December 2011
FROM WEAK AI TO STRONG AI
Selected and edited extracts
Paul Budding
Introduction
In this paper I have sought to find relevant extracts from Ray Kurzweil and
others… concerning the progress being made in Artificial Intelligence. (AI)
More specifically, I have selected (and edited) comments, quotes and lengthy
extracts from interviews that (I believe when put together) gets across the
progress being made concerning going beyond Weak/Narrow (or Task-Specific)
AI on the path towards the holy-grail of Strong AI.
No specialist technological knowledge is required to follow what is written - - -
although those who are familiar with the Technological Singularity discourse
will find this more comprehensible than those unfamiliar with the concept.
1
About David Ferrucci‟s Latest Project: Extracted from:
Website: IBM
Title: IBM Watson: Research Team: Dr. David Ferrucci
Address: http://www-03.ibm.com/innovation/us/watson/research-team/dr-
david-ferrucci.html
Dr. David Ferrucci is a Research Staff Member and leader of the Semantic Analysis and
Integration Department at IBM‘s T.J. Watson‘s Research Center. His team of 25 researchers
focuses on developing technologies for discovering knowledge in natural language and
leveraging those technologies in a variety of intelligent search, data analytics, and knowledge
management solutions.
In 2007, Dr. Ferrucci began exploring the feasibility of designing a computer system that can
rival human champions at the game of Jeopardy! Dubbed DeepQA, the project focused on
advancing natural language question answering using massively parallel evidence-based
computing. After winning support, Ferrucci has set and driven the technical agenda for
Jeopardy! The IBM Challenge.
The Watson computer system designed by Ferrucci‘s team represents the integration and
advancement of many search, natural language processing, and semantic technologies.
Following the Jeopardy! challenge, Dr. Ferrucci and his team plan to apply DeepQA
technologies to areas like medicine, government, and law to drive advances in computer
supported intelligence and decision-making.
―The opportunity to pursue an exploratory project that took an area of science that I was most
interested in, and to bring together a team of world class people, and push the limits – it
doesn't get any better than that.‖
Dr. David Ferrucci
Principal Investigator,
DeepQA/Watson project
2
The following is extracted from Exponential Times: The Future Comes
Faster than you Think. (Website)
Title: A Computer called Watson
Speaker: David Ferrucci interviewed by Richard Waters
Date: 2011
Address: http://exponentialtimes.net/videos/computer-called-Watson
Where Ray Kurzweil refers to ‗Weak‘ or ‗Narrow‘ AI… David Ferrucci refers to ‗Task-
Specific‘ AI. Both Kurzweil and Ferrucci are meaning the same thing here. And while
Ferrucci is focused on Task-Specific AI (Kurzweil of course is hungry for progress towards
Strong AI) Ferrucci is nevertheless in agreement with Kurzweil that Watson has excellent
implications for healthcare. Kurzweil goes further in declaring Watson a milestone on the
path to Strong AI. (See later in this paper). Ferrucci isn‘t focused on Strong AI but discusses
it in this video and is (again) in agreement with Kurzweil concerning what it would entail.
Ferrucci informs us that so far AI has been characterized as ―Task-Specific‖. And ―a program
is artificially intelligent if it performs tasks [in the sense of] if that task was performed by a
human you would consider the human intelligent.‖ Ferrucci gives the example of Deep Blue
beating Gary Kasparov at Chess. Kasparov was then the best human player on the planet.
Deep Blue was therefore a Task-Specific/Weak AI. Its abilities are limited (to chess) as
opposed to possessing ‗general‘ intelligence like a human. Concerning Watson, Ferrucci says
―…Watson‘s performing like a human [when it plays] the game of Jeopardy. But can Watson
sit down and have this conversation? No. […] You end up with a very Task-Specific
perspective.‖ Consistent with Kurzweil, Ferrucci points out that it would take the passing of
the Turing Test for a machine to go beyond Weak Task-Specific AI and become Strong AI.
Afterall ―…we expect a lot more from humans than just to do these specific tasks.‖
Ferrucci discusses the problem of establishing Strong AI. Strong AI must be flexible, broad,
general, not narrow, not limited. This is the ―hard part‖ of artificial intelligence. ―How do we
get computers to understand things in human terms […] as something that can‘t be reduced to
well-defined mathematics.‖ Ferrucci thinks that the example of chess contrasted with
―understanding human language‖ has ―framed this problem nicely because […] chess is this
3
mathematically well-defined problem. There‘s a certain number of pieces, there‘s a finite
grid, each piece has an exact mathematically defined pattern in which you can move on the
grid. You can imagine how a computer can follow these rules, move those pieces around and
say did I win yet? […] There‘s no ambiguity. [… It…] doesn‘t [require] any other
background knowledge to play the game.‖ Chess is a machine-like function. But when you
get into human language, machines aren‘t there yet.
Ferrucci gives an example of how difficult it is to get machines to think like humans. If
someone shouts to their child hey come here, this is really interesting and the child rushes to
see what it is that is interesting… but finds it boring… and this same event repeats itself a
number of times… then the child might respond to the parent shouting hey come here, this is
really interesting… by saying ―Daddy, interesting things are boring.‖ Ferrucci says that this
is a real-life example involving himself and his daughter. His daughter associated the word
interesting with boring. So language has rules but the rules change depending on the context.
So it is clearly difficult to get computers to understand the context that language is embedded
in.
Skepticism over the possibility of a computer like Watson existed within the field of AI itself.
Executives searched in vain for two years in their efforts to find AI researchers who believed
in the technology. This demonstrates that it isn‘t only technologies that really are decades
away that attract nay-sayers. On the contrary, Ferrucci informs us that ―there were a lot of
nay-sayers‖ who were declaring that ―…Your going to fail, it‘s pure folly, you‘ll destroy your
career.‖ Thankfully, concerning Watson AI, Ferrucci himself felt up to the task. After
conducting a feasibility study he came back and made the case for the technology to the
executives. Even then people disagreed with him!
Ferrucci and his team did not try and model the human brain. Computer scientists are credited
with Watson‘s success as opposed to neuroscientists and cognitive psychologists. Ferrucci
admits that he doesn‘t know ―how a brain is organized physiologically‖ but that ―I could tell
you […] that I‘m conscious of the similar procedure so the computers using a similar
procedure that you might be conscious of that you use when you answer these questions.‖
Watson‘s similarity of method (concerning intelligence) is amplified (Kurzweil refers to
amplifying the human brains intelligence in AI later in this paper). Watson had access to the
equivalent of approximately one million books.
4
Watson‘s success has benefitted health-care. Ferrucci makes it clear that Watson is an
advance on Expert Systems. What Watson AI does is say ―Here is why it might be Lyme
Disease. Here‘s why it might be Arthritis. And my explanation is from these references. [e.g.
Medical Journals]. If you read this and then read this and read this you will know exactly why
I was more confident in Lyme Disease than Arthritis. So your bringing an explanation to the
human, to the decision-maker based on natural language content.‖
So Watson is useful for ―anytime your exploring possibilities and you know that there may be
different alternatives, different options and you want to evidence those options […] from a
body of structured and unstructured knowledge.‖ Human decisions become more informed…
Watson supports and enables humans to reach a decision. The technology underlying
Watson‘s answers ―is generated by many […] hypotheses, gathering and scoring evidence
and that‘s where we are going with the technology.‖
The following is extracted from IBM Developer Works Website.
Title: Watson and Healthcare: How Natural Language Processing and
Semantic Search could Revolutionize Clinical Decision Support.
Author: Michael J. Yuan
Date: 12th April 2011
Address: http://www.ibm.com/developerworks/industry/library/ind-watson
Watson in action
To see the kinds of questions Watson can answer, check out the two example questions Dr.
David Ferrucci showed to German Chancellor Merkel and Turkish PM Erdogan at the CeBIT
2011 Opening Ceremony. Try typing those questions into Google to see how hard it is to
parse the answer from a list of documents, and how easy it is to miss less frequent diagnoses.
1. Question: Streptococci cause this childhood "fever" characterized by a bright red rash
and high temperature.
Answer: 98% Scarlet fever, 15% Rheumatic fever, 8% Strep throat
5
2. Question: This disease can cause uveitis in a patient with family history of arthritis
presenting circular rash, fever, and headache.
Answer: 76% Lyme Disease, 1% Behcet's Disease, 1% Sarcoidosis
Extracted from Mountain Vision (Website)
Title: Ray Kurzweil: Humans Will Become Cyborgs (Video) Humanity's
transbiological future
Speaker: Ray Kurzweil interviewed by David Orban.
Date: March 16th 2011
Address: http://mountainvision.blogspot.com/2011/03/ray-kurzweil-
humans-will-become-cyborgs.html
Kurzweil said IBM's Watson technology "would still not pass a valid Turing Test. It's not at
that level yet. But it's a very important milestone towards getting there."
Kurzweil has issues concerning key concepts and terminology used by some or many
advocates of the Technological Singularity. For example he is not keen on the use of virtual
reality because it implies that virtual reality technology is not really real… when Kurzweil
says that the telephone is a virtual reality and is a real extension of self. However he uses the
term itself because we are stuck with it. In this video Kurzweil says:
―One [concept] I really don't use and object to is transhumanism because it implies we are
going to transcend our humanity. I think we are actually going to enhance our humanity. We
are going to transcend the limitations of biology and be transbiological.‖
Elsewhere on the Stepcase Lifehack Website a commenter argues in an
article titled „Can the Life-hacking concept help you live until the
Singularity?‟ that mind-uploading will be possible in the future, because…
―By then the brain may have a way of being transferred synapse by synapse, nerve by nerve,
cell by cell, piece by piece to retain our sense of conscious identity and 'self' or 'soul' instead
6
of simply making a 'copy' …‖ (http://www.lifehack.org/articles/lifestyle/can-the-
lifehacking-concept-help-you-live-until-the-singularity.html)
Extracted from Kurzweil, 2005, The Singularity is Near: When Humans
Transcend Biology (Penguin)
―A good example of the divergence between human intelligence and contemporary [narrow]
AI is how each undertakes the solution of a chess problem. Humans do so by recognizing
patterns, while machines build huge logical ―trees‖ of possible moves and counter-moves.
Most technology (of all kinds) to date has used this latter ―top-down,‖ analytical, engineering
approach. Our flying machines, for example, do not attempt to re-create the physiology and
mechanics of birds. But as our tools for reverse engineering the ways of nature are growing
rapidly in sophistication, technology is moving toward emulating nature while implementing
these techniques in far more capable substrates. [i.e. characteristic of Strong AI].‖ (p146)
The following is extracted from the Singularity Hub:
Science/Technology/The Future of Mankind. (Website)
Title: The Mind and how to Build One (Video)
Author: Ray Kurzweil
Date: December 21st 2010
Address: http://singularityhub.com/2010/12/21/ray-kurzweil-the-mind-
and-how-to-build-one-video/
We will reverse engineer the human brain in order ―to understand the principles of operation
[…] Then we can take these principles and focus on them and amplify them and leverage
them. That‘s basically what technology does.‖
They will of course they amplified in Strong AI.
Extracts from the World Future Society (Website)
7
Title: The Futurist interviews Ray Kurzweil
Speaker: Ray Kurzweil
Address: http://www.wfs.org/content/futurist-interviews-ray-kurzweil
―… if we understand the basic principles by which the brain creates intelligent behavior, we can focus
and leverage it and create much more powerful systems.‖
―We have hundreds of examples today of Narrow AI—programs doing tasks that used to be done by
human intelligence but doing them better and less expensively—and the narrowness is gradually
getting less narrow. And this intelligence is deeply integrated with our own already, even if, for the
most part, it‘s not yet in our bodies and brains. There‘s going to be a continuous exponential
progression of computers getting more powerful, getting smaller, and we‘re going to become more
and more integrated with them. And they‘ve already made us smarter, and I don‘t just mean as
measured by IQ tests. I mean by measurement of [the] intellectual […] capability of our civilization,
which includes all of the things that we can do with biological and non-biological intelligence
working together.
That integration is going to become more and more intimate. In 2035, you‘re not going to be able to
walk into a room and say, ―humans on the right side, machines on the left.‖ It‘s going to be all mixed
up and integrated—one complex, dynamic, chaotic human/machine civilization.‖
The following is extracted from : Kurzweil Accelerating Intelligence
(Website).
Title: Ray Kurzweil Responds to “Ray Kurzweil does not understand the
Brain”
Author: Ray Kurzweil
Date: August 20th 2010
Address: http://www.kurzweilai.net/ray-kurzweil-responds-to-ray-
kurzweil-does-not-understand-the-brain
8
―It is true that the brain gains a great deal of information by interacting with its environment
– it is an adaptive learning system. But we should not confuse the information that is learned
with the innate design of the brain. The question we are trying to address is: what is the
complexity of this system (that we call the brain) that makes it capable of self-organizing and
learning from its environment? The original source of that design is the genome (plus a small
amount of information from the epigenetic machinery), so we can gain an estimate of the
amount of information in this way.
[…]
Halfway through the genome project, the project‘s original critics were still going strong,
pointing out that we were halfway through the 15 year project and only 1 percent of the
genome had been identified. The project was declared a failure by many skeptics at this point.
But the project had been doubling in price-performance and capacity every year, and at one
percent it was only seven doublings (at one year per doubling) away from completion. It was
indeed completed seven years later. Similarly, my projection of a worldwide communication
network tying together tens and ultimately hundreds of millions of people, emerging in the
mid to late 1990s, was scoffed at in the 1980s, when the entire U.S. Defense Budget could
only tie together a few thousand scientists with the ARPANET. But it happened as I
predicted, and again this resulted from the power of exponential growth.
Linear thinking about the future is hardwired into our brains. Linear predictions of the future
were quite sufficient when our brains were evolving. At that time, our most pressing problem
was figuring out where that animal running after us was going to be in 20 seconds. Linear
projections worked quite well thousands of years ago and became hardwired. But exponential
growth is the reality of information technology.
We‘ve seen smooth exponential growth in the price-performance and capacity of computing
devices since the 1890 U.S. census, in the capacity of wireless data networks for over 100
years, and in biological technologies since before the genome project. There are dozens of
other examples. This exponential progress applies to every aspect of the effort to reverse-
engineer the brain.‖
9
The following is extracted from Kurzweil: Accelerating Intelligence
(Website)
Title: Kurzweil Responds: Don‟t underestimate the Singularity
Date: October 20th 2011
Address: http://www.kurzweilai.net/kurzweil-responds-dont-
underestimate-the-singularity
―Allen articulates what I describe in my book as the ―scientist‘s pessimism.‖ Scientists
working on the next generation are invariably struggling with that next set of challenges, so if
someone describes what the technology will look like in 10 generations, their eyes glaze over.
One of the pioneers of integrated circuits was describing to me recently the struggles to go
from 10 micron (10,000-nanometer) feature sizes to five-micron (5,000 nanometers) features
over 30 years ago. They were cautiously confident of this goal, but when people predicted
that someday we would actually have circuitry with feature sizes under one micron (1,000
nanometers), most of the scientists struggling to get to five microns thought that was too wild
to contemplate. Objections were made on the fragility of circuitry at that level of precision,
thermal effects, and so on. Well, today, Intel is starting to use chips with 22-nanometer gate
lengths.
We saw the same pessimism with the genome project. Halfway through the 15-year project,
only 1 percent of the genome had been collected, and critics were proposing basic limits on
how quickly the genome could be sequenced without destroying the delicate genetic
structures. But the exponential growth in both capacity and price performance continued
(both roughly doubling every year), and the project was finished seven years later. The
project to reverse-engineer the human brain is making similar progress. It is only recently, for
example, that we have reached a threshold with noninvasive scanning techniques that we can
see individual interneuronal connections forming and firing in real time.‖
The following is extracted from Wired.com (Website)
Title: Reverse-Engineering of Human Brain likely by 2030 Expert Predicts
Speaker: Ray Kurzweil in Priya Ganapati
Date: August 16th 2010
10
Address: http://www.wired.com/gadgetlab/2010/08/reverse-engineering-
brain-kurzweil/
―The singular criticism of the singularity is that brain is too complicated, too magical and
there‘s something about its properties we can‘t emulate,‖ Kurzweil told attendees at the
Singularity Summit over the weekend. ―But the exponential growth in technology is being
applied to reverse-engineer the brain, arguably the most important project in history.‖
For nearly a decade, neuroscientists, computer engineers and psychologists have been
working to simulate the human brain so they can ultimately create a computing architecture
based on how the mind works.
Reverse-engineering some aspects of hearing and speech has helped stimulate the
development of artificial hearing and speech recognition, says Kurzweil. Being able to do
that for the human brain could change our world significantly, he says.
The key to reverse-engineering the human brain lies in decoding and simulating the cerebral
cortex — the seat of cognition. The human cortex has about 22 billion neurons and 220
trillion synapses.‖
Extracted from Kurzweil: Accelerating Intelligence (Website)
Title: Doubling Time for Neuroscience: 20 Years
Author: Extropia DaSilva (a prolific blogger who writes “about how
technological development in areas like nanotech, biotech, infotech,
robotics and computing may lead to redefinitions of what life is, what it
means to be human.”
Address: http://www.kurzweilai.net/forums/topic/doubling-time-for-
neuroscience-20-years
―Actually, Kurzweil does a lot more than invoke the magic of exponential increases. What he
and his team do is to map progress in a wide variety of scientific and technological fields and
11
how all that diverse research is accumulating and converging on an understanding of the
brain's principles of operation.
An expert in neuroscience is usually an expert in a particular aspect of brain architecture or
functionality. She tends to be focused on this particular area and does not have much time to
be more than vaguely aware of all the research going on outside of her particular area of
interest. As Kurzweil explained:
"There are more than 50,000 neuroscientists in the world, writing articles for more than 300
journals. The field is broad and diverse, with scientists and engineers creating new scanning
and sensing technologies and developing models and theories at many levels. So even people
in the field are often not completely aware of the full dimensions of contemporary research".
Note that this does not mean Kurzweil knows everything every expert in brain science
knows. That is impossible. He lacks fine-grained knowledge regarding this or that aspect of
the brain. What he and his team do is observe the general direction all the research and its
related fields of study are heading in, while lacking the clarity one would have focused on a
microscopic part of the whole structure.‖
12
Conclusion
The following is extracted from Authors@Google (YouTube Video)
The three predictions listed below are from a slide displayed by Kurzweil,
shown to the audience at Google‟s Mountain View, CA Headquarters. The
event occurred on 1st July 2009.
Address: http://www.youtube.com/watch?v=43zo82W7aPI
Kurzweil sees the next major milestone year as being 2029. It will not be the Singularity yet
as the human/machine merger will not be complete until 2045. Nevertheless Kurzweil says
that in 2029:
Computers [will] pass the Turing Test.
Non-biological Intelligence [will] combine.
- The subtlety and pattern recognition strength of human intelligence, with
- The speed, memory, and knowledge sharing of machine intelligence.
Non-biological Intelligence will continue to grow exponentially whereas biological
intelligence is effectively fixed.
13
14