INTELLIGENCE

1
INTELLIGENC
E
HOW A NEW UNDERSTANDING OF THE BRAIN WILL LEAD TO
THE CREATION OF TRULY INTELLIGENT MACHINES

• Jeff Hawkins
• with Sandra Blakeslee

2
INTELLIGENC
E
HOW A NEW UNDERSTANDING OF THE BRAIN WILL LEAD TO
THE CREATION OF TRULY INTELLIGENT MACHINES

• Jeff Hawkins
• with Sandra Blakeslee

3
On Intelligence
Jeff Hawkins (his ideas)
Sandra Blakeslee (her style)
4
On Intelligence

• http//www.onintelligence.org/about.php

5
On Intelligence

• A new theory of what intelligence is and
• how the brain thinks.

6
On Intelligence

• A new theory of what intelligence is and
• how the brain thinks.

Its magic!
7
On Intelligence

• A new theory of what intelligence is and
• how the brain thinks.
• Anyone can understand this theory and it explains
  much of what we do.

8
Prologue

• Jeff Hawkins passions
• mobile computing (Palm Computing Handspring
• brains (Redwood Neurosciences Inst. Numenta)
• He wants to understand
• intelligence
• how the brain works (in an engineering way)
• how to build machines that work like the brain
• how to build truly intelligent machines

9
Prologue

• Intelligence is the last frontier of science.
• Everyone has a brain.
• You are your brain.
• We can understand intelligence now.
• Large societal benefits beyond health issues.
• There are lots of neuroscientists.
• We have no theory of intelligence.
• Neurobiologists are fixated on subsystems.
• Computers and brains are different.
• The problem of intelligence can be solved.

10
(No Transcript)
11
(No Transcript)
12
(No Transcript)
13
This is a mirror
14
(No Transcript)
15
(No Transcript)
16
(No Transcript)
17
(No Transcript)
18
Prologue

• How come kids can hop rocks but robots cant?

19
Prologue

• How come kids can hop rocks but robots cant?

Nothin to it!
Forget about it!
20
Prologue

• How come 3-year olds are learning languages?

21
Prologue

• How come 3-year olds are learning languages?

• Jeff Hawkins memory prediction framework
  requires that pyramidal neurons can detect
  precise coincidences of synaptic input on thin
  dendrites. (Like Searles Chinese Room
  translator, I dont actually understand what that
  means. Mr. Hawkins paid me a lot of money to say
  that.)

22
Prologue

• How come 3-year olds are learning languages?

Jeff Hawkins memory prediction framework
requires that pyramidal neurons can detect
precise coincidences of synaptic input on thin
dendrites.
23
Prologue

• How come 3-year olds are learning languages?

Like Searles Chinese Room translator, I dont
actually understand what I just said.
24
Prologue

• How come 3-year olds are learning languages?

I did this because Mr. Hawkins is such a nice
man. Besides, he paid me a lot of money.
25
Prologue
How come you can tell a cat from a dog
26
Prologue
How come you can tell a cat from a dog
... dog ... cat dog cat cat
27
Prologue
but a computer cant?
Id say, Its raining cats and dogs.
28
And Id say the computer is more right.
I know syntax.
Wheew! Im sooo glad hes not a cat!!!
29
Prologue

• We have clues we need insights.
• August 2002 - RNI dedicated to brain theory.
• Neocortex - part responsible for intelligence.
• Dedicated to understanding the neocortex.
• Book describes a theory of how the brain works.
• What is intelligence?
• How does the brain create intelligence?
• He does not claim this is all new.
• He hopes we will get insights into why we think
  and behave the ways that we do.

30
Prologue

• He hopes some readers will build intelligent
  machines based on principles in the book.
• He is interested in real intelligence in
  contrast to artificial intelligence.
• Real because it starts with the brain.
• book starts by describing early failures in AI
  and ANN.
• core Idea memory prediction framework
  introduced.
• How physical brain implements this model, i.e.
  how the brain actually works.
• We will build intelligent machines.
• We will not be overrun by robots.

31
Free me! Dont turn off the power!
32
Prologue

• Questions covered
• Can computers be intelligent?
• Werent ANNs supposed to lead to intelligent
  machines?
• Why has it been so hard to figure out how the
  brain works?
• What is intelligence if it isnt defined by
  behavior?
• How does the brain work?
• What are the implications of this theory?
• Can we build intelligent machines and what will
  they do?
• He will explain this new theory of intelligence.

33
Prologue

• It will take years to build truly intelligent
  machines.
• This doesnt diminish the power of the core idea.
• He couldnt find a good book that described how
  the brain works.
• The most powerful things are simple.
• This book proposes a simple and
• straightforward theory of intelligence.

34
Prologue

• Social and other implications of the theory which
  for many readers might be the most thought
  provoking section of the book.
• Ends with discussion of intelligent machines
  how we can build them and what the future will be
  like.

35
I disagree with his attacks on artificial
intelligence (AI) and artificial neural networks
(ANNs). Most of us working in these fields were
not looking for a generalized model of
intelligence but rather robust tools for
practical problem solving. At least two of
tonight's presenters have demonstrated successes
with these tools me and Judy Dayhoff. Check out
our publications.
36
I dont like the high school debate about what
intelligence is. Is it passing Alan Turing's
test, John Searles Chinese Room test or Jeff
Hawkins prediction score? This is a pointless
semantic argument. I am interested in having
computers help us to improve the health of
everyone. Id rather focus on useful rather
than intelligence.
37
I like this book a lot for its practical
implications! I agree with Jeff that it is
worthwhile to build hierarchical memory models
based on knowledge of neocortex functioning. I
hope Jeff will come to the NIH and brainstorm
with us about how we might do this and how we can
approach some hard biomedical problems like
protein folding, signal enhancement in microarray
data, and data mining.