread take
- Think of people exploring tunnels in Scooby Doo.
- One of the problems of concurrency -- you find the villain
down your tunnel, how do you let the others know?
- If you could get enough
threads, & you had some
way to let all of them know when one of them found a solution,
then you could turn all NP problems into P problems.
- Not something you can do on one PC -- threads share the same
processor, no big win.
- Even massively parallel computers usually only have 1,000
processors, doesn't get you that
far in chess.
- Solutions people are currently working on:
- Quantum computing :
- particles can be potentially in many places at one
time, wave distributions.
- weird! but has
really been done for a small number of bits.
- Biological computing:
- use genetics to create and & or gates in bacteria.
- Takes maybe 48 hours for one operation, BUT can be done
billions of times in a single testtube.
- Again, people are really working on this (for example, Ron
- Evolution / memetics:
- If a few members of a species or a culture find a good /
winning solution, the entire species or culture may
eventually adopt that solution.
- This works much faster in memetics (ideas) then genetics
(requires waiting for children to grow up.)
- One consequence of increasing communications in the world
is that solutions with clear impact spread much more
quickly, many expect this to fundamentally change society.
- Of course so do other ideas, e.g. fashion.
- Some of the best AI works by exploiting human
culture. Like Google spell checking.
- Cloud computing
- The richest companies in the world are running on (Google)
or at least inseparably from (Apple, Microsoft, Amazon,
- Many have their own rivers to power their server farms.
- Trying to come up with good design rules is like memetics.
- This is
why sometimes programmers can chance on the right answer to a
question even sooner than they can understand how to solve the
java are designed to give you a heavily pruned search space.
technique makes it likely you will find good answers!
- Using very limited search of known good strategies is the key
to several other kinds of AI:
- Will talk about this Thursday.
III. Searching in Advance III: Natural Selection
(background for GAs)
- Natural Selection is a theory for explaining the fact of
evolution which is observed in nature and the laboratory.
- Fact is based on changes
in morphology over time,
- more recently we can also measure changes of gene
frequencies, the technical definition of biological
- Darwin guessed this all not from fossils or measurements,
but just on how animals look that live on islands – as if
one happened to get there from nearby & then speciated.
- Darwin's theory has three parts
- Every population has variation
- Parents have children that are more like them than like
the average member of their population – heritability.
- They also have more children than survive to reproduce
To the extent that which children survive is
systematic, species will change
(evolve) to follow what works well.
- Will also get some random change even when there's no
- Evolution doesn't necessarily discover optimal solutions.
- "I only have to be faster than you."
- Can think of it as an anytime algorithm.
- Can think of it as a search for heuristics.
- Sometimes heuristics work well in some environments
and not others.
- Other organisms or ideas are evolving too, will find hacks
around what used to be a good solution.
- E.g. giant birds as predators in SA, nailed when cats got
across from Asia.
- Evolution does come up with a bunch of `good tricks' (Dennett)
- This is true of cultural evolution as well.
- Although we think we are rational agents who do things that
provably make sense, in fact much of what we do and know we
get by imitation of authority.
- Learning when you don't even know you've learned is called
- Adopting mannerisms of people you admire or just observe.
- Language -- you pick up new words and even dialects
- May be most of what you learn.
- Intelligent behaviour is not just brains
- As per the first lecture, intelligent behaviour derives from
the interaction between agents and their environment.
- Brains co-evolved with the body.
- Big Dog
(from Boston Dynamics) works by exploiting both evolved
control (captured with motion capture) and evolved
dog throwing bricks. This isn't really to teach
you anything (except maybe fear).
IV Biologically Inspired (& Parallel) Learning
- Evolution-like rules.
- Pick a representation of variables that affect the
behavior of an individual agent.
- Start with a population of possible values for those
- Run all the programs and evaluate how well they do.
- Select the ones that have done the job best,
- Create `children' by systematically permuting the
variables of the winners
- crossover : choose two (or more) parents, choose some of
the variable settings from each.
- mutation : just change a small number of variables
- any other operator -- doesn't have to be naturally
- These children make a new population, now iterate! (go
back to 2)
- Good for searching around likely solutions
- Need to have a good idea what the right representation /
set of variables is.
- Need to have a good idea of how to tell if the individuals
are doing well
- (tricky when starting from a solution too far from right
- what solves the first steps well may not be able to
solve the last steps.)
- Many variations / applications:
- See demo from Karl
Sims (you'll have to chase some links...)
- Just the tip of the iceberg.
- Neural Networks
- start from production
- more inputs
- more outputs
- Applied Statistics (Chris
Bishop Book NN Book)
- Category Perception (Pattern Recognition: A second year
- importance of category perception in
- progress from perceptrons to formal
- Learning and intelligence are forms of search.
- But it's specialized so likely to succeed.
- Still any bias to search may mean that you miss a good
- Intelligence is (NP) hard!
- at least, the parts we call "intelligent".
- Come to the seminars if you like.
page author: Joanna
14 April 2015