How many computer-GM games have ever gone 100 moves?
humans better than computers at chess (my argument)

2 points.
1) 2 of the 4 games went 100+ moves Hikaru vs Stockfish
2) Even if the game only goes 50 moves, the human player feels compelled or has learned how to play chess in a far more forward manner than the chess which is as level minded as possible.
Ex. If this was basketball and a winning strategy involved dribble in zig zags or in one corner of the court, a CPU would do that, but a human would have un-learned to do that, so it may fundmentally be correct, and the rules should dictate the pace of the game, but people go beyond

My argument that they're not:
http://en.wikipedia.org/wiki/Deep_Blue_versus_Garry_Kasparov
http://en.wikipedia.org/wiki/Hydra_%28chess%29#Tournaments_and_matches
CTRL+F Adams.
Keep in mind that these are old games.

Ok, but if Kasparov approached the machine as if he [kasparov] was also a machine, I think he would draw or win.
He didn't try to beat it at its own game (to my understanding), he played like Kasparov has learned to play over his career and would approach a human opponent. If he re-learned to play against a machine style, that would be possible

No.
Stockfish 5 on a strong computer vs anyone right now: 1-0 almost every time. A draw is the best anyone can hope for.

I don't believe anyone/machine has beaten Stockfish yet...nor do I think Carslen has played Stockfish in a match game

I often wonder if a computer was boxed up, sealed with no input from programmers (perhaps we could allow it to store new games) and let loose into the tournament world, where would it end up?
my guess is it would start out at 3200 performance, but drop pretty quickly, and there'd be a lot of GMs getting draws against it.
Take for example stockfish on a macbook, seal it up and stop the programmers tweaking it all the time, and let it threaten to take away tournament prizes and then see what top-class GMs can do about it, as a group, each studying the other's games.
even stockfish has many flaws, it's just that there isn't really an incentive for people to find out what they are.

If a CPU was compelled as a human is to win a game in 100 moves, a human Super GM would rather easily beat a Super computer.
Sorry to say but this sentence alone demonstrates how little you appreciate the enormous gulf between contemporary programs and the best humans. Don't forget Nakamura was receiving big odds in his games.
I'd back Stockfish for a 10-0 win against any human past or present, even conceding odds of having to win the games by move 100.
I've figured it out.. plot twist. Humans are actually better than computers at chess.
I made note of this personally during the candidates tournament or the Zurich tournament upon comparing the games to the evaluation and making one preliminary finding; and that's the computer shuffles pieces and humans make direct moves.
Now here's the bigger piece: computers are set with no priority over excitement or making the game interesting, they calculate a winning strategy against the clock with the intention of making 2000 moves if necessary. They are not emotionally indebted to the game nor have any limit to energy.
If a CPU had a rule implemented that said the CPU HAD to win the game in 100 moves or less or it would forfeit it would begin to play like a human with the "feeling" to be compelled to make the game interesting/ exciting or at the very least to play for a a more concise win.
Instead of grinding down opponents with brute calculation it would make human irrational aggressive moves being compelled to achieve a result.
To continue to make things interesting the human player would lose after the 99th move, meaning before the game reached 100 moves the player would forfeit, but the CPU wouldn't be made aware of this, and would play with the incentive of achieving a result in 100 moves or facing defeat.
Most human GM games last under 100 moves and human's aren't built for games that are 100+ moves of building small advantages with infinitely small incremental gains.
If a CPU was compelled as a human is to win a game in 100 moves, a human Super GM would rather easily beat a Super computer.
The one flaw is, humans can make mistakes, CPU's are calculators 4+4=8