Advice is a funny thing. When people take your advice, they often feel like they're doing you a favour. And when it doesn't work out, you tend to get the blame.
Moreover, most people don't want advice. One of the most important things I learned in Accenture was that there's a time and place to offer advice. The right advice at the wrong time falls flat, or, even worse, causes irreparable damage to relationships. One should be even more careful dispensing advice than receiving it.
Don't give bullshit advice. Don't tell an entrepreneur whether you think his idea will work. You don't know. You have absolutely no idea. Entrepreneurs have to see around several corners. They have visions for a future that doesn't currently exist. That vision currently doesn't exist because the product hasn't been made in just the right form and/or because the world is just not ready for it. Yet. And you can't predict how or when the world will change. You may have decent product intuition, but the great achievements in innovation are so massive precisely because everyone else got it wrong at the time. Don't be that guy.
Jason suggests that instead of answering the common "Do you think it's a good idea?" question, one should focus on clarifying hypotheses, advising on process (do they know their lean methods from their elbow?), making introductions to other entrepreneurs, potential customers, and investors, and providing support.
Those are all very good points, but going so far as to suggest you can't offer an opinion as to the validity of the idea is silly. Opinions are just that - opinions. No entrepreneur worth his salt will give up on the next Facebook because someone else (however important) said it was a bad idea. In fact, every entrepreneur working on anything important will be told over and over that their idea stinks and can't possibly work.
The reason you shouldn't declare assertively that such-and-such idea has no chance is really because you'll look like a complete idiot when it does turns out to be the next Facebook, after the other guy/gal ignores you (like the other 500 people who said it couldn't work) and goes ahead anyway. Save yourself the trouble, and instead, offer a tempered opinion: "Well, I can't say for sure, but I think you may have a lot of trouble with X, Y, and Z, but at the end of the day, that's just a high-level guess".