"If someone in the future was to take AlphaZero as an algorithm and run it with greater computational resources than we have available today, then I will predict that they would be able to beat the previous system 100-0. If they were then to do the same thing a couple of years later, that system would beat the previous system 100-0. That process would continue indefinitely throughout at least my human lifetime."

I read this quote from David Silver (from Deepmind) which caught my attention (https://www.chess.com/news/view/david-silver-alphazero-reinforcement-learning):



I was puzzled about the "process would continue indefinitely throughout at least my human lifetime" part. Why would subsequent generations of AlphaZero beat older ones 100-0 - does it imply that older ones hit some sort of "bug" in the learning process that newer ones just learn to exploit?
If the learning process is "smooth", I expect that it would be harder and harder for newer generations to beat older ones, and they would't do that at 100-0 rate, but more "human-like", say 80-20, etc.

I guess no one knows "the right answer", but I am curious what others think about this statement.
(David Silver is talking about the chess version of the engine in this particular article, but I think the same would apply to go as well.)

What puzzles me is that AlphaZero cannot be many handicap stones away from "perfect play": a 100-0 score suggests more than one stone handicap difference in strength, so hearing David Silver saying that he expects at least 20 or so more generations where each beats the previous one 100-0 seems very counter-intuitive.

Even if 100-0 doesn't mean one handicap stone difference, but only means 1 point difference in final scoring, it doesn't seem to me that we can expect a gap between current top AlphaZero-like engines and "perfect play" to be 20 points or more...

As a player gets stronger they become more consistent so a particular winning percentage in games becomes less and less points / handicap stones difference.

The strong Go programs, to me anyway, are not likely to be impacted by this variability.

To me, Silver's claim seems reasonable (at least to the point of the game being solved).

They are not likely to misread a life/death problem, or a ladder, or miscount.

I guess my point is that there are differences between humans and machines.

Even though Human A is better than Human B, there is still often some nontrivial chance Human B will win. And many of the reasons for this do not apply to machines.

In stones there are unlikely more than a few above today's best bots, but strength classes defined as having ~75% winrate above the previous one are harder to tell.

Defining "moha" classes as separated by 75% winrate is more or less the same as saying that these classes are 200 Elo points wide.

ranks around 7d EGF are about 200 Elo points wide and in the limit of perfect play, rank width approaches infinity Elo points.

This I find hard to believe. As mentioned I think 1 stone (14 pts) cannot worth more than 28 classes, thus quite finite Elo-wise.

Consider the last class difference. A player getting 25% against perfect play is something like dropping a point in every other game (thus 0%*50% + 50%*50%, either as draws with correct komi or W wins with fractional komi). This already means more than half point strength difference, and it seems reasonable to assume that preceding classes only get wider pointwise.

But I don't understand why 1 point komi cannot exceed 400 Elo points (1 point = 2 classes and 1 class = 200 Elo), or vice versa, why 400 Elo points is always less difference than 1 point komi.