"AR/VR comes to mind, but I think it's too early to make the call there" Are you willing to make calls for things that were so many decades early that even if

"AR/VR comes to mind, but I think it's too early to make the call there" Are you willing to make calls for things that were so many decades early that even if they do eventually happen they aren't really the same tech anymore? People put real money into AR/VR in the late 1990s and had a real marketing push behind it. It was definitely wrong. If it happens to be right a literal human generation (~40 years generally) later, well, I don't really think it made the people pushing it in the 90s right about what they were saying then. AI has an even bigger one, with its own name: AI Winter. https://en.wikipedia.org/wiki/AI_winter https://en.wikipedia.org/wiki/AI_winter Another case of what they were promising simply wasn't possible at the time and skeptics are right. That one we're already a full human generation away from, and again, nobody at the time was going "Give me money now, and in 40 years we'll definitely have AI going!" And even now, while we have tech that does some things they wouldn't have dared to dream of, our current tech also still can't do some things they promised were just around the corner. There's also a class of "technology failure" I'm a bit of a fan of myself, which is, the technology that fails to live up to its promises even though it doesn't die out. SQL is a failure... it is not possible for casual business people to do away with database engineers because the language is so easy to write. Obviously, SQL lives on even so, but it failed out of its original goal. In fact there's several things that started life as "we'll get rid of programmers" like that, such as COBOL. Object orientation is a failure; the promise was that we'd end up with huge libraries of reusable components such that all programming would be just wiring them together. We have a huge number of libraries of basic "objects", we have a huge number of libraries of frameworks and other really big things, but there's a vast wasteland in the middle. You can never just go up to the Object Repository and grab a User object for your application. You also need some bespoke user model for everything. You can't just grab a permission system. You can grab a permission engine, but you have to wire a whole bunch of things up because even if you do grab a framework that tries to do this all for you the third thing the PM does is spec a permission like "The user is allowed to delegate to other users that they're allowed to see a subset of their content based on a regular expression, as long as it's authorized by their team administrator" and you're back to writing real code. Semantic web has been around multiple iterations of the hype loop by now. It was a promise to turn the entire web into a graph database. It is reduced to claiming success when Google writes very powerful ML stuff to extract basic information about a business from their home page and presenting it as a widget on the search results, using essentially no "semantic web" technologies in the process. It isn't anything like what it was supposed to be.