3 ms·

I'm not questioning the validity of the research. I'm questioning that one statement (a statement similar to one that appears in so many pop-sci articles). The

by sambeau 14y ago

I'm not questioning the validity of the research. I'm questioning that one statement (a statement similar to one that appears in so many pop-sci articles). The £5,000 to £100 estimate is based solely on the hope that the raw component costs will come down (through bulk-buying and the natural trend of micro-electronics to get smaller, more powerful and cheaper). Unless I missed something, this is not because they have some groundbreaking method that will reduce the number (or price) or components — it's just because there is an assumption that they will. Any IT researcher can make similar claims so they are essentially meaningless. A comparison between the £5000 and the component cost of a rival system would be more valid and would allow those of us who understand the possibilities for future component cost reduction to extrapolate. However, there is no guarantee that component costs will continue to fall indefinitely. Who knows what wars or shortages may befall us in the future. Prices may even go up. As this system relies on an Apple iPad does this price calculation assume that in a few years Apple will be selling a model for less than $50? I would be extremely cautious about making such an assumption. This cost also doesn't take into account the reliability level of the components that would be required to run such a system on a daily basis. You can put together a working prototype system for £5000 but as soon as suppliers are providing components that are mission (and life) critical there is no way that they will that cheap: the insurance each supplier would need would make a huge difference to the price. Add to that the high quality control & testing and the price is unlikely to fall much over the next 10 years. Common sense also tells us that we'd never be able to buy a system for so little. $150 is around the price of a replacement battery for my car or two cheap tyres. I cannot buy a replacement CD/radio for that much. This is for a car that is 7 years old. Surely over the last 7 years my car parts should have got cheaper? Newman and his group know all of this — and you know this — but you choose to ignore it as it doesn't make a good quote of a Pop-Sci article. Promising excited less-technical readers a self-driving car for £100 is just too tempting. As I said, I don't blame Newman (although I do think it detracts from their real achievements), I blame the reporter who should have gently called them on it and clarified what they really meant.

robotresearcher 14y ago
The article was pretty normal for its slightly crappy coverage of science/engineering. But the potential for extreme low cost of the mono-SLAM approach is actually the news here. It's almost all software. 1) You seem to be missing the important point - the approach they are using uses ordinary cameras and a computer. This really, truly can be very, very cheap to implement, once we have the (very expensive to develop but near-free-to-deploy) software. They are trying to see how much of the driving problem can be solved without using expensive laser rangefinders, etc. Your cellphone has all the hardware needed, but is not yet fast enough to do it in real time. OK, it needs more RAM too. 2) You have the prototype vs. production cost exactly inverted. For example, the iPad is much more expensive than the component cost of a dedicated touch display, which your new car already has installed for the GPS map anyway. 3) Your car's electronic traction control system is mission critical and uses very cheap electronics and near-zero-marginal-cost software. It did not add $5000 to either the build cost or purchase price of your car. 4) When an engineer talks about cost, she is not thinking about retail price.
- sambeau 14y ago
  Traction control cost $100 when it was introduced in 1971 which is ~$567 in today's money. A new traction control system costs £500-£800. Clearly those components aren't as cheap as you might think. When a reporter writes about a cost in a PopSci publication she should be thinking in terms of retail price as her readers will be unless it is explicitly stated.

by sambeau 14y ago

robotresearcher 14y ago

The article was pretty normal for its slightly crappy coverage of science/engineering. But the potential for extreme low cost of the mono-SLAM approach is actually the news here. It's almost all software. 1) You seem to be missing the important point - the approach they are using uses ordinary cameras and a computer. This really, truly can be very, very cheap to implement, once we have the (very expensive to develop but near-free-to-deploy) software. They are trying to see how much of the driving problem can be solved without using expensive laser rangefinders, etc. Your cellphone has all the hardware needed, but is not yet fast enough to do it in real time. OK, it needs more RAM too. 2) You have the prototype vs. production cost exactly inverted. For example, the iPad is much more expensive than the component cost of a dedicated touch display, which your new car already has installed for the GPS map anyway. 3) Your car's electronic traction control system is mission critical and uses very cheap electronics and near-zero-marginal-cost software. It did not add $5000 to either the build cost or purchase price of your car. 4) When an engineer talks about cost, she is not thinking about retail price.

sambeau 14y ago
Traction control cost $100 when it was introduced in 1971 which is ~$567 in today's money. A new traction control system costs £500-£800. Clearly those components aren't as cheap as you might think. When a reporter writes about a cost in a PopSci publication she should be thinking in terms of retail price as her readers will be unless it is explicitly stated.