The assumption of a Go playing computer program is, that the machine provides an agent who plays the game. Man or machine have to prove inside the same rules of Go that they are able to win the challenge. The idea is, that the Deepmind Go playing engine is a narrow AI which is working inside the environment known as “game of Go”.
This assumption has to be questioned if it's true. In most cases, AI game playing agents are more than only a synthetic player but what they are providing instead is a learning environment for humans. From the game of chess this extra feature is known from the past. Most AI chess engines are equal to a computer based training in which the human player learns to improve his chess skills. In the mainstream discussion around AI engines this extra feature is ignored as not relevant. But it has to do with the social implication of an AI software.
In the first case, the AI is simply an agent who fits into an existing game, in the second case, the AI is producing a new kind of game not known before. The hypothesis is, that it's not possible to create a narrow AI Go playing agent, and each attempt in doing so will look like a strong AI in which a human player is put into the social role of learner.
Let us analyze the relationship between Lee Sedol and the Deepmind Alphago team by it's social roles. They aren't playing against each other, but Lee Sedol was embedded in the project. That means, the Alphago software is a computer based training and Lee Sedol is using the software to increase his strength. It makes no sense to rate the strength of a computer based training system but it's the other way around. Different human go players will use the AlphaGo software with a different purpose. Somebody gets bored while other can profit from it.