So, you want to learn AI …

School daze.  Learning new technologies in late middle age is like playing the old greased watermelon game.  The game was played at a swimming pool by rubbing Vaseline all over a whole watermelon, tossing it in the pool and telling the kids to jump in and wrestle it out of the pool.  The resulting mayhem and general buffoonery were priceless.  There were plenty of theories as to how to get the watermelon out of the pool.  Pull it underwater and swim underwater.  Push it toward the ladder and carry it up the ladder.  Squeeze the watermelon between two kids.  While no best practice was ever discovered the watermelon did always end up coming out of the pool.  I suspect the same will be true for learning AI without completing a master’s degree in Computer Science.  Mistakes will be made but learning will occur.

Learning by writing.  This blog is testament to the idea that a person can make some progress in leaning a new area of technology by writing about that area of technology.  You not only have to read enough to learn something worth writing about you also have to be able to explain it to others.  Reading loads the gun.  Writing pulls the trigger.  Unfortunately, the resulting intellectual weapon is more of a squirt gun than a Sig Sauer.  Getting real traction with a new technology area can start with writing but has to progress to building solutions.   But where to start?

When in doubt search it out.  A quick query of “Learn AI” in the Google search engine seemed like a good place to start.  The resulting 133 million results begs the question of whether one needs AI to sort through the ways to learn AI.  A few hours of link chasing surfaced some patterns.

Just do it.  One school of how to learn AI involves picking a problem that interests you and building an AI based solution to solve the problem.  Ben Hammer, co-founder and CTO of Kaggle, put forth his “Eight Easy Steps To Get Started Learning Artificial Intelligence”.   Needless to say, several of the steps involve Kaggle.  This seemed a bit like teaching a kid to swim by throwing him into the deep end of a pool.

Audacity.  Online educational outfit Udacity offers a free Introduction to Artificial Intelligence course.  It requires intermediate skills and will take 4 months.  Anytime something substantial is offered fro free Spider Senses should start tingling.  The prerequisite to the class is an Intro to Statistics course that requires beginning skills, is also free and will take 2 months.  Coursera also offers a Machine Learning course.  There is a free option and an option for $79 whereby you get a certificate at the end if you pass.

Water water everywhere.  There seem to be multiple good ways to get started learning AI.  Updates from the attempt to teach an old dog new tricks will be posted as the educational process unfolds.

 

 

 

Who’s who in the AI zoo?

Teach a man to fish.  The old adage holds that if you give a man a fish you’ll feed him for a day but if you teach a man to fish you’ll feed him for a lifetime.  Researchers from Universidad Politécnica de Madrid and University of Florence have a new spin on the old saying … build a robot fish to tend to real fish and there will be more fish for everybody.  Science Daily is reporting on a “New Robotic Fish for Environmental Monitoring  Given that aquaculture furnishes about half of the seafood eaten the researchers wanted to create a environmental testing tool to ensure that the farm raised fish, crustaceans, etc live in clean, healthy water.  So, why a robot fish instead of more traditional monitoring probes?  “In order to minimize the inconveniences and possible stress in fish, the developed robot is bio-mimetic, that is, that mimics both its appearance and its functioning.”  One suspects that it’s also a lot more fun to build robotic fish than monitoring probes.

Pass the Cheetos.  Chester the Cheetah now has some competition.  Science Daily is reporting that engineers at The University of Twente have created a robotic cheetah.  Geert Folkertsma has dedicated four years of research and development to constructing a scaled-down robotic version of the fastest land animal in the world, with a view to replicating its movements.  Geert explained his interest in CyberChester by saying, “My robot vacuum cleaner, for example, cannot climb stairs or even cope with thresholds. We therefore need to develop robots that can walk and when it comes to moving around efficiently, there’s a lot we can learn from the cheetah.”  Hmmm, true ‘dat, Geert.  Although it’s early in the development cycle the mini-cheetah operates more like a mechanical sloth than the fastest mammal on Earth.  It runs at one kilometer per hour.  At that rate the mechanical cheetah would need about 1.25 years to walk back to its ancestral home in Africa from its birthplace in Enschede, Netherlands.    For now, the wildebeests migrating across the Serengeti Plain will be safe from robo-Cheetah.

 

Donald J. Rippert

AI – Nirvana or Apocalypse?

Smells like tech spirit – Artificial Intelligence may be on its way to becoming the buzziest buzz-term in the buzzword laden history of the buzz-o-sphere.  No prior trend has engendered the societal debate that AI has sparked.  Scientists, billionaires, politicians, poets, priests, butchers, bakers and candlestick makers have all gotten into the game.  Ok, the candlestick maker reference was hogwash but give that industry time … I am sure something will come up.  Everybody has an opinion and the opinions are “all over the map”.  Artificial intelligence will either be the recreation of Eden on Earth (without the troublesome snakes and apples) or the kind of zombie apocalypse that gives zombies nightmares.  The goal of the re-launched Smart Ubiquity blog is to examine the technical, economic, industrial, governmental and societal impacts of AI.

“I’m sorry Dave, I’m afraid I can’t do that.”   Concerns about computers getting too big for their britches go back a long way.  Generation after generation has had their fears of computer overlords generally mucking things up.  My generation first learned the perils of artificial intelligence from HAL of 2001: A Space Odyssey fame.  Thirty three years later everybody laughed when 2001 came and went without any psychotic computers in evidence.  But here we are 16 years later and there are some very serious people with some very serious concerns.  Why did concerns about AI go from the realm of entertainment to a serious debate about the end of mankind?

The winter of their discontent.  AI has gone through a series of boom and bust cycles over the decades from the hype of the 1970s and 80s to the so-called AI winter from about 1990 through 2011.  In some ways the public’s fascination with AI elevated the highs and made the lows all that much lower.  In 1981 Japan’s MITI funded the Fifth Generation Computer Systems project with $850M.  The ambitious program would build a new generation of computers designed for AI along with the AI software needed to make the dream come true.  An impressive list of goals was drawn up.  Ten years later the goals had not been met.  Twenty, even thirty years later many of the goals from 1981 were still elusive.  Then, in 2011, came one of those bizarre occurrences that sort of change everything.

Your answer must be in the form of a question.  In January 2011 IBM’s AI platform, named Watson, played Jeopardy! against the two best human Jeopardy! players in history and beat them soundly.  The AI winter was over.  In reality, AI research had been going on at IBM and elsewhere during the so-called AI winter.  The research was often called something other than AI during the AI winter because of the stigma AI had developed.  However, it was AI.  The Watson Jeopardy! match put AI back in the public’s imagination and it’s been “off to the races” ever since.

The Last Question.  Google followed IBM with a more impressive AI demonstration.  In 2016, using its Deep Mind AI platform, Google defeated the reigning human Go master.  Go is a 3,000 year old Chinese board game that has been notoriously hard for AI platforms to successfully play due to the mind-boggling number of possible moves.  These advances, and many more, explain why the debate over AI and the future of mankind has reached such a fever pitch.  AI is finally real.

Future columns will discuss the nature and timing of AI’s progress but it seems that Isaac Asimov’s Multivax and The Terminator’s Skynet are both reasonable proxies for the future of AI and human civilization.  One is nirvana and the other an apocalypse.  But, which one will it be?

 

Donald J. Rippert

Author’s note:  I am an employee of the IBM Company.  However, on this blog I write as an individual rather than a representative of my employer.  All opinions are my own and do not necessarily (or even probably) represent the position of the IBM Company.