The document discusses the concept of a singularity of evolution, where the rate of evolution formally tends to infinity. It presents 3 potential ways this could occur: 1) through demographic growth, 2) technological advances leading to superintelligent AI, and 3) a general acceleration of biological and social evolution. It warns that accurate predictions beyond the singularity are impossible. The text also examines potential crises for fundamental science like information overload and resource limitations. It questions whether advances like AI could fully compensate for these issues and overcome human intelligence. Overall, the document analyzes challenges for predicting and supporting continued scientific progress.
The Future of Science and the Singularity of Evolution
1.
2. СИНГУЛЯРНОСТЬ ЭВОЛЮЦИИ И
БУДУЩЕЕ ФУНДАМЕНТАЛЬНОЙ НАУКИ
А.Д. Панов, МГУ, Москва, Россия
THE SINGULARITY OF EVOLUTION AND
THE FUTURE OF THE FUNDAMENTAL SCIENCE
A.D. Panov, MSU, Moscow, Russia
3. The evolution is an accelerating process.
The singularity of evolution
is a point of time where the predicted rate
of evolution formally tends to infinity
and simple extrapolations behind this
point are impossible
4. Various ways to the singularity of evolution.
1. The demographic singularity.
The hyperbolic law of the growth
of the Earth population →
t* - the point of singularity
H. von Foerster, P. Mora, L. Amiot.
I.S. Shklovsky,
Doomsday: Friday, 13 November, A.D. 2026
1965
Science, 1960, V.132, p.1291
t* = 2026
I. S. Shklovsky.
The Universe, Life, Intelligence.
1965.
t* = 2030
5. Various ways to the singularity of evolution.
2. The technological singularity.
Smarter-than-human intelligence
→ prediction of future is
impossible
Irving John Good, 1965 -
intelligence explosion
Vernor Vinge, 1988 -
technological singularity,
2005-2030
Hans Moravec, 1988 -
technological singularity,
2030-2040
Raymond Kurzweil, 1990th -
technological singularity,
http://en.wikipedia.org/wiki/File:PPTExponentialGrowthof_Computing.jpg
2045
6. Various ways to the singularity of evolution.
3. General singularity of evolution.
Brain and
humankind
evolution
Graem Snooks, 1996: The evolution of the biosphere and then
the evolution of the humankind is a joint accelerating process
expressed in terms of 'waves of life' with the
acceleration factor ~3.0
7. General singularity of evolution.
Ray Kurzweil, 2001
The Law of Accelerating
Returns.
'Paradigm shifts‘ unite
the biological and the
social evolution into
one chain.
8. General singularity of evolution.
8 phase transitions in
humankind history from
I.M. Diakonov, 1994.
Historical singularity -
was predicted but the
position in time was not
calculated
9. S.P. Kapitsa, 1996
General singularity of evolution. Mustier
Acheul
Chell
Palaeolithic revolution
Anthropogene
A.D. Panov, 2005
All biological points
t* = 2004 y. - all points
t* = 2015 y. - A.D. points
Properties of
phase transitions:
- overcome of evolution
crises (endo-exogenic,
thechno-humanitarian)
- using of superfluous
variety factors
- Sedov's law of hierarchical
compensations
10. General singularity of evolution.
The singularity is not a point -
it is a period of time,
approximately from 2000 to 2050.
The law of all planetary evolution from
the origin of life must be changed
during the period of singularity -
the 'weight' of the present time is
comparable with the 'weight' of origin
of life.
Any exact predictions over the
singularity based only on the
scale-invariant law of evolution
t* ~ 2000-2050 before the singularity is impossible.
11. Examples of post-singular
troubles:
Post-singular evolution Depletion of mineral resources →
closed-circuit production
Depletion of fossil fuels →
+ What may be a base for
renewable energy sources,
predictions? thermonuclear energy
+ The singularity is a region of Environmental protection →
concentration of crises. general humanization,
possible prohibition of
+ If the humankind survives after experiments on any animals,
the singularity, all crises must O
Other prohibitions (web etc.)
be compensated. The rate of exploration of outer
space in XXI has slowed down
+ Numerous of 'compensators'
dramatically compared to the
must be supported XX century →
permanently after the humankind will be restricted
by mainly planetary evolution
singularity point →
during a number of decades
SUCH A LIFE IS NOT EASY! or even more
.
12. Information crisis (S. Lem) and the future of science.
(
(Is the future civilization 'a civilization of science'?)
+ Progressiveness is limited in time We should expect signs of a
for any evolution factor. The law crisis in the science.
of leadership change.
+ The science is a typical progressive
S
Stanislav Lem (1963)
factor of the evolution:
+ Each solved problem bears a number
- Science method is related to industrial
of new unsolved scientific problems →
revolution of XV-XVI century (resolution
o
of agrarian crisis of Middle Ages)
+ Number of problems grow exponentially,
- The ancient mathematics and astronomy – but the number of scientists is restricted →
the factor of surplus variety + There is a lack of scientists to study
- The science became a leader in formation of all actual problems →
the vector of evolution of the civilization + 'Disrupt' of the front of science -
+ Conclusion: the science could not be
Information crisis
eternal leader of the evolution (???)
(
(predicted to the beginning of XXI c in 1963)
+ Sedov’s law → a change the place of
A sort of a lack of resources!
science in the social history in the future.
13. Resource restriction and possible collapse
of
o funding of fundamental science (micro world and cosmos)
During accumulation of the science knowledge about the Nature,
new fundamental knowledge become more and more expensive.
More and more perfect methods and innovations cannot solve the
problem of the cost rising of the fundamental science.
Examples:
L
Larger and larger accelerators of particles (like LHC)
Larger and larger telescopes (cosmic and ground)
L
However the resources (number of scientists, money)
are restricted from the top.
14. Positive feedback loop could produce collapse of funding
Stabilization of the funding
of science implicates reduction Decrease of funding
of the amount of discoveries implicates further reduction
due to increase of mean cost of the amount of discoveries
of one discovery
Positive feedback Collapse of
loop funding
Reduction of the amount of
Decrease of interest of the
discoveries implicates
society to the fundamental
decrease of interest of
science implicates decrease
the society to the fundamental
of funding of science
science
Crisis of loss of interest to the science
16. Predictions of the model
(January 2006)
The absolute funding of science increases
but the upper level is restricted.
The number of discoveries increases
due to funding increasing, but
then begin to decrease due to
increasing of the cost of one
discovery.
There is interval of time where the
funding increases but the number of
discoveries decreases.
The final collapse of funding (near t ~ 500)
is a result of positive feedback loop
17. Number of
papers per year
1817-2010
1817 2010
http://www.wired.com/wiredscience/2011/03/best-science-maps?pid=1052 - from M'hamed el Aisati
19. After 2006
the number of
papers decreases
for the first time in
all science history.
But funding of
science increases
USA science funding
http://www.nsf.gov/statistics/nsf11313/
20. The dynamics of electronic papers
(arXiv, http://arxiv.org/ )
21. Increasing of funding of science
A 'paradoxical' result: increasing of A practical example:
funding implicates more early collapse Freezing of funding of SSC collider
of funding of science with almost the same in USA and particle physics on
total sum of results as for lower funding. circle colliders
Increasing of funding of science 'brings the future closer' and makes it more safe.
22. Could Artificial Intelligence (AI) compensate the crisis of science?
= Information crisis (Stanislav Lem)
= Crisis of loss of interest to the science
= Resource crisis of science
Could AI-robots compensate a lack of number
of alive scientists?
Could AI 'grow knowledge' instead of study of nature
with real experiments?
...........
23. A prediction that AI will
'exceed human mind in
all parameters' is based
The Moor's law and AI capabilities mainly on Moor's law.
1. A question:
Is the estimation of brain’s
rate correct? An amoeba
has comlicated behavior,
but it has not neurons at all.
The amoeba’s thinking is
molecular, whit is the rate?
2. Actually the Moor's law
provide only necessary
but not sufficient condition
for AI to exceed human
mind in all parameters.
24. За прошедшие 15 лет «разум» наших The «mind» of our computers was
электронных вычислительных машин improved million times during the last
улучшился в миллион раз... В течение 15 years... A new improvement of
нескольких следующих десятилетий computer's «mind» no less than a
следует ожидать увеличения number of thousand times more
характеристик «разума» машин еще should be expected within the nearest
по крайней мере в несколько decades. The «mind» of such computers
десятков тысяч раз. «Разум» таких will definitely overcome the human
машин по основным параметрам mind in basic parameters.
будет заведомо превосходить разум
человека.
И.С. Шкловский, 1975 I.S. Shklovsky, 1975
37 years have passed! An improvement about million times since 1975 took place.
Where are the expected computers to overcome human mind?
25. What is a source of mistake in predictions?
The necessary and sufficient
condition for computers to overcome human mind in all respects
is sufficiently fast and powerful hardware (Moor’s law)
together with software that can reproduce human’s mode of reasoning.
But software is much more conservative than hardware.
26. maxima — one of the better systems of analytical computing now.
A classical AI system (heuristic programming).
Was written in 1972, 40 years ago.
Computer power was improved more than one million times.
Many other contemporary systems of analytical computing
have same core.
Microsoft Word — windows version was written in 1989, 23 years ago.
Computer power was improved about 105 times.
No changes in main functions of the Word system up to now.
The main system of documents preparation in the world now.
Computer translators from foreign languages — now are almost so feeble as
at the beginning of 1990th were
Computer power was improved about 105 times after 22 years.
27. Main AI technologies:
•Neural network All are known since
•Heuristic programming late 1950-th - early 1960-th
•Expert systems (more than 50 years no essential news)
•Evolutionary programming
There is hard stagnation in the field of AI programming ideas.
It is unknown what is the human’s understanding.
Nobody know exactly what problem should be solved
to reproduce human’s understanding.
A problem could not be solved if it was not formulated.
28. What is a source of troubles?
•A computer operates with information.
•A man operates with meanings.
•It is supposed by default that human meanings
may be represented in information terms.
•But nobody proved that human meanings actually
may be represented in information terms.
One possible counter-example:
If meanings are represented in brain by quantum states
(not classical bit-like states) than meanings are not represented in
information terms, since such states have not properies of information:
- information is something that can be copied (duplicated)
- quantum states is something that cannot be copied due to
no-cloning theorem of quantum theory.
Quantum state is not information.
Real situation might be (and possibly is) even much more complicated
29. Rodger Penrose’s no-go theorem for AI
Theorem: The rough idea of proof:
Any finite computer system constructed The theorem is similar to the Noether’s
with usage of any known physical incompleteness theorem.
A finite computer system is described
principles cannot reproduce some
like a finite system of ‘axioms’.
special mathematical capabilities of Then there exist Noether’s propositions
a human mind. that may not be deduced in the
system but can be understood by a
Corollary 1: human mind.
Operation of human mind is based on
unknown physical principles
Nobody could predict when and whether
Corollary 2: the new required principles will be
To reproduce or overcome human mind discovered. Therefore nobody could
predict when a computer could
with computer one should discover some overcome human mind.
new physical principle(s) and construct a
‘computer’ system based on it. The collapse of fundamental science
funding might prevent the discovery
Penrose’s hypothesis: new unknown of this new principles at all.
principles related to quantum gravity.
30. Contemporary direction of evolution of AI is not at all one
to overcome human mind in all respects.
Rather, the actual direction is to integrate the
humankind into one unit information system.
AI is only an instrument in man’s hands in this context.
The acivements in this direction are actually huge:
- online-communications and virtual societies
- fast search and indexing of information in the web
- direct network democracy become compete with
usual representative democracy
“Towers and the Moon” metaphor
31. My private opinion:
AI alone may not overcome possible crisis of science.
It may be used only as a kind of an instrument.
However, I can’t propose any absolutely firm way
to overcome this crisis…
But there are a number of other possibilities
to discuss in this respect.