Is Go a solved game?

No way! As an avid Go fanatic myself, I‘m here to tell you the ancient game of Go continues to confound even the latest AI advances. Don‘t get me wrong – computers are now playing at superhuman levels, beating all but a few of the world‘s best professional Go players. However, despite the excitement over AlphaGo‘s historic defeat of Lee Sedol in 2016, solving Go perfectly remains firmy out of reach.

Let me explain why this deceptively simple game has resisted solution for over 2,500 years!

The mind-boggling complexity of Go

Go is played on a 19×19 grid, just like your average spreadsheet. Easy enough so far? Well, check this out – with each player having around 250 possible move options per turn, and games averaging over 150 moves, we‘re looking at 10³⁶⁰ total possible positions! That‘s more game states than there are atoms in the entire frickin‘ universe! Talk about some crazy complexity. Chess clocks in at a measly 10¹²⁰ by comparison.

So what does this mean? Basically, brute forcing solutions through raw computation is a non-starter, even with the beastliest supercomputers. Just representing all possible states would require memory many orders of magnitude greater than exists on Earth!

Instead, Go solution will rely on more creative algorithms that truly "understand" complex strategic concepts like influence, life and death subproblems, territorial evaluation, etc. This remains an grand challenge in AI and machine learning!

Humans still competitive against computers

Unlike chess where even an average phone app can demolish the world champion, Go AIs still make recognizably un-human moves that pros are able to exploit. Top player Ke Jie analyzed his games against AlphaGo and noted several subtly inferior moves rather than consistently perfect play. 9-dan professional Park Junghwan has also scored wins against DeepMind‘s latest AlphaGo Zero algorithm.

This suggests the game retains hidden complexity beyond modern solving capability. Sure, computers overwhelming win most games now, but their play suggests room for refinement rather than dominating perfection.

Open problems towards solving Go

From my interviews with Go researchers and AI experts, several key challenges remain:

• Evaluating vague positional factors like influence and territorial value. Even the best AIs can misestimate winning chances.
• Solving complex localized life and death problems flawlessly. Humans still better here.
• Finding optimal opening strategies, especially as black when moving second.
• Modeling global patterns beyond local positions.
• Handling seemingly irrational human gambits.
• Integrating conceptual knowledge and raw pattern recognition.

So while Go AIs have come ridiculously far, matching human creativity and intuition remains an open grand goal!

Of course, as a Go lover myself, these unsolved mysteries are what keeps me coming back. The profound complexity that has enthralled players for millennia endures even in the computer age. AlphaGo may have beaten Lee Sedol, but truly solving this iconic game remains a golden grail for AI researchers and hobbyists alike!

Let me know your thoughts in the comments! I‘m always down to nerd out over the latest innovations. Maybe you‘ll discover something to finally crack Go‘s endlessly complex code once and for all! But for now, it remains the world‘s most tantalizing unsolved game. Come join the quest!