Why do computers lose against people

So I've been thinking, how can a computer lose against a player when the computer can calculate 3million moves per second? You might say, well a computer isn't calculating for moves that are to be unexpected... Well I counter that and say with 3 million moves per second it's searching every possible combination of moves/counter moves to at least 6 moves each to a depth of 8 - 10 places.

For example, even if a player decided to place a peice in a retarded position, the computer would have already took that move into account. I just don't understand how a human, with the ability to calculate roughly 3 different moves, each with a depth of about 6 - 8.

I'd like to see Carlsen against a computer, talking through his logic while playing and seeing if he can predict what the computers move will be.

While I can't give you an exhaustive answer, I think one reason that computers lose to people is because they simply don't understand certain positions. I beat Fritz 9 at maximum strength with the King's Indian Defense simply because it couldn't see all of my attacking tricks and resources. I believe the Stonewall Attack is yet another position which computers are terrible at, as they don't realize their king is under attack until it's too late. 

I personally think it is simply because computers cannot "learn" a person's playing personality, and that gives humans a big advantage.

They don't.

I agree.

Samurai Jack... I don't know what to make of this. It's so random.

...

For that, I think we first have to know how does a computer software figures out the best move in the first place. I know that there is something called evaluation points, where a software takes into account several advantages and disadvantages of a position to come up with a numerical value. However, as per my observation, no two chess softwares comes up with the same evaluation points for a given position, which shows that there exists discrepancies in the best move, that different softwares come up with. Computer chess tournaments are actually a battle between these eval points. A better chess software comes up with accurate eval points. It then goes on to evaluate deeper variations and searches for the best move. The searching strategy plays a key role in determining how fast the evaluation completes, and hence the algorithm's efficiency. I found this topic very interesting. I would like to have more insight into chess algos.

Well I think the problem lies within the computational speed of a computer. A computer wanting to calculate 1 set play to a depth of 8 moves. That's 8! which is 40,320 different moves for that 1 peice of play. Now, we allow the computer to calculate 3 peices of set play to a depth of 8. That is 8!x8!x8! = 65,548,320,786,000 which is a staggering amount of moves. So of course we can see this is outrageous and so the computer must decide upon the most likely moves, which let's say will be 4. A total of 4 likely moves to a depth of 5 places.

5!x5!x5!x5! = 207,306,000 moves which is much more managable. I've been reading up and apparently this is the biggest issue with people beating computers, people like kasparov play anti-computer moves and so it's extremely unlikely that the computer would have considered that chain of moves that would follow. So the computers biggest weakness by the looks of it is the fact that the computer tends to 100% calculate positions but only for the chains that are likely to be played. 

Before you assume you have beaten the computer, check your task manager. If antivirus program is running, a chess program would use as little as 5% of cpu power.

Chess computers have a horizon effect. Basically if a computer is set at a low level, it will only see say four or five moves in advance. A human looking at the board with his eyes may see visually, for example,  that he can advance his pawn say 6 moves and make it a queen. But if the computer can only see 5 moves in advance it doesn't see the winning 6th move.

Ask a chess computer programmer?