Artificial Intelligence and other Myths

AI Costs - KISS

Humans have real intelligence. Computers, on the other hand, can only follow a program of instructions. Computer programs are written by humans and can only impart a small fraction of the human programmer's skill into the resulting program run.

The advantage that a computer has is the penchant for working full speed non-stop without getting the human feelings of boredom and fatigue.

I believe that computer programmers of today have the ability to make computers do almost anything that humans can do - but at a cost. Part of the cost is 'bloat' other cost is complexity of the human interface and the balance of the cost is reliability.

Bloat can be seen in the hardware requirements of modern operating systems and applications. They either don't run at all, or run extremely slowly on older hardware.

Human interface complexity is also a necessity in order to communicate to the computer what your requirements really are. Compare the keyboard of a standard calculator to that of a financial or scientific calculator.

Reliability suffers with complexity both due to the greater probability of program bugs as well at operator errors.

The bottom line is that computers should not try to replace human intelligence. The old computer adage K.I.S.S. (Keep It Simple Stupid) applies as much today as ever before.

Non-stop pushups

A few months ago I showed photos of my coral tree saplings in late afternoon and early evening. See related link.

Now they are bigger, and winter is approaching, but they are still flapping their leaves.

This video shows three days of action taken at one frame every ten minutes.

Notice how the leaves start rising before two o'clock in the morning, hours before it starts getting light - indicating that it is controlled by time rather than light. Yet if artificial light is turned on while the leaves are moving down, they start to rise - showing that there is a degree of light sensitivity. During the day the leaves appear to be shaking due to breeze - but it is dead calm and at normal speed no movement can be detected. Night images are iluminated with infra-red.

Real AI

A "robot modelled after Albert Einstein with specialised software that allows it to interact with humans in a relatively natural, conversational way". Click on related link, below, to see full story and video.

I hope it is as good as the animated Abraham Lincoln that I saw at Disneyland in 1966. Hmmm, not likely - Abe had a full body and stood up, gave a speech, and sat down again. I was convinced that it was a human actor. A detached head and shoulders won't fool anyone!
.

Want some money?

Read good science fiction stories to get good ideas for inventions.

Research (hint: Google) other research in related ideas to gain ideas for how the sci-fi inventions can be realised.

Write a proposal, adding a liberal sprinkling of well-known theories such as nano-technology, quantum mechanics and biological neurons.

Make a model. Photograph it and call it a prototype. (It can be the wrong size and need not work, but it must look mysterious.)

Request research funding from any of a large selection of agencies. (hint: Google). DARPA may be a good one - see related link.

Funding grants for impossible research, such as Smart Dust, are a waste of money. But it does entertain many people who watch Discovery Channel! It is surprising how many people really believe the futuristic inventions may one day become practical.

Research Fraud

The New York Times of 29/12/2008 reported "Universities to Pay U.S. $1.6 Million in Research Fraud Case".

I think that this is the tip of the iceberg and many more research projects that are being publicly funded are just Sci-Fi ideas that are made believable by fraudulent claims.

Univertity professors should be teaching the students useful skills and not wasting time playing Sci-Fi games with expensive toys funded by research grants.

Mellon with no Intel-ligence

The Carnegie Mellon University, in collaboration with Intel, is doing research into what they call Claytronics.

This project combines modular robotics, systems nanotechnology and computer science to create the dynamic, 3-Dimensional display of electronic information known as claytronics.

Our goal is to give tangible, interactive forms to information so that a user's senses will experience digital environments as though they are indistinguishable from reality.

Claytronics is taking place across a rapidly advancing frontier. This technology will help to drive breathtaking advances in the design and engineering of computing and hardware systems.

Realizing the vision of claytronics through the self-assembly of millions of catoms into synthetic reality will have a profound effect on the experience of users of electronic information. This promise of claytronic technology has become possible because of the ever increasing speeds of computer processing predicted in Moore's Law.

The axiom that they are building on is Moore's Law; an observation that since the invention of the integrated circuit in 1958, the number of transistors that can be placed inexpensively on an integrated circuit has approximately doubled every two years. The problem is that Moore's Law is not an axiom based on any physical or mathematical law.

I believe that Moore's Law has already stopped working, in spite of what Intel may say - or rather in spite of what Intel do not say. Take microprocessor speed as an example. Pushing speeds beyond about 3GHz increases power consumption exponentially. Microprocessor manufacturers are working around this barrier by making multi-core processors - essentially more than one relatively slow microprocessor sharing the workload. For multicore processors to show any improvement over the older, single-core, processors, software must be specially adapted. As a result, older software runs just as fast, or faster on older computers as it does on new multi-core ones (especially if you do a fresh install of Windows).

Only the transistor count per chip is still increasing in line with Moore's Law and will soon make mechanical hard disk drives obsolete. But memory isn't everything. Overall size is the real measure that the researchers at Carnegie Mellon should be looking at. According to Moore's Law, size for equivalent power should be halving every two years. since my first cellphone in 1994, size should have halved seven times, and that would mean a cellphone less than a millimetre in length (with a relatively huge antenna).

Carnegie Mellon are welcome to sue me, but I say they are making fraudulent claims that their technology is realistic. In fact the more I think about it the more it looks like someone's April Fool joke that is persisting years later - it even made it to Discovery Channel :)

The October 2007 issue of Pittsburgh Magazine has a more complete story under the title "Sci-Fi". At least the title of this story is honest.

R/C Flying Simulator

I recently bought myself a 4-channel radio controlled model helicopter. After a few minutes of unsuccessful attempts to control the model I became convinced that my brain was not wired correctly to be a helicopter pilot and I was sure that I would totally destroy the machine before I could get any fun out of it. That's when I decided to spend a bit more and buy an E_Sky flying model simulator.

This controller comes with freeware software that surpassed my highest expectations. The simulation is so accurate that it made me feel like I was controlling the real thing. After a few hours of fun over the next week I managed to master the basic skill of flying a remote controlled helicopter. I crashed more often than not, but just simulated.

Now, thanks to the simulator training, I am now getting full enjoyment from my real model helicopter and I still use the simulator to practice more advance flying skills. Hobby shops should rent or sell these simulators as an essential pre-requisite for every novice R/C pilot. Or for wanabe R/C pilots who cannot afford the real flying models - fixed wing or helicopter.

Fantasy Research

I always read news articles in technical magazines. Some are both believable and have potential to be commercially viable, but usually the researchers at universities make announcements that are so weird that if I was directly involved I would only announce them on April Fool's Day.

Here is one that I read today:

A team of researchers from Rensselaer Polytechnic Institute have developed a unique type of camera lens, composed of a pair of water droplets which vibrate upon exposure to sound, thereby changing the lens's focus. Although liquid lenses are not new, this system doesn't attempt to maintain focus the way other designs do. Instead, it captures images in a continuous stream. Because the lenses are constantly vibrating, some frames are in focus and some are not, and advanced imaging software ignores the blurry frames and retains the in-focus ones. Consuming less energy than traditional lenses and capable of capturing 250 pictures per second, the technology could find its way into systems such as cellphones, automobiles, miniature spy planes and autonomous robots.

What hints that the story is flawed is that it "could find its way into automobiles" - ha ha, cars don't use camera lenses.

Obviously students are just doing this sort of 'research' to earn their qualifications - but why do the professors allow such futile projects? Well, it turns out that it is the professors that are doing the research and it was U.S. National Science Foundation that funded the project. Go figure!

Less is more

I have been a radio ham since my teens, but never thought of this until my daughter asked me today; "why is long distance radio called short wave?". Well, in radio broadcasting, 'Long wave' is used for medium distance, 'Medium wave' is used for short distance, and 'Short wave' is used for long distance. (Actual distances depend on factors that are seemingly totally unrelated to wavelength, such as sun spots and the time of day.)

And, while the longer waves are use for small amounts of data transmission, it is the 'Micro waves' that can carry huge amounts of data.

It's the same with computers - the smaller they get the more powerful they can be. Science can be so mystical! I love it.

Humans vs Machines

I realised today that artificial intelligence can be far more intelligent than human intelligence. But not because AI is humanlike.

The fact is that humans are not all that intelligent. Humans think relatively slowly and make many mistakes. Humans also suffer from delusions, paranoia and other psychoses. Humans tend to 'think' with their hearts rather than their brains. If humans were reasonably intelligent we would all agree with each other, there would be no arguments as to how best to spend money or who has the best political policies. There would be no civil unrest and no wars.

To prove the point, Vodacom recently rolled out a "Content Adaption" system. CA has the job of "reformatting web pages on-the-fly to fit better on small screen devices (SSD) and thus give an easier browsing experience."

Further quoting an article by a Vodacom spokesperson:

The idea is to seamlessly take the HTML stream between the SSD and the web site and reformat the layout of the page to fit better on the SSD.

Especially, it should make navigating the site easier by removing the need to pan all over the original page.

It will also reduce the amount of traffic flowing down to the SSD as the CA takes place on the Vodacom network and only the resultant new, small screens are sent down to the SSD and are counted as part of your data use.

To do this, the SSD must first connect to an intermediated box where the CA will take place. This box will then connect to the requested web site on the internet, retrieve the web pages, re-format them and then pass it down to the SSD.

This concept of implementing intermediate, in-line devices in the data stream to perform specific functions is well known and employed by all ISP's. Examples include Firewalls, Virus Scanners, Caching Engines and Proxy Servers. These boxes all perform some important function, either to improve the customer experience, reduce network traffic or both. The CA engine is just another of these in-line engines albeit with quite a complex task; trying to think like a human!!

I believe that they succeeded admirably in "trying to think like a human". It was a big mistake. They should rather have left it alone.

A survey on the mybroadband.co.za web site showed that only 1 out of 35 respondents found the CA system to be successful.