It won't even amount to a statistic. Yes, there's plenty of work to continue to do, but even today cars can drive well enough to avoid the vast majority of accidents that humans get into because they're attentive and alert at all times, and they have much better adapted sensors, deployed in more and better locations, etc. The driving record of the existing autonomous systems is really incredibly good. Yes, they may only drive in conditions they're already prepared for, but as I said, most driving is of that sort anyway. Given another 5 years of polish and good heuristics on when NOT to use autonomous driving I think the accident rate of these vehicles will be a fraction of a percent of what it is for humans.

that 99.9% of the routes that cars drive today are on the same few miles of road? Do you know that its really easy for different computers to exchange information. In other words if its pre-mapped, its pre-mapped for EVERYONE, and you really don't need to pre-map a ton of routes hither and yon to get everyone doing 98% automated driving. That's 98% fewer accidents.

Yeah, its hard to believe fixed routes won't be entirely mapped soon. Aren't there city buses that can already do that? I believe so. Because its a small fixed route it can be completely mapped and analyzed to the point where there aren't any surprises except what normally isn't there, you can pick it all out, you already know all your navigation decisions, etc. It will still take a couple more decades for the whole thing to get routine. I doubt truck drivers are losing TOO much sleep yet.

I use a number of products like that which really don't get much in the way of updates. Even so, at least your project has a maintainer that probably answers the very infrequent question and obviously addresses any bug reports in some fashion. Lots and lots of sourceforge projects never ever release code. I expect many die at the "I had an idea" stage, but others just never really sort out the organizational or "marketing" issues (IE getting people interested and trying the code so that something grows).

http://www.lazard.com/PDF/Leve...

Clearly the LCOE for SPV and ST (even with storage) is NOW competitive with nuclear power. Its probably not yet competitive with NG, but again NG is getting a free pass on CO2 and other issues. Given that recent research suggests that the actual social cost of carbon may be as high as $1000/ton we're pretty sure at this point that SPV is a huge good deal. Why do you think it is growing by leaps and bounds?

Yes, of course subsidies help, but they don't even cancel out 10% of the subsidy that coal/gas/oil get. Just the EXPLICIT subsidies on fossil fuel use are on a par with ALL the subsidies for renewables, so it isn't even clear to me that in terms of incentive we wouldn't be best off just getting rid of everyone's subsidy, not even counting carbon costs.

The truth can be determined with FIVE MINUTES of looking, and invariably the people who are trotting out this "solar panels use more energy than you get" BULLSHIT aren't interested in facts. Its an outright lie that someone has perpetrated on the public and the people who continue to circulate it are either VERY ignorant, like they haven't read Wikipedia, or they're actively dishonest. There's no legitimate skepticism involved at this point. If someone had brought this up 20 years ago we'd have all scratched our heads and thought "well, that's a good question, sounds wrong, but we'll get back to you." Today, in 2015? Its just a scurrilous lie.

All you probably NOTICE are the more successful FLOSS projects. For each one of those, there are 100, possibly even 1000 that languish, maybe make a release or two, and then lie in their shallow graves on sourceforge for the next 10 years.

Yeah, a couple, like since the 1970's. This old canard is just PURE FUD, it was never true, never even half true, etc.

Yeah, I had a National Semiconductor programmable calculator, forget the model number. It was pretty much the same thing, nice RPN with a big stack. All you had to do to program it was hit the 'start program' button and the 'save program' button when you were done. It was pretty slick. Thing lasted forever too.

Why would anyone have to choose? There are numerous languages out there which will do basically all of these, starting with FORTH, which was released in 1968 IIRC...

I'm sure you can find some person somewhere who's mentality is so limited they don't make basic connections between actions and consequences, or fail to make basic generalizations. I don't think that means such things aren't part of basic human intellect.

You're right of course, but that's the SIMPLE PART. The harder part is judgement. Even the stupidest human being has a vast amount of common sense, masses of rules of thumb which they have internalized and a deceptively deep understanding of context. How would a robot even know how to classify things as clothing or not clothing? Or more to the point washable or not washable? All but the stupidest humans would hesitate to throw piece of clothing with a large wet ink stain into a laundry machine with other clothes for instance, and said humans could reason this out from first principles (IE an understanding of how the washing process works, what ink is, etc). The level at which even the most sophisticated software operates is nowhere near robust enough make those sorts of reasoned decisions except in very carefully set up situations.

And then there are the higher level dimensions to the whole thing. When is it appropriate to wash things and when not? Which things do you have a RIGHT to wash and which things do you have a RESPONSIBILITY to wash? Since the 1950's people have gone on about the "3 laws of robotics", but Asimov would have been the first to point out that such things couldn't possibly ever be imbued into a machine. Its not even just the logical and epistemological limitations of those sorts of strictures themselves, but simply that we cannot define the situations wherein they would operate or determine when they were being violated. We can't make a self-driving car because we would have to teach it things like "Its better to run over the old man than to run over the baby when you cannot avoid them both." Obviously we'll live with robot cars that simply do one or the other by chance, but to imagine that anything short of a fully conscious general AI could make that sort of decision in a 'human-like' way is patently ridiculous, and we haven't got even the slightest idea how such an intelligence would be developed.

You say 20 years, but I say 100 years. We've barely set our foot on the first step of the path to understanding how to make something like that, and the most critical challenges involved have barely been imagined.

lol, yeah, that's what I was thinking, they must be doing this research at NIMH!

Gold isn't any more 'real' or 'funny' than any other sort of money.

I mean basically we have $1 trillion worth of funny money from QE1-42 that has to be burned sooner or later anyway. If no actual real money is spent, is it really a valuation? Is it really a bubble?