I heard yes.

As I know that there are professional ...

This is really interesting!

The computer has not only changed chess openings, it has changed everything. To a computer, chess holds no more mysteries. If you want to know if your move was right, you ask the computer for a 5 minute analysis, and the answer will be definitive.

I think this idea is subtle. Just because computers can beat us at chess (and probably in the near future go) doesn't mean they play really well in an absolute sense. It might be that we are just really bad ("puny humans!"). There still might be a long way to go in playing strength. For example in computer chess the top program beats the 3rd best with some regularity and ratings of computers chess machines are going up rapidly.
see: https://praxtime.com/2014/03/24/chess-technology-progress/

For example, suppose that deep learning ala alphago get to the point that they can be pros every time. It might happen that some completely different approach might crush alphago and that approach could play go very differently than humans do, while alphago plays a lot like humans since its neural net was trained on human play.

Could Magnus Carlsen defeat a computer with a handicap? Say the computer was down a pawn. A knight?

Could Magnus Carlsen defeat a computer with a handicap? Say the computer was down a pawn. A knight?

Could Magnus Carlsen defeat a computer with a handicap? Say the computer was down a pawn. A knight?

It is pretty rare that a move formerly considered reasonable has been completely refuted by a computer and abandoned (I can't think of an example off the top of my head). Consider the Go analogy; AlphaGo probably isn't going to discover that some established joseki actually ends in disaster for one side due to some incredible tesuji.

What happens more often is that computers can pretty exhaustively search the possibilities in a position or line and confirm that it is promising or not. A typical opening line might obtain some static advantage for one side but concede some dynamic possibilities to the other, and the computer can check whether those dynamic possibilities really come to anything in the end with best play. If not, the player can go ahead and play the risky move (if he remembers or can reproduce the computer analysis over the board!). As a result, chess style has become more "concrete", focusing on whether moves work out in actual variations, rather than evaluating them with rules of thumb and pattern-matching. (Of course people have always done both; it's just that the balance has shifted more to the former.)

One interesting result is that many top players now play openings that are less sharp and concrete, avoiding computer-assisted preparation on the part of the opponent and trying to "just play chess", relying on their more human strategic skills. Magnus Carlsen (the current world champion) in particular is the poster child for this approach.

My bold face. Even among human pro players a joseki might be abandoned because one side has what seems a very small advantage, e.g. one point or less. AlphaGo might abandon a joseki because of even smaller advantage, not due to the joseki being "crushed". However, can AlphaGo make such quantified judgments?

The most recent handicap games against a top-level player were against Hikaru Nakamura (currently world #6) earlier this year.

The handicaps were pawn-and-move (Komodo played Black and without one pawn), pawn (Komodo played White and without one pawn), exchange (Komodo played White without a rook, Nakamura played Black without a knight), and 4-move (Nakamura played White and played 4 moves in a row before Komodo started to play).

The first three games were drawn and Komodo won the fourth. All games are at the link I provided earlier and are fun to play through if you are interested in top-level handicap chess.