AlphaGo strategies and innovations – DeepMind friendly interpretations

DeepMind Blog is another site preferred by The Information Age. This is the blog of the Google’s most significant acquisition within the Artificial Intelligence (AI) space of ventures or companies.

DeepMind is also the venture of AlphaGo achievement. To the less informed about what AlphaGo is, here it is a reminder: AlphaGo is an AI implementation of a machine player able to play the ancient game of Go. Go is a chinese and oriental ancient board game. It is a remarkable board game, the complexity of which surpasses that of Chess; the number of possible moves in Go surpasses even the number of particles in the entire Universe. This is an instance of an all human and material design with a complexity beyond its own material possibility. It makes us wonder as to what is the nature of mathematical imagination or its ability for thinking about the infinite or about its own constitution and essence.

The post from DeepMind that I would like to write about here today is precisely about the innovations that AlphaGo brings to the table of discussions about Go – this game is widely discussed in the oriental cultures and societies that have played it. Ancient philosophers like Confucius even recommended it as an imperative requirement for the educated mind.

But the AI implementation of an autonomous computer able to play Go promises to both take the game to another level of analysis, and of helping AI researchers achieving AI milestones that one day will further AI to a level of superintelligence or artificial general intelligence (AGI). Indeed the innovations we are introduced into in this post are of complementary and interdisciplinary nature – prerequisites for AGI. We glimpse a combination of computation, learning algorithms, game theory (strategic thinking, economic or otherwise) and even political science in this post while our eyes parse through it eagerly:

Indeed the interdisciplinary capacity of research efforts such as these is one more – if we needed one more – vindication of how clever investments in research & development (R&D) sooner or later will reap benefits somehow, somewhere. How could we, on first impression thought about a benefit, for instance, in energy optimization in large scale facilities as data centers of other complex manufacturing units? The only caveat here is the inherent uncertainty of these outcomes, but my view is that this is a warranted uncertainty: it is more a question of timing and/or opportunity than an invest/not invest tradeoff. Obviously more complex inputs should also be taken into account, the quality of the science and technology involved or the right mix of human resources coupled with the right cultural background being indispensable.

One of the great promises of AI is its potential to help us unearth new knowledge in complex domains. We’ve already seen exciting glimpses of this, when our algorithms found ways to dramatically improve energy use in data centres – as well as of course with our program AlphaGo.  

Since its historic success in Seoul last March, AlphaGo has heralded a new era for the ancient game of Go. Thanks to AlphaGo’s creative and intriguing revelations, players of all levels have been inspired to test out new moves and strategies of their own, often re-evaluating centuries of inherited knowledge in the process.

Ahead of ‘The Future of Go Summit in Wuzhen’, we summarise some recent examples of AlphaGo’s strategic and tactical innovations, and the new insights they have revealed.

dia1.width-1500_Q2AjGgO

The political science impression mentioned earlier were dawn on your friend’s mind when reading the paragraph below:

Although Go is a game of territory, most decisive battles hinge on the balance of power between groups, and AlphaGo excels in shaping this balance. Specifically, AlphaGo makes masterful use of “influence,” or the effect of existing stones on surrounding areas. Although influence cannot be measured exactly, AlphaGo’s value network enables it to consider all stones on the board at once, endowing its judgment with subtlety and precision. These abilities let AlphaGo convert local regions of influence into coordinated global advantages.

In this game (Dia. 1), Black (AlphaGo) has little secure territory, while White has three corners, but Black’s influence radiates across the entire board. In particular, while the marked exchange solidifies White, it also improves Black’s potential. Go players usually shy from such exchanges, which pay a definite price for uncertain profit, but AlphaGo combines its sterling judgment with a keen sense of risk and reward to make such moves possible.

The newfound capacities of AlphaGo go as deep as putting in check centuries old Go players recommendations. Machines are proving to not being only slaves of human masters: machines with this kind of intelligence are more of an enhanced (and how enhanced that might be… ?) partnership for humans. We should reckon though that at our current level of development this is only a simulation of a future partnership; it is not a great stretch of the imagination to envision a future where the autonomy of these machines will be on par with our own human autonomy. The question then will be if we are really in control of a simulation or not… I promise to one day post about these delicate ethical issues here in The Information Age.
deepmind_logo
However, the value of influence depends entirely on context, and AlphaGo relinquishes influence freely when it can be effectively mitigated. In the the game displayed in Dia. 2, one of the most surprising in its oeuvre, AlphaGo has just played an incredible six stones along the second line. Go players have a saying: on the fourth line there is influence, and on the third line there is territory, but on the second line there is only defeat. AlphaGo’s play at first looks deserving of such censure, as these moves give White both strength and influence in exchange for Black’s paltry 4 points of side territory. Most players, unwilling to bear the ignominy of playing the marked stones, would reject this line in an instant. Yet AlphaGo judges it worthwhile to keep White’s stones separated, and in the following exchanges, slowly erodes White’s influence from the top and bottom to secure a winning advantage.
dia2.width-1500_w5yXikQ

New Moves, New Patterns

AlphaGo has also played several opening novelties in its recent games, the most salient being the early 3-3 invasion and a new variation of the “Magic Sword”. Each defies conventional theory, but proves sound on deeper inspection.

dia3.width-1500_rzlxptT

The Early 3-3 Invasion

One of the most territorial joseki (corner sequences) in Go is the 3-3 point invasion, shown in Dia. 3.

(…)

dia6.width-1500_cUqzJlo

Though slightly less secure, Black retains miai (options) to escape on the left or finish the joseki later, and has gained territory while ceding only moderate influence. This strategy has created a great stir among professionals, and at least one has already tried it in an official game (Dia. 6).

The New Magic Sword

Originally trained on human data, AlphaGo knows modern joseki and usually plays accordingly. However, in the “Magic Sword,” a famously complex joseki family named for the cursed sword of Muramasa, it diverges. 

(…)

Most Go players would not consider playing this variation, as it gives Black a powerful wall, but White’s follow-up approach declares that Black’s influence is not as valuable as it looks. If Black does not reinforce the wall, it may even become a target. Kim Jiseok, one of Korea’s top professionals, recently played this line in a tournament game (Dia. 10), which he went on to win.

dia10.width-1500_wLKaRNs

The true nature of Go in an ever more improved partnership between human players and AlphaGo is what is promised in the end of another rewarding read from DeepMind. We hope that the Future of Go Summit in Wuzhen will feature these and more innovations of AlphaGo. And after these developments the future of AI R&D will surely brighten further, with applications of AI we now can scarcely fathom with our also enhanced human minds…

features image: AlphaGo beats Lee Se-dol again to take Google DeepMind Challenge series

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s