IT: Status and Outlook of AI

Forbes recently had a nice collage of articles on AI and the status of AI :
http://www.forbes.com/2009/06/22/singularity-robots-computers-opinions-contributors-artificial-intelligence-09_land.html

One problem with many critics of AI is that they overestimate the "Human Understanding"! And they are not fair to computers - finally, also we humans do not understand some words or jokes or some CAPTCHA graph (intended to inexpensively differentiate humans and simple computers).

We have to see that some appliactions (e.g. numeric computing, storage and storage-based applications as, e.g., Google) directly benefit from Moore's law and grow exponentially, others only much weaker and grow only linearly in power (as, e.g., speech recognition and natural language understanding).

Therefore, Computers can often directly solve problems by "brute force", e.g. in translation by learning from all accessible sentences Chinese-English. In this Forbes series, Peter Norvig from Google describes this as "unsupervised learning": Given a sufficiently large base of material, e.g. master chess games or spoken English, the computer solution becomes great! The computer is in many cases by orders of magnitude better in pattern finding and association detection!

But orthogonal to this success is improvement by method i.e. in software:

Every software programmer knows that the performance of a simple approach and a better programmed approach can differ by many orders of magnitude even in elementary tasks!

The Turing Test will go the same fate as other negative predictions e.g. "computers will never be able to drive a car" (I remember hot discussions on this): The issue will just become meaningless and uninteresting because it will be obvious that computers can. Then the next human task will get in the focus, e.g. (some degree of) creativity or biomimetic personal robots as Kevin Warwick explains in the parallel article. And finally, there will not be "natural" and "artificial" intelligence, just intelligence - if you use some paper and pencil to make notes and support some thoughts, you are not ashamed either.

Private: ALS and Dog Therapy

There is nothing positive to report from the ALS disease of my wife - it is just strugglng with more and more problems. Now she cannot even stand any more, I have to carry her. It is now very difficult to put clothes on etc..

One highlight was yesterday: our friendly neighbor visited us with the wonderful huge white dog Don Merlin. This Pyreneen shepherd dog is so friendly and beautiful - this was for my wife (and me) a great highlight! Don Merlin is a therapy dog mainly working with disabled children.
A wonderful creature - hope to see you two again soon!

Physophy: The computer science paradox - negative view and less interest, more important

Computer science (I prefer the European term "Informatics") has a strange position in science and society: Everybody agrees that IT is ubiquitous -
but Informatics is less and less visible in the society:

Where are big inventions of computer science visible? Internet and computers are no real computer science results (maybe Google's search is ?).

And Informatics is less and less appreciated: See also the decreasing number of students!
Just saw a note by a scientific colleague working on history and future of science: Computer science is not on his radar - "it is just some auxiliary tool".

Indeed, for many practical purposes, the current IT progress and the professional work done by IT people for business is commodity:

the same ideas as 30 years ago, new trials, and cost reduction is the main driver!
There are people stating that computers anyhow have only two hand full of ideas which are repeated over and over since 30 - 40 years, e.g.

• virtualization,
• caches and working sets,
• hierarchies,
• distribution of control and work (local - remote),
• componentization and integration,
• simple structures (no goto's, today no threads) .

Three more negative arguments:

Informatics builds large systems - in order to do this, it must be strictly and simply structured; this implies a lot of repetitive work (it does not help that this is automated - automation is just lifting the level, the simplicity at the human interface has to remain by definition).

For many applications, IT is under the surface, and the applications are in the foreground (and are visible, and in a company will earn the career).

IT is so popular and fast progressing, that a large part of the pragmatic progress is known to almost everybody who wants or needs to, computer scientists or laymen, - with a relatively low entry level (cp. this to quantum physics, for example!): the professional advantage is often small.

But apart from this unappreciative economic and social role, Informatics and IT become more and more fundamentally important: Informatics and IT are the science and the engineering discipline to organize every work done in society, and because all changes in nature can be described as ongoing work, all nature.

The scientific importance of Informatics cannot be exaggerated:

• basic infrastructure of economy and society (wait for personal robots!),


• new pillar of science and engineering (repeating nature in the computer)


• continuation of the biological evolution:
  Don't forget - life is software, evolution was and will be informatics.

Even the connection between information and physics is not satisfactorily understood - and no limits of IT systems are visible: Informatics builds ultra-large-systems and larger.
But the science behind this is just in status nasciendi!

Therefore in daily life, IT is just infrastructure - but it is also the infrastructure of the human future!

Physophy: A Scientific Hardness Index

Recently, I have read in a blog on a conference on Pseudosciences:
"Nothing against pseudosciences pls - Einstein's theories are probably also pseudosciences" - and I was shocked. (The post referred to the effect "you think about a person, and then he or she calls you", and claimed that this is causally connected).
In my understanding, we have roughly three classes of systems (or sciences):

1. Superhuman systems as, e.g. fundamental science, e.g. particle theory -
   no individual could produce it alone. Mathematics (or computer software of large complexity) is the glue,
   
2. Human systems of varying stability and hardness (established or under construction),
   
3. Pseudosciences (besides science or in contradiction to science).

To have a simple index, I propose to have a hardness scale, similar to the Mohs hardness for minerals, with:

• positive values: scientific statements
  
• zero: neither scientific nor obvious nonsense
  
• negative: non-scientific (para or pseudo).

This gives my proposed scale "Scientific Hardness Index" from +3 to -3:

+3: Fundamental science (superhuman), e.g. fundamental physics

+2: Science established, probably high precision, e.g. astronomy, evolution

+1: Scientific theory under investigation e.g. extended longevity

0: Neutral -neither scientific nor obvious nonsense e.g. visitors from other stars

-1: Beside science but not hard contradiction (e.g. astrology)

-2: Hard contradiction (e.g. predictions, telekinesis)

-3: Proven wrong (or obsolete) (e.g., "earth is a hollow sphere")

These numbers cannot show the tremendous nonlinear difference in the system strength of these levels: I would like to compare

• a level 3 - area with a world class building with concrete and steel,
• level 2 with a solid house,
• level 1 with a play with Lego blocks to try what is fitting,
• level 0 with a heap of cotton balls,
• negative levels just with rings of smoke (more or less hazardous)

I know, this will not end the discussion what is pseudoscience and not, on the contrary, - but it should make clear what one is talking about, about a 800 m height modern construction or about a heap of cotton balls (or mozzarella?) or just about smoke (unfortunately, as you know, you can earn a lot of money just with smoke and cigarettes ...).

Physophy: Evolution continues dramatically with IT

Biological evolution meets IT evolution (the flat blue arrow from the right) in this figure:



A hard consequence of evolution is a major change in the vision of human history:

Look at the historical, clear picture as, e.g., claimed by the Abrahamitian religions:

First, a world is created with a hard and fixed hierarchy of species (the humans are kings and queens and can do with the lowers almost everthing they want - and they do or did),

second, the world runs for a certain time to separate the good from the bad, and

third, the world stops with extinction and nemesis and paradis for the selected.

With evolution, the world has been created in a big bang as a basis (if you like), and then a machinery for evolution has been generated - and with this machinery, there is no reason that evolution will stop (with us "original" humans as final state).

Apart from other (more conventional) driving forces for evolution, we are in an epoch where IT technologies explicitly intervene evolutionarily and accelerate evolution:


Hitting the individual or the population altogether are, for example:

1. Understanding genetics and epigenetics (genetic engineering),


2. Enhancing humans with neuro-devices and more (e.g. neural implantations),


3. Virtual lifes with realistic minds merging to real (and competing with real life),


4. Real virtuality with personal Robots (e.g. competing with real human partners),


5. Social communities allowing collaboration,


6. Mind sharing technologies starting to merge individuas and IT.

In the figure, the evolution in biology up to now ("in carbon") and the evolution in IT ("in silicon", at least for the time being) are joining and opening a spectrum of channels for evolution, from close to IT to close to flesh and blood (in the figure, the green area): It seems that we take all of them. Nobody knows the overall outcome - this is the singularity! And we are not sure that the resulting evolution will be guided (or "intelligent") - neither biology nor IT was "guided" although the latter was even made by humans.

As Ray Kurzweil states: "Life is Software" - and when we are able to change the software of life, improve (?) or add, we perform evolution. Given the many influences and influencers of the coming evolutions made by humans, it is probably again no "intelligent design" - it will be non-intelligent, I am afraid, but it will hopefully be successful whatever this means.

Innovation: Déjà-vu, vu jà dé and jamais vu

A futuristic picture from 2002: A wearable computer with communication ...

Two annoying points on Innovation:

• Innovations take often such a long time to succeed, e.g. 15 years,
• and new people think that this is brand new.

The idea to write this blog post was from the LinkedIN forum "Greater IBM" where younger colleagues show that they think "this is new" or "this should come" - and its already here, and waiting for success. The specific innovation discussed there is the open mobile phone wave - now the mobile phone finally becomes a regular computer. This is correct and important, but it is not the first time:

• starting about 1990, the [wintel] computer became mobile as "wearable computer".
  Xybernaut was the leading company (I had once dinner with the CEO).
  But the product remained a niche, no success, bad ending of the company.
• about 2004/2005, IBM Research had a modular mobile computer "Metapad" (with touch-sensitive screen). A start-up trying to distribute the Metapad failed.

Now technology and context (mobile device penetration) and entrepreneurship (Steve Jobs) make the open mobile computer Me.C possible ...

Another long-running innovation are RFIDs, starting e.g. with Paul Moskowitz' United States Patent 5,528,222 - filed 09/09/1994 . And I read today "70% of RFID projects fail" - although not for technical reasons! And RFIDs (and their variants) are still futuristic for many people!

Therefore the title of this post on innovation and these notions:

• Déjà-vu something new with the feeling already been seen
• Vu jà dé something old and often seen which suddenly looks different
• Jamais vu something simply new

Physophy (and IT): Evolution as a software system

Evolution is . . . Evolution is the main part of modern biology, and one main part of modern biology is bioinformatics! The core of the evolution is the development of a software system (based on genes and proteins etc), but function-wise it is a compact, special, adaptive software system with controlled adaptability:
This software orders the detailed construction of organisms.
What is missing today, is an analysis and modeling of evolution as a software system!
There are several aspects of commercial and industrial (human-made)software which are helpful in understanding, e.g.

• it is easier to reuse running code than to develop new,
• similar: reusing established subroutines allows hierarchical structures,
• flexibility on the code level requires building blocks,
• flexibility on the execution level requires configurator capablity.

But you need more (and most of it is new and non-trivial):

• preparation of self-modifying code
• close interconnection between program and storage
• mechanisms for easy reconfiguration,
  adaptive to "more or less mutations wanted",
• i.e. management of the "mutation distances".

The "mutation distance" determines the probability of again useful software after a random software mutation. What is needed are software engineering models how to achieve these functional features - and then see how much is directly implemented in genomics, proteomics and transcriptomics.

This software engineering science is the engineering discipline of biology - a wonderful field of exciting and relevant research, much more scientific than ruminating the (of course also important) human software generation processes...

Many people are not aware that evolution takes place on the software level, not on the level of visible features (phenotypes)!

“I made this program longer than usual because I lack the time to make it shorter.” - paraphrasing Blaise Pascal (after softwarecreation.org ).