TPM has a nice history of AIG Financial Products. In essence, AIGFP's major sins were (1) to commit "ratings arbitrage"—exploit the fact that AIG the insurance company was AAA rated to allow AIG the trading/CDS-writing company to do all sort of things more cheaply than it might otherwise have to—(2) to fail to model housing prices correctly, in which they were not alone even though people on the outside looking in thought the whole enterprise was crazy, and (3) to assume that events that would trigger payments on some of their insurance/CDS contracts would require an economic collapse that would take out the counterparties first (oops!). Separately you have periods of time where AIG has trouble with its auditors or Eliot Spitzer, though it's unclear to me if their level of trouble was any higher than the average financial firm.
I keep groping for some middle ground, not out of instinct but because I know some of this work is actually valuable. Packaging up diverse loans to reduce risk makes a lot of sense; it just went bad when the packagers started making crazy assumptions about the housing market and their borrowers' ability to pay that everything went awry. Compensating traders fairly well makes senseit's just that when the compensation comes in the form of "guaranteed bonuses" that it looks crazy. I'm actually pretty confident that the people running the show understand that there's some value in the changes in Wall Street over the past thirty years, but the question is how to preserve all that value while taming the excesses.