Transcript Doawnload PPT file

Why General Artificial
Intelligence (AI) is so Hard
Theo Pavlidis
Distinguished Professor Emeritus
Dept. of Computer Science
[email protected]
http://theopavlidis.com
Definitions of Artificial
Intelligence (AI)
• General or Strong AI: A machine that
replicates the functionality of the human
brain. “Around the Corner” since about 1945.
• Narrow or Weak AI: A machine that does a
specific task that traditionally has been done
by humans. Each specific application is
treated as a separate engineering problem.
Numerous successes.
10/9/2009
Why General AI is so Hard - CS talk
2
Successes in Narrow AI
(Seen in daily life)
• Restricted Speech Recognition (in Banking
and Airline reservation systems, etc)
• Credit Card Fraud Detection
• Web Tools (Shopping Suggestions, Mechanical
Translation, etc)
• Simple Robots (Roomba house cleaner)
• 1D and 2D Bar Codes (in stores and in
shipping)
10/9/2009
Why General AI is so Hard - CS talk
3
Successes in Narrow AI
(Not Seen Everyday)
•
•
•
•
•
Chess Playing Machines
Optical Character Recognition
Industrial Inspection
Biometrics (Fingerprints, Iris, etc)
Medical Diagnosis
10/9/2009
Why General AI is so Hard - CS talk
4
Features of Narrow AI
• Each Problem is Solved Separately even
though certain common mathematical tools
may be used (statistics, graph theory, signal
processing, etc).
• Each Solution Relies Heavily on Specific
Environment Constraints and performance
(compared to that of humans) drops when
these constraints are relaxed.
10/9/2009
Why General AI is so Hard - CS talk
5
Why Not General AI?
• Why “waste” time with all the special cases and
not solve the general problem once for all?
• Why not use a “brain model” to solve all these
problems?
• Are advances in general computer technology
(hardware, systems) likely to help? Why not
wait for them rather than solving problems
piecemeal?
10/9/2009
Why General AI is so Hard - CS talk
6
Humans may be machines, but they
are very different from computers
10/9/2009
Why General AI is so Hard - CS talk
7
Understanding the Difference
between
Humans and Computers
• We will start by looking at the problem of
content-based image retrieval to obtain an
understanding of the difference.
10/9/2009
Why General AI is so Hard - CS talk
8
Content-based Image Retrieval
(CBIR)
• Given an image find those that are similar to it from a
data base of images. (If the images are labeled, the
problem is reduced to text search.)
• Systems do not perform as advertised. For a
collection of critical writings see
– http://www.theopavlidis.com/technology/CBIR/index.htm
• The difficulty of image retrieval should be contrasted
with the success of text retrieval, not only Google, but
also earlier programs such as the Unix grep.
10/9/2009
Why General AI is so Hard - CS talk
9
Example
10/9/2009
Why General AI is so Hard - CS talk
10
Reasons for the Poor Results in
Machine Vision and CBIR
• Images are represented by statistics of pixel
values (e.g. color histogram, texture
histogram, etc)
• Such statistics are unrelated to human
perception.
• Papers describing CBIR methods use trivial
queries (e.g. “show me all pictures with a lot
of green”).
10/9/2009
Why General AI is so Hard - CS talk
11
Perceptual versus Computational
Similarity
• Two pictures may differ a lot in their pixel
values but appear similar to a person. (“They
have the same meaning”.)
• Two pictures may differ in very few pixels but
they have different meaning. (Face portraits
of two different people in front of the same
background.)
10/9/2009
Why General AI is so Hard - CS talk
12
Perceptual versus Computational
Similarity
Perceptually close
10/9/2009
Pixel-wise close
Why General AI is so Hard - CS talk
13
Text versus Pictures
• In text files each byte (or two) is a numerical
code for a character. Therefore strings of
bytes correspond to words that carry
semantic meaning.
• In pictures each byte (or group thereof)
represents the color at a particular location
(pixel). Pixels are quite far from the
components that have a semantic meaning.
10/9/2009
Why General AI is so Hard - CS talk
14
We do not do that well in text!
• If it is hard to search for concepts unless we
can map concepts into words.
• Example 1: Find all articles critical of the
government policy in dealing with the banking
crisis.
• Example 2: Find all articles about a dog
named Lucy. Amongst the Google returns
was an article with the phrase: “Lucy and I spent
the weekend alone together. We have a dog named
Kyler.”
10/9/2009
Why General AI is so Hard - CS talk
15
Human Intelligence made simple
Input
Concept
Input
Output
10/9/2009
Why General AI is so Hard - CS talk
16
The Big Difference
• The transformation of input to concept is a complex
process (binding), barely understood by
neuroscientists. (In spite of claims to the opposite by
some computer scientists.)
• It is hard to develop algorithms
for a barely understood process.
• Humans can transform concepts into formal entities
(words in a language) and then code them in
computer readable form.
• Computers can deal with such formal input.
10/9/2009
Why General AI is so Hard - CS talk
17
What Neuroscientist Say
• “Perceptions emerge as a result of
reverberations of signals between different
levels of the sensory hierarchy, indeed across
different senses”. The author then goes on to
criticize the view that “sensory processing
involves a one-way cascade of information
(processing)”
• Source: V.S. Ramachandran and S. Blakeslee Phantoms in the
Brain, William Morrow and Company Inc., New York, 1998 (p. 56)
10/9/2009
Why General AI is so Hard - CS talk
18
What Do You See?
10/9/2009
Why General AI is so Hard - CS talk
19
Reading Demo - 1
10/9/2009
Why General AI is so Hard - CS talk
20
Reading Demo - 1
Tentative binding on the letter shapes (bottom
up) is finalized once a word is recognized (top
down). Word shape and meaning over-ride early
cues.
10/9/2009
Why General AI is so Hard - CS talk
21
Reading Demo -2
New York State lacks proper facilities
for the mentally III.
The New York Jets won Superbowl III.
• Human readers may ignore entirely the shape of
individual letters if they can infer the meaning
through context.
10/9/2009
Why General AI is so Hard - CS talk
22
The Importance of Context
• “Human intelligence almost always thrives on
context while computers work on abstract
numbers alone. … Independence from
context is in fact a great strength of
mathematics.”
• Source: Arno Penzias Ideas and Information,
Norton, 1989, p. 49.
10/9/2009
Why General AI is so Hard - CS talk
23
The Challenges
• We need to replicate complex transformations
that the (human/animal) brain has evolved to do
over millions of years.
• We have to deal with the fact the processing is
not unidirectional and also affected by other
factors than the input (context). (Such factors
cause visual illusions.)
10/9/2009
Why General AI is so Hard - CS talk
24
A time scale
• The human visual system has evolved from
animal visual systems over a period of more
than 100 million years.
• Speech is barely over 100 thousand years old.
• Written text is no more than 10 thousand years
old.
10/9/2009
Why General AI is so Hard - CS talk
25
A note on brain models
• There is a history for considering the latest
technology to be a model of the human brain,
for example in the 16th century irrigations
networks were considered to be models of
the brain.
• If someone claims to have a machine
modeling the human brain, ask how could the
machine be modified to model the brain of a
dog (since a dog cannot learn to write poetry,
play chess, etc)?
10/9/2009
Why General AI is so Hard - CS talk
26
A Note on Neural Nets
Is this a model of the brain?
As much as a table is a model of a dog.
10/9/2009
Why General AI is so Hard - CS talk
27
Simplified model of a small part of the brain
10/9/2009
Why General AI is so Hard - CS talk
28
A Dubious Approach
• “Training” on large numbers of samples has
been used as a way out of finding a way to
understand what is going on.
• But humans (and animals) do not need to be
trained on large numbers of samples.
• Rats trained to distinguish between a square
and a rectangle perform quite well when
faced with skinnier rectangles. They have the
concept of rectangle!
10/9/2009
Why General AI is so Hard - CS talk
29
Distinguish Rectangles from Squares
The Artificially Intelligent Approach
• Take a hundred (or more) pictures of
rectangles and squares, compute several
statistics on each picture and for each picture
create a “feature” vector F. Then compute a
vector W so that
F’W > 0 for squares and
F’W < 0 for rectangles
10/9/2009
Why General AI is so Hard - CS talk
30
Distinguish Rectangles from Squares
The Natural Approach
• Find the outline of a shape (if one exists in a
picture) and fit a rectangle to it. Then
compute the aspect ratio of the rectangle. If it
is near 1 (for some given tolerance), then it is
called a square, otherwise a rectangle.
• Criticism: Method lacks generality!!!
10/9/2009
Why General AI is so Hard - CS talk
31
No Generality in Nature
• The animal visual systems has many special
areas for visual tasks (about 30 in the human
case).
• We have already seen examples where “high
level” (context) recognition takes quickly over
the low level data processing.
10/9/2009
Why General AI is so Hard - CS talk
32
Negator of Generality
10/9/2009
Why General AI is so Hard - CS talk
33
The Learning Machine (neural net)
Approach
• It has the appeal of getting something for
nothing, so it is kept alive.
• We can “solve” a problem without really
understanding it.
• Give a learning machine “enough” samples
and a classifier will be found!!!
• (Forget about the rat who only needs two
samples.)
10/9/2009
Why General AI is so Hard - CS talk
34
Criteria for Choosing a
Problem to Work on
• Context should either be known or not important.
• Processing of the input should be relatively simple
(it should be clear what kind of information we
need to extract).
• For an example relying heavily on context see:
technology/BoxDimensions/overview.htm
on my web site.
• Comments on major areas in the next few slides.
10/9/2009
Why General AI is so Hard - CS talk
35
Speech Recognition
• Grammar driven models (using low level
context) have been quite successful.
• High level context is even better. For
example, matching a speech fragment to a
name on a list.
10/9/2009
Why General AI is so Hard - CS talk
36
Optical Character Recognition
(OCR)
• Printed text characters have small shape
variability and high contrast with the
background.
• Spelling checkers (or ZIP code directories in
postal applications) introduce low level
context.
10/9/2009
Why General AI is so Hard - CS talk
37
An example of heavy use of context
• Reading of the checks sent for payment to
American Express.
• Because payments are supposed to be in full
and the amount due is known, the number
written on a check is analyzed to confirm
whether it matches the amount due or not.
• (But direct payment is used more and more!)
10/9/2009
Why General AI is so Hard - CS talk
38
An Aside: Why did OCR mature when
the need for it was diminished?
• The algorithms used in the products of the 1990s
were known earlier but they were too complex to be
implemented effectively with the digital technology of
earlier times.
• When computer hardware became cheap enough for
good OCR, it also became cheap enough for direct
text entry through PCs and the Internet.
• Keep this in mind in your business plans!
10/9/2009
Why General AI is so Hard - CS talk
39
Face Recognition
• It took over thirty years to built acceptable
quality machines that recognize printed
symbols. What makes us think that we can
solve the much more complex problem of
distinguishing human faces?
• Neuroscientists point out that humans have
special neural circuitry for face recognition.
10/9/2009
Why General AI is so Hard - CS talk
40
How these two faces differ?
10/9/2009
Why General AI is so Hard - CS talk
41
How about these two?
10/9/2009
Why General AI is so Hard - CS talk
42
Face Recognition and
Scalability
• The population samples in published studies
are relatively small and include men and
women of different races with different
hairstyles, etc.
• I have never seen a study where all the
subjects are similar. For example, white blond
men between the ages of 20 and 30 with long
hair and beards.
• Subjects in published studies are
cooperative.
10/9/2009
Why General AI is so Hard - CS talk
43
How About Deep Blue?
• In 1997 a chess machine (IBM’s “Deep Blue”)
beat the human world champion Garry
Kasparov.
• This resulted in a lot of publicity on how
computers had become smarter than
humans.
10/9/2009
Why General AI is so Hard - CS talk
44
However
Chess is a deterministic game, so a computer
could derive a winning solution analytically.
On the other hand the number of all possible
positions is so large (10120) that using even
the fastest available computer it will take
billions of years to consider all possible
moves.
• Skilled players may look at 20 moves ahead
by pruning, i.e. ignoring non-promising
moves.
10/9/2009
Why General AI is so Hard - CS talk
45
Chess Playing Machines
• Around 1980 Ken Thompson developed a
chess playing program called Belle based on
a minicomputer with a hardware attachment
used to generate moves very fast.
• Belle defeated all other computer programs
and became the world champion.
• The use of special chess knowledge and
special purpose hardware became the
preferred approach since then.
10/9/2009
Why General AI is so Hard - CS talk
46
More on Deep Blue
• A major focus of the effort was the
development of special purpose hardware.
• An expert chess player (Murray Campbell )
contributed the evaluation functions of the
moves generated by the hardware.
• The project had as a consultant an
international grandmaster (Joel Benjamin
who had played Kasparov to a draw in 1994).
10/9/2009
Why General AI is so Hard - CS talk
47
Concluding Remarks
• Before we try to built a machine to achieve a
goal we must ask ourselves whether that goal
is compatible with the laws of nature . (Not
because “people can do it”.)
• While such laws are clear in Physics and
Chemistry, there are not in the field of
Computation except in some extreme cases.
10/9/2009
Why General AI is so Hard - CS talk
48
Human Credulity - 1
• In spite of well understood laws of physics
“inventors” persist in offering designs that
violate them and they find takers.
• Therefore fundamental advances in
Computer Science are likely to reduce but not
to eliminate preposterous claims.
10/9/2009
Why General AI is so Hard - CS talk
49
Human Credulity - 2
• 50 years ago Langmuir (in “Pathological
Science”) debunked UFOs but also predicted
that UFOs will be with us for a long time
because it is too good a story for the news
media to let go.
• The view of computers as giant brains that
are able to out-think and replace humans is
about as valid as visits by extraterrestrials,
but it makes too good a story for the news
media to let go.
10/9/2009
Why General AI is so Hard - CS talk
50
The End
That’s all folks
10/9/2009
Why General AI is so Hard - CS talk
51