re: Google Unveils Self-Driving Cars

quote:

---

25 mph is absolutely practical for in city driving and as a taxi service

---

quote:

---

Also, the Lexus and Toyota's that are 700,000+ miles accident free go well over that speed

---

Again, all under very controlled circumstances.

quote:

---

yes. Hell, they already do this.

---

Really now? They can accurately interpret a policeman's hand signals and other gesticulations (which can get quick and ambiguous depending on the situation)? Are they able to "see" who the signals are meant for?

Will these sensors be able to see/interpret a traffic wand at night, amongst a background of headlights, flashing emergency lights, and street lights? What if it's a regular joe (w/o a wand) doing the traffic control until the cops arrive?

We're not talking about the simple act of a cyclist signaling a left or right turn here.

quote:

---

quote:

---

How many airline disasters have been linked to malfunctioning software or "sensors?"

---

I bet that failure rate of the software is lower than humans.

---

Let's combine software failure with sensor, electronic and mechanical failures and see how they compare then.

quote:

---

The big issue has actually been how reliable the software has been.

---

The flip-side argument to all this is that I can come up with a ~dozen cases where an airline disaster was averted because of the improvised actions of a brilliant flight crew.

quote:

---

Pilots aren't as skilled outside of the autopilot.

---

I have a really hard time accepting the notion of an autopilot being able to successfully pull off a "Sullenberger" all by itself.

quote:

---

A self-driving car will never get tired, distracted

---

They're also incapable of adapting to unforeseen (i.e. not programmed) situations, and on-the-fly problem solving.

quote:

---

or drunk

---

Much tougher DUI laws would go a longer way in eliminating DUI related fatalities.

quote:

---

and it can see farther than its human counterparts, see at night, and see in 360 degrees.

---

I see they left bad weather out of the equation... again.

quote:

---

but keep comparing it to windows.

---

Are Micro$oft products the only ones that are buggy?

quote:

---

They have radar sensors, roof sensors, camera sensors, etc.

---

All of which are susceptible to software glitches, hardware malfunctions, blockages from snow, ice, mud... not to mention regular ol' road damage from flying rocks.

quote:

---

Right... Because Me, XP, Vista, 7, 8... have all been so bug free.

---

The cars don't run Windows at all. The way that consumer software is written vs. mission-critical software is so profoundly different that you can't really compare the two. The vast majority of Windows issues are due to poorly-written 3rd party software and drivers, and also due to the fact that the OS has to be open to allow for this vast array of software to work, and to do so on an unlimited variety of hardware configurations. When the exact hardware and software configuration is known, probably north of 90% of bugs can be eliminated. Another 90% of the remaining bugs can be eliminated by extremely strict programming guidelines. For consumer software, time is of the essence, so if it kinda works, it ships. There are no checks and double and quadruple checks over each and every path the code may take. That last 1% of bugs can then at the very least be caught and handled, if not eliminated entirely through redundancy in both hardware and software.

It's just a totally different set of requirements and goals, and that's why most of the software you use is buggy.

quote:

---

Even military software controls (hard to get more "stringent" than mil-spec!) aren't immune to bugs. There have been many incidents where bugs in a modern fighter's FLCS has led to the loss of the plane.

---

Mil-spec? You mean the specs that get passed on to the lowest bidder?

And when was the last incident? How much has been learned since then? How much more can go wrong flying vs. driving?

quote:

---

Even an established avionics powerhouse like Garmin issues regular software updates and bug fixes for its EFIS and GPS/NAV/COMM units.

---

Link to these bug fix change logs? Also link to the software safety requirements for such systems?

quote:

---

How many airline disasters have been linked to malfunctioning software or "sensors?"

---

You tell me.

I would venture a guess that it's a lot fewer than are due to human error or mechanical failure.

quote:

---

And you and I differ greatly in how much credit we give computers (or automation in general).

---

I see that.

Computers are better than humans at one thing: doing the same thing, over and over, exactly as instructed. Driving is very "computable". Keep it between the lines. Do what the sign says. Don't break the rules. Keep your eyes on the fricking road.

We let computers reshape our fricking eyeballs.

Do you look both ways before crossing any railroad tracks, or do you trust that the ding-dings are going to work every time? Or that the traffic lights will never give green both ways at the same time?

How many automated systems do we trust our lives to every day? A lot. And when an automated system kills someone (even if it's the person's fault), it makes big news because it's so rare. People die due to human error constantly.

Some things just beg to be automated. Driving is one of those things.