Once upon a time the Efficient Market Theory (EMT) was developed by Harry Markowitz. Other mathematically inclined people expanded his ideas into various specific market models and investment processes. EMT became very popular because it was intellectually attractive and mathematically was simple enough for people to understand and use.
The weaknesses of EMT, and other models that had been derived from it, was grounded in some of the assumptions made by EMT creators. Specifically, the assumption, most vulnerable to critique, has been a concept of all investors being rational and acting in a way that maximizes their returns and minimizes their risk (defined as volatility). Incidentally, using volatility as a measure of risk is another vulnerable assumption used when developing the EMT.

The investors eventually started to understand the limitations of the EMT and other models developed on it's basis by several famous researchers including quite a few Nobel prize winners. The assumption of market participants are rational and acting in there own the best interests has been disproved in several studies (as if it was not an obviously far from reality in the first place).
To overcome the shortcomings of the investment processes built on the EMT, behavioral finance was developed and became quite popular because it can explain deviations of price action and investors behavior from what would be suggested by the efficient market theory. The deviation of actual investors behavior from the rational actions has been creating opportunities for some managers to outperform the others consistently. One can assume that the out-performers could deliver the so-called "alpha" (excess return over what the EMT would suggest was possible) because these managers could ignore the crowd mentality and be more rational than their peers.

My guess is that the advantages of using behavioral finance, technical analysis and even long term statistical models are going to dissipate quite soon and EMT rule will be restored. The reason for that is an accelerated invasion of machine learning combined with algorithmic trading. Like any other algorithms, the AI driven algorithms are not emotional. They will easily pick up the inefficiencies created by irrational human behavior and will arbitrage those inefficiencies away. As the result we will be left with much more efficient markets.

Of course volatility still is not a good measure of risk. Maximal mark-to-market loss is probably a much better one. And even a better measure seems to be an irrecoverable loss that can be a result of forced portfolio liquidation or a jump to default in some fixed income instruments. However, in any case we now have enough computing power to incorporate any conceivable measure of risk into our optimization models. If only we could predict the expected returns better ...