>I don't understand why it means that Bitcoin is insecure for people to be able to mine in their gpu.
What do you mean? People can mine with a gpu, but they won't make much because of the many miners using much faster ASICs. If you think Bitcoin should have been made so that ASICs weren't possible: a number of altcoins agree with that idea, but there's an argument to be made that optimized hardware is always possible, and by making it hard to make, it makes it more likely that a single small group who manages to make optimized hardware will be the only one to do so for a large time period, will be able to make huge amounts of profits funding more hardware, and at that point will easily be able to get up to >50% of the network hash rate (which breaks Bitcoin's decentralization and allows rewriting the blockchain). By using a proof-of-work that's relatively straight-forward to optimize means that when the possible mining rewards eventually incentivize development of ASICs, it's likely that multiple groups will be able to build them.
Yes, but this way you're trading an elite made of hardware specialists and people that can afford to hire hardware specialists, with an elite made of people that can afford to fill datacenters with the most specialized hardware. Mining quickly becomes a game of "who can throw the most money into the problem" and ends up skewing the pool significantly because you are rewarded with money for winning the game, which means you can afford even more of the specialized hardware. That's the big thing with Bitcoin: All the miners are rich folks (these days, mostly Chinese) with huge datacenters and tons of cheap electricity. I have 12 of the USB ASIC miners that are many times unable to even pay for themselves in electricity costs because the difficulty is so high, which is the problem that squeezes out grandma's GPU miner and pushes even more people out of the mining game (not to even mention the ballooning size of the blockchain).
Sort of. There is a potential that if you take over the network that it all becomes worth nothing because users understand the trust problem and cash out in the lead up. Or the users elect to not use your currency, it could still hold value, as the blockchain until that point is valid, but it remains paused until the 51% is resolved. This encourages you to work covertly to organize the 51% stake and you have to hold it in secret or all your coin/power may become worthless. It's essentially the cold war and in the same way, it's in everyone's best interest to avoid destroying the value of the coin by being selfish in an attempt for differential advantage.
Sure, and to this point, there have been at least a handful of times when the 51% attack could have happened, as well-known, related mining pools could have combined their power to achieve the majority. It has been plausible before, and I'd imagine what you mention is the reason it still hasn't happened. It would destroy the coin, because who knows what happened while they had majority. There would probably be some kind of fork, and I'm sure the value would diminish. It would be messy for everybody.
What do you mean? People can mine with a gpu, but they won't make much because of the many miners using much faster ASICs. If you think Bitcoin should have been made so that ASICs weren't possible: a number of altcoins agree with that idea, but there's an argument to be made that optimized hardware is always possible, and by making it hard to make, it makes it more likely that a single small group who manages to make optimized hardware will be the only one to do so for a large time period, will be able to make huge amounts of profits funding more hardware, and at that point will easily be able to get up to >50% of the network hash rate (which breaks Bitcoin's decentralization and allows rewriting the blockchain). By using a proof-of-work that's relatively straight-forward to optimize means that when the possible mining rewards eventually incentivize development of ASICs, it's likely that multiple groups will be able to build them.