Artificial Intelligence: The second sixth day - Part 2

Its Taking more than a day

God may have created man in one day, but its taken us hundreds of years to get to the present state of AI. And even then, todays applications are impressive but primitive compared to organic intelligence and limited in scope.  

AI really took off with the explosion of computing in the mid-twentieth century. But its been a slow plodding progress, quiet unlike the developments in other areas of computing eg micro-processor development. 

Early optimism fell into frustration and cynicism. But AI's development was held back mainly by the lack of sufficient computing power. Once computing power became powerful enough AI progressed rapidly.

Lets chronicle this journey - atleast the modern part of it;

• 1956 Allen Newell, Herbert Simon, and Cliff Shaw wrote a computer program which they named the Logical Theorist. It went on to prove 38 mathematical theorems by exploring a search tree, the root of which was the initial hypothesis and each branch was a deduction based on the rules of logic. It used 'educated guesses' [heuristics] to eliminate unlikely search paths and prevent the search tree from growing exponentially.

• In 1959 Arthur Samuel an engineer at IBM programmed an IBM 701 computer to play the game of Checkers (a.k.a Draughts). His program implemented a scoring function that assessed the board position, allowing it to determine its chances of winning from that board position. It also used Alpha-Beta pruning to reduce the chance exploring 'bad moves' and focus on better-more optimal moves.

Arthur Samuel plays Checkers against the IBM 701

• From 1966 to 1972, researchers at Stanford Research Institute worked on a robot named Shakey. The Robot, that was capable of reasoning out its own actions to achieve a goal. Shakey combined research in robotics, computer vision and natural language processing.
• Researchers at Carnegie Melon developed a program that used knowledge of human experts to determine what components a computer should be made from to meet a clients requirements. This was a classic example of an Expert System - popular AI agents in the 1970's and 1980's. Digital Equipment Corporation (now defunct) used this software to great effect.
• Perhaps AI's most famous historic milestone was IBM's deep Blue just barely defeating Gary Kasparov in a six game chess match in 1997. Its worth mentioning that Deep Blue had lost their first match a year earlier.

Deep Blue Main Frame

•  In  2005 a completely autonomous vehicle programmed by a team from Stanford University navigated a 212 Km desert off-road course to win a $2million prize in a Pentagon AI competition. The complexity of the challenge proved that AI and the fact that three vehicles completed it showed that AI was finally starting to deliver on the promises we had hoped for 50 years earlier. The vehicles navigated deserts, salt pans, rough terrain - all the while judging how to get over and around obstacles while staying on course - without any human help!
• Decades of the high expectations of AI being let down had lead to pessimism about developments and the future of the field. For example researchers believed the Asian board game of Go was way too complex for current technology to defeat a human Grand Master. But advances in machine learning had outpaced expectations and in 2016 an AI agent named AlphaGo developed by UK based company Deep Mind trounced the reigning world champion Lee Sedol 4-1 in a five game match.
  

AlphaGo takes on Lee Sedol (left)

 These are just some of AI's historic highlights, but only some of many many highlights - and lowlights too.

 

Now this may all seem very geeky and for a bunch of nerds - but . . . AI is the future. It currently underpins most of our digital lives from communication to online search, social media, e-commerce, share/bond trading, politics, transportation etc.

AI will be part of more and more of our real lives soon. Cars will drive themselves, planes will land themselves, cyber AI agents will control weapons systems and make kill/no kill decisions on the battle fields, AI doctors will treat illness, administrative Bots will formulate public policy, predict criminal activity etc.

You need to care about AI. Care about what part it will play in your life and what controls must we place on it - because we are creating something much much smarter than we are. . . .