Monday, March 12, 2012

Driverless cars and speech recognition that "work"

"If you're still not convinced that driver-less cars are almost there: http://ping.fm/PAnrz ("The fact that you’re still driving is a bug")"
To which @avernet replied (noting that "almost" is ambiguous):
"@obruchez Almost here, like voice recognition that works?"
My reply (pinpointing another ambiguity):
"@avernet Define "work" and I'll answer. :) But, more seriously, I think that average voice recognition is way harder than average driving."
A definition of "work" is given:
"@obruchez Let's define "work", very informally, as "you and I use it on a regular basis, and are happy with it"."
I was then asked to predict when speech recognition and driverless cars will be "good enough":
"@obruchez With that, in which year would you predict speech recognition and self-driving car will "work"? #longbets"
This is me thinking out loud:
"@avernet This is tricky. I really feel there's a way bigger gap between "working" speech recognition and "really good" speech recognition..."
"@avernet ...than between "working" self-driving cars and "really good" self-driving cars. Let me think about it."
Finally, here's my prediction:
"@avernet I asked my Magic 8-Ball and here's what I got: speech recognition that "works" -> 2018 and self-driving cars that "work" -> 2024."
"@obruchez I'd say speech recognition I'd use is just around the corner, but it's been 15 years I say that, so you might be right with 2018."
"@avernet My reasoning exactly ("just around the corner" for years). Plus speech recognition in French is lagging (and that's what I'd need)."
"@obruchez It might be longer for self-driving car. For legal reasons manufacturers won't risk it until they are *much* better than humans."
"@avernet I agree. But we have "working" prototypes right *now*. And humans are really *bad* at driving (>30000 deaths per year in the US!)."
And here's the reason for this blog post (a tweet by @ebruchez):
"@obruchez @avernet We should be tracking those predictions."
So, see you in 2018 for a debate about the meaning of "regular basis" and "happy", and about the reasons why long-term bets are difficult (for more serious examples, see "How My Predictions Are Faring" by Ray Kurzweil). ;-)

Update (August 28, 2012). See the post "Self-driving cars in 2019, report says" on kurzweilai.net. I think I'm being a bit too pessimistic with my prediction (2024).

Update (September 27, 2012). And yet another optimistic article: "Self-driving cars a reality for 'ordinary people' within 5 years, says Google's Sergey Brin" (within 5 years = by 2017, i.e. seven years before 2024 - my prediction).

Update (January 3, 2014). Another more conservative article: "Fully self-driving cars expected by 2030, says forecast" ("Self-driving cars (SDC) that include driver control are expected to hit highways around the globe before 2025 and self-driving “only” cars (only the car drives) are anticipated around 2030").

2 comments:

hsa said...

While the driverless car sounds cool I would be terrified riding in one

Olivier Bruchez said...

I'm often terrified by humans driving cars... :)