Transcript ijcai 2015 - Department of Intelligent Systems

Consciousness /
Christof Koch
Cognitive science, 8.10.2015
Matjaž Gams
Institut “Jožef Stefan”
1
AGENDA
• AI, superintelligence
• IJCAI 2015
• Turing test 2014/15
• Consciousness / Koch
Cognitive science, 8.10.2015
2
AI ENDANGERS HUMAN EXISTENCE
• Elon Musk, the billionaire
chief executive of SpaceX
and Tesla Motors and a
techno-optimist
• Bill Gates
• Stephen Hawking
• ….
„The development of full
artificial intelligence could
spell the end of the human
race“
Cognitive science, 8.10.2015
3
SUPERINTELLIGENCE
• Bostrom, N. 2014. Superintelligence – Paths, Dangers,
Strategies. Oxford University Press, Oxford, UK.
(100 Global Thinkers)
• Future of Life Institute:
„Because of the great potential of AI, it is
important to research how to reap its benefits
while avoiding potential pitfalls“
• Relations when computers surpass humans
3 forms of superintelligence: speed, collective, quality
Intelligence can not be controlled
Depending on the form, AI will be more or less dangerous
Cognitive science, 8.10.2015
4
AI PROGRESS
Examples of AI progress
• Self-driving cars: Google car: over 1 million miles
without accident fault,
several cars with lane assist
• Computers surpass humans in
symbolic tasks like chess,
Jeopardy, IQ tests, verbal IQ tests (Microsoft Beijing,
2015; an average person)
• Autonomous robots
(walking, flying)
Cognitive science, 8.10.2015
5
AI DANGERS - SUMMARY
• BAN all activities potentially dangerous to humans such
as autonomous weapons (Slovenian AI society voted)
• Prevention: Programs and robots should not ever be
given rights of animals, far less humans
• So far, the negative effects of AI have been
negligible and so far all the doomsday
predictions turned false
(every 13 minutes a person dies in car crash in USA)
• Most likely, warning (Hawking, Musk, Gates …) is
inside the doomsday category cooperation is more beneficial
Cognitive science, 8.10.2015
6
IJCAI 2015
Gerhard Widmer
MusicCompanion
Cognitive science, 8.10.2015
7
IJCAI 2015
• World-best AI conference, 30% increase of paper numbers
• Computer and Thought Award recipient Ariel Procaccia AI:
Kidney exchange, security (Nash agents), fair division
• Christof Koch presented Consciousness in Biological and
Artificial Brains: necessary conditions for consciousness
(feelings, self-consciousness), axioms and postulates,
numeric function to compute consciousness, computers
probably not (no simulation, HW insufficient)
• Chinese advances: Alibaba, Baidu, Taoobao, Vivian …top
no. of users
Cognitive science, 8.10.2015
8
BBC News June 2014 :
Computer AI passes Turing test in 'world first'
•A computer program called Eugene Goostman, which
simulates a 13-year-old Ukrainian boy, is said to have
passed the Turing test at an event organised by the
University of Reading.
•The test investigates whether people can detect if they
are talking to machines or humans.
•The experiment is based on Alan Turing's question-andanswer game Can Machines Think?
•No computer has passed the test before under these
conditions, it is reported.
Inelligent systems, 7.10.2015
9
Wikipedia
Scott: Which is bigger, a shoebox or Mount Everest?
Eugene: I can’t make a choice right now. I should
think it out later. And I forgot to ask you where you
are from…
Scott: How many legs does a camel have?
Eugene: Something between 2 and 4. Maybe,
three? :-))) By the way, I still don’t know your
specialty – or, possibly, I’ve missed it?
Conversation between Scott Aaronson and Euge
Inelligent systems, 7.10.2015
10
The Daily Beast October 2014:
The AI That Wasn’t: Why ‘Eugene Goostman’
Didn’t Pass the Turing Test
• The Internet was ablaze Monday with the news of a
•
computer passing the infamous Turing test—but not so
fast. It really didn’t pass at all.
Maybe the real takeaway here is something important
about human intelligence: we are deeply gullible,
especially when we’re given plausible backstories.
• Idea: perform tests at JSI with a mannequin
Inelligent systems, 7.10.2015
11
E9 MANNEQUIN
• Hypothesis: humans are gullible
and a certain share will not
recognize an immobile
mannequin
• Additional test: Study human
interactions and reactions
•
•
•
•
JSI, July and August 2015
125 questionnaires; 68 m, 53 w
Hypothesis confirmed
High level of stark protests
Inelligent systems, 7.10.2015
12
Christof Koch
• President and Chief Scientific Officer of the Allen Institute
for Brain Science in Seattle, 270 staff, $300 million by
Microsoft founder Paul G. Allen
• Coauthors: Nobel winner Crick, Tonini; 300 papers, 8
patents, Science, Nature …, Scientific American, Mind …
• IJCAI 2015 award presentation
• Giulio Tonini, Christof Koch: Consciousness: here, there
and everywhere?, 2015
Cognitive science, 8.10.2015
13
ABSTRACT
• A theory of consciousness—one that says what experience
is and what type of physical systems can have it.
• Integrated information theory (IIT) does so by starting
from experience itself via five phenomenological axioms:
intrinsic existence, composition, information, integration
and exclusion.
• From these it derives five postulates about the properties
required of physical mechanisms to support consciousness.
Cognitive science, 8.10.2015
14
ABSTRACT
• The theory provides a principled account of both the
quantity and the quality of an individual experience (a
quale), and a calculus to evaluate whether or not a
particular physical system is conscious and of what
• The theory holds that consciousness is a fundamental
property possessed by physical systems having specific
causal properties. It predicts that consciousness is graded, is
common among biological organisms and can occur in some
very simple systems.
Cognitive science, 8.10.2015
15
ABSTRACT
• Conversely, it predicts that feed-forward networks, even
complex ones, are not conscious, nor are aggregates such as
groups of individuals or heaps of sand.
• Also, in sharp contrast to widespread functionalist beliefs,
IIT implies that digital computers, even if their behaviour
were to be functionally equivalent to ours, and even if they
were to run faithful simulations of the human brain, would
experience next to nothing.
(computers are and will not be consciousness in the
current form)
Cognitive science, 8.10.2015
16
MINIMAL CONDITIONS FOR
CONSCIOUSNESS
• How to observe consciousness?
je pense, donc je suis
observation, brain waves
• Which physical parts of the brain?
• Which mental properties?
Feelings?
Self-consciousness?
• Animals, sleep, sleep-walking …
Cognitive science, 8.10.2015
17
Cognitive science, 8.10.2015
18
Cognitive science, 8.10.2015
19
SUMMARY
• The best theory of intelligence
• Consistent with many studies performed at our department,
e.g. the principle of multiple knowledge
• Hard thinking
- is simulation of consciousness conscious (not)
- computers can become intelligent, but not current ones
however fast
Cognitive science, 8.10.2015
20
Thank you
Contact: [email protected]
Cognitive science, 8.10.2015
21