He expected a social backlash to the march of technology, like the anti-facebook groups and then violent sabotage. He expected it to slow things down, but ultimately technology is irresistible. That's not a statement by a person who cares about the details of his predictions.
This whole prediction accuracy threw my point off topic. The point is startup ideas are just as interesting as Ray's predictions about the future. The number of people a company's idea impresses doesn't predict success of that particular company.
The bigger factor is probably the type of people the startup world attracts, the reorganization. They don't think like the people in larger corporations, they're more motivated, addicted to stress, whatever. The ideas they have aren't more amazing than the ideas in big corporate R&D departments, the difference is now the projects are tested out before getting cancelled by people who saw them only in a spreadsheet.
So VCs don't care about ideas, like they don't care about general predictions. They're mostly investing in small companies, not setting up trading strategies. In the current world where VCs invest in particular companies of people, not ideas, it makes more sense to invest in companies that are somewhat functional.
Even if their idea seems dumb or has to change, competent people will make something work. People that bet it all on an idea aren't flexible, aren't a wise investment. That's why investors prefer simple, practical ideas when nothing is built yet. More depends on the people than the idea and a fast to build idea lets people show their important skills, like salesmanship.
---- As far as "safe" stock market predictions, that's your interpretation since you're interested in money centered economics. His "safe" prediction is that money would change step by step, become obsolete sometime this century. The fancy financial instruments made by quants which screwed up the stock market fit right into that "safe" prediction. It's safe because money is a bad measurement system and as we get more information we'll be able to measure better.
I could predict today that in 50 years, the earth will still exist. I am almost certain it will be true. But it's also meaningless, because everyone could predict that, and there's really not much you can do with that prediction.
The same with many of the predictions that you mention. It's a fairly good bet that in 10 years, people will use computers to communicate. But that's a useless prediction - so what? Maybe it's useful fodder for cocktail-party conversation, but maybe not even then - there's really not much meat to discuss in that.
In fact, Kurzweil was wrong about even one of his "safe" predictions. He predicted that the stock market would be higher in 2009 than in 2001, something that's been true of every decade since the 1930s. It wasn't.
The details matter. There's an amazing human trend toward confirmation bias - when people find that one of their predictions hasn't panned out, they try and broaden it out, ignore a few details, and fit the data to their predictions. But if you want your data to have any validity at all, you can't do this. Your predictions need to be specific enough in both details and timeframe that you can unequivocally state whether or not they're true, in their original form. And if they aren't, well, they're false.
(And I suspect your overall point, that VCs don't care about these predictions, is something I'd totally agree with. But we have a different idea of what a prediction is. I don't think pundit wankery that speaks in vague generalities about the future counts as a prediction at all. I do think VCs care about predictions, but they care about prediction in the more specific sense, something that is actionable and will be either right or wrong.)