AlphaGo Proves A Few Steps Forward, But AI Still Poses No Threat

The recent news that Google’s AlphaGo program beat Go champion Lee Sedol brings up the most common questions, unfortunately even with people who’ve never played the game, which is likely all of Western civilization.

The Inquisitr covers the basics of the beat down a visibly shaken Lee Sedol took earlier in the week.

But the Economist further examines the showdown between man and machine as a leap forward in artificial intelligence (AI), whether AlphaGo won or not.

The article states that the traditional stone and board game is a hit in Japan and Korea, much like chess is in the Western world, but it also makes the comparison between how a program differentiates moves depending on preset filters accumulated by previous games with the value of those chess pieces and how AlphaGo approaches its moves using two algorithms: the policy network and the value network.

AlphaGo artificial intelligence possibilities
AlphaGo's brain could easily resemble this if it was all on one board. But the technology simply isn't there yet. [Image by By Tej3478 (Own work) [CC BY-SA 4.0], via Wikimedia Commons]

For details of these learning networks AlphaGo uses, a paper written by the software creators is available on Nature, and it shows the very real advances of AI without having to know how the game is played which the article by The Economist simplifies.

To the more common questions however, which stay within the realm of the hypothetical of whether machine will ever defeat man, while the fictionalized worlds of movies and books continue to advance towards that goal, the reality appears to be the complete opposite.

At the very worst, the threat a win by AlphaGo might consider is more on the field of workforce than war with the machines where we use these programs to our advantage, which has already been utilized with similar AI programs like IBM’s Watson.

To be more specific, University of Sheffield professor of machine learning Dr. Neil Lawrence does provide some insight with a question and answer session he does regularly via Quora.

AlphaGo puts the world's best minds to use.
University of Sheffield where Dr. Neil Lawrence professes about machine learning. [See page for author [Public domain], via Wikimedia Commons]

In one session related to the workforce, Dr. Lawrence offers insight into the evolution of the workforce with machine technologies.

“With machine intelligence, as we do start to better emulate various characteristics that *today* are considered distinctly human, I think our the way in which we connect with other humans will evolve. You hear people talking in rather simplistic terms about ‘productive’ jobs being done by machines and humans becoming focused on ‘entertainment’. But those terms are rarely defined satisfactorily: what is a ‘productive’ job and what classifies as ‘entertaining’? My own job often feels like both, but sometimes feels like neither. Maybe I’m too optimistic, but I hope we’ll evolve to better understand, respect and enjoy each other and that the work we find satisfying will evolve to build on that understanding.”

Dr. Lawrence recently retweeted a tweet by Andrej Karpathy, a CS PhD Stanford student who specializes in OpenAI and deep learning, about AlphaGo, which comes to show he’s obviously paying attention.

But the article by the Economist also explains that the AlphaGo learning process is similar to human learning, by mimicking or familiarizing itself with human moves — though the moves on the board would require a different analysis — which is processed by the value network.

Thus far the record AlphaGo has in winning is promising yet slow as this is not the first time the program has “had a go” at Go as it did play against and beat European Go champion Fan Hui in October of last year, who was at a lower rank than Lee Sedol is.

In order for AlphaGo to continue to beat players — which it may not have to do anymore — the processing power in both hardware and software would have to be maintained as it takes 1,920 standard processing chips and 280 special ones, aside from special coding in the software.


An article published on The Future Of Life Institute website provides more specifics on the CPU and overall hardware power required to operate a program like AlphaGo.

If the sentiment the LA Times shows is the shock and expectation anyone had that a machine would ever learn the game well enough to defeat the 9 dan champion, then perhaps those expectations could use better guidance, as AlphaGo was running on the bare minimum when compared to Watson; if Sedol had not made the one mistake he claims to have been the cause of his defeat, Sedol could have had an extra million in his bank account.

At the very least, this latest news about AlphaGo will certainly entice some proactive people to play the game themselves.

In comparison with IBM’s Watson, AlphaGo’s hardware is minimal and clearly isolated to the game itself; although if one could imagine the marriage of the two, perhaps even a monopolized AI industry would result in the sum of all fears.

[featured image by Lee Jin-Man | AP Photo]