24 ms·

A Pentagon-funded contest spawned many of today’s self-driving startups

195 points by TrickyRick 9y ago

guicho271828 9y ago
27 years. The first famous "proof of concept" autonomous driving is Navlab (Thorpe, 1990). (Are there any prior projects?)
- Retric 9y ago
  A little more than that. Navlab 1 was built in 1986 using a Chevrolet panel van.[2] The van had 5 racks of computer hardware, including 3 Sun workstations, video hardware and GPS receiver, and a Warp supercomputer.[2] The vehicle suffered from software limitations and was not fully functional until the late 80s, when it achieved its top speed of 20 mph (32 km/h).[2] https://en.wikipedia.org/wiki/Navlab https://en.wikipedia.org/wiki/Navlab
  - grkvlt 9y ago
    The Warp supercomputer is interesting [0] as it's apparently a 'systolic array' computer. Never heard of that architecture before...? 0. https://en.wikipedia.org/wiki/WARP_(systolic_array) https://en.wikipedia.org/wiki/WARP_(systolic_array)
- s-macke 9y ago
  1989, ALVINN, an autonomous land vehicle in a neural network https://www.youtube.com/watch?v=ilP4aPDTBPE&t=32s https://www.youtube.com/watch?v=ilP4aPDTBPE&t=32s
  - f00_ 9y ago
    the paper, for those interested: http://repository.cmu.edu/cgi/viewcontent.cgi?article=2874&context=compsci http://repository.cmu.edu/cgi/viewcontent.cgi?article=2874&c...
- dahauns 9y ago
  There's VaMoRs from Ernst Dickmanns (1986) https://en.wikipedia.org/wiki/Ernst_Dickmanns https://en.wikipedia.org/wiki/Ernst_Dickmanns Already driving on the Autobahn in '87: https://www.youtube.com/watch?v=2EH3R6c7Ufg https://www.youtube.com/watch?v=2EH3R6c7Ufg
- startupdiscuss 9y ago
  Good thread. What was the school affiliation?
cschmidt 9y ago
I would argue that the 2004 DARPA challenge was a failure with the best car, Sandstorm, only going 7 miles before getting stuck on a rock. That led to a do-over in 2005 that was really the first success. https://en.wikipedia.org/wiki/DARPA_Grand_Challenge#2004_Grand_Challenge https://en.wikipedia.org/wiki/DARPA_Grand_Challenge#2004_Gra...
- Fricken 9y ago
  I would argue that no one who has commented in this thread so far has actually read the article.
  - jessriedel 9y ago
    Ha, yes it's especially obvious in this case since the HN-chosen title is opposite the general flow of the article. I think people would be more likely to read the bulk of articles if they actually followed the inverted pyramid structure. Instead, nearly every one begins with a meandering human-interest story to draw the reader in. Whatever happened to those proposals to annotate the web?
bmcusick 9y ago
It was 13 years from the Wright Flyer to the first commercial fixed-wing aircraft. And that was still very dangerous. Jet engines (which really introduced the safety we are used to in airlines) took 50 years.
- byebyetech 9y ago
  This analogy can't work in 21st century. First this is linear thinking, it does not take exponential progress into account. Second, in the case of Autonomous vehicles, the problem is purely software related and does not have any limitations in terms of hardware.
  - tyingq 9y ago
    >does not have any limitations in terms of hardware. That seems a little strong. Lidar tech, for example, seems to still be a differentiator between competitors.
    - mturmon 9y ago
      Good observation. Compounding progress in other sensors (high-resolution stereo cameras, GPS, radar/proximity) has also been important, but none more so than lidar. Also, increased on-board processing power for computer vision/machine learning. In the early 2000s we had nowhere near the onboard compute that we have now.
  - Spooky23 9y ago
    The problems are more than hardware and software. Hardware and software is designed against ethical, legal and liability standards. This is illustrated by the high number of rear end collisions that robot cars are prone to vs human drivers. Experience in the field doesn’t happen exponentially. Development of law relating to these matters doesn’t happen exponentially.
  - Verdex_3 9y ago
    I don't know, it looks like it's holding true so far. If it was 13 years from the Wright brothers to the first commercial flight and now we're seeing 10 years from the "Wright brothers" of self driving cars to so far not having a commercial product. Why do you think we are experiencing exponential progress in the area of self driving cars? Why do you think that the Wright brothers era weren't experiencing exponential progress with respect to the technology they needed to make commercial flight? Finally, with hardware we normally only have to worry with 3 dimensions plus time. I guess fluid dynamics is kind of complicated, but we can at least put the thing into a wind tunnel. With software things can get more complicated because they're not constrained by physics anymore. But at least we have debuggers. With self driving cars we're going to have a support vector machine or a neuro net or something like that running. Some of these techniques can't even be debugged ... they just work or they dont. I'm not sure being free of the limitations of hardware is an asset.
djrogers 9y ago
Proving that one (or something) can drive is a far cry from proving that one can safely drive in a wide variety of changing conditions around other drivers. I would however argue that many people don't actually possess the latter ability, and I think 'robots' will get there before we as a species do.
barnabee 9y ago
In 2007 (when the first iPhone was released), it had been (at least) 15 years since "phones proved they could run apps" [1]. Knowing a limited version of something is possible doesn't tell you a lot about the road to commercialisation and widespread acceptance. [1] https://en.wikipedia.org/wiki/Smartphone#cite_ref-schneidawind_8-0 https://en.wikipedia.org/wiki/Smartphone#cite_ref-schneidawi...
- sandworm101 9y ago
  And proving that something is possible is far from proving that it is actually safe. 15 years after kittyhawk aircraft were widespread and common, but still deathtraps by our modern standards.
tw1010 9y ago
Putting what everyone is saying more abstractly; solving a constraint satisfaction problem in a nice and closed domain should not make us surprised that it isn't as quick to deal with when lifted to an unbounded and messy space.
virgil_disgr4ce 9y ago
I was just telling someone about this the other day. In college at UF, I produced dozens of video segments on engineering student projects, and the DARPA challenge was among the biggest deals, and that was now a decade ago. It wasn't on anybody's radars (so to speak). It's funny how even a little bit of progress suddenly flipped the concept into the popular imagination.
wyldfire 9y ago
There was a great "Nova" episode on the Grand Challenge [1]. [1] http://www.pbs.org/wgbh/nova/darpa/about.html http://www.pbs.org/wgbh/nova/darpa/about.html
ChuckMcM 9y ago
It certainly seems to validate the strategy of "grand challenges" to focus research. One could argue that the X prize helped spawn today's commercial space companies. So does this mean a good way for benevolent billionaires to improve the world would be to sponsor things like this? Certainly Gate's sponsoring the toilets prize was influential in bringing out better sanitation ideas. I wonder if anyone would sponsor a prize for an 'off grid' sustainable living space. People create slums and homeless encampments out of a variety of cast off materials, is there a way to create housing for people which would be safer and enhance their quality of life without also creating a public burden? Perhaps that is a pipe dream but it is something I wonder about.
- rory096 9y ago
  >One could argue that the X prize helped spawn today's commercial space companies. That'd be a stretch. The Ansari X Prize led to a bunch of companies pursuing a goal with little commercial value and few designs with orbital aspirations. The winner led to a costly, dead-end design that has still yet to fly again for suborbital tourism, killed four people along the way, and will never service orbit. The rest of the companies are either dead or vaporware. That said, I'm generally in favor of prize-based challenges to spur development at lower cost, as long as you aren't forced to set a goal that's entirely tangential to what you want out of industry. Newt Gingrich is a notable supporter of the idea, but I'm not aware of any other prominent politicians who support it.
  - ChuckMcM 9y ago
    Agreed it is a stretch, but the argument would be that Virgin Galactic is exactly the follow-on to the rocket that won the x-prize and that Blue Origin's effort came out of their work to win the x-prize but didn't get it done in time. That said, the challenge is a goal that, if accomplished, results in some foundation capability (like vehicle navigation in the 'real' world). I was hoping the robotics challenge that DARPA has been running would have similar benefits but those efforts don't seem nearly as applicable to everyday stuff.
  - jkimmel 9y ago
    Sanders proposed a prize system for pharmaceuticals. It's an interesting alternative to patents for rewarding innovation in that space particularly, which is entirely patent dependent. https://www.govtrack.us/congress/bills/113/s627/text https://www.govtrack.us/congress/bills/113/s627/text
aresant 9y ago
Keeping up with Darpa's challenges is a good way to see into the future a couple of their current projects: 1) A machine-learning competition to overcome scarcity in the radio frequency which is INSANELY fascinating and, if they pull it off, hugely impactful -> https://spectrumcollaborationchallenge.com https://spectrumcollaborationchallenge.com 2) A program to build technology to drive “swarm sprint” exercises to inform tactics and technologies for large groups of unmanned air and ground robots in urban environments. Yes, like how do we build a collaborative air drone army. And PS they are having a hacker fest in the bay in a couple of weeks, registration is closed but bet you can get in if you email via - https://darpahackfest.com https://darpahackfest.com
- vkou 9y ago
  The technologies that will be developed in #2 will be absolutely, unquestionably used by governments to repress their own people.
  - JumpCrisscross 9y ago
    > The technologies that will be developed in #2 will be absolutely, unquestionably used by governments to repress their own people One can go to the dawn of self-driving cars, the Internet, really any potentially game-changing technology, and throw this out. It's an inconsequential assertion that provokes no follow-on thought. Just emotion. Graduating this seed from FUD to a chain of critical thought involves asking how the technology in question might evolve detrimentally and what we can do, today and specifically, preferably in a technical capacity, to mitigate that risk.
    - vkou 9y ago
      Every technology has a spectrum of usefulness, between applications that make people's lives better, and worse. It's not FUD to point out that a particular technology will be primarily used for the latter. We generally don't hold people responsible for direct effects of their work. Working on drones that will kill people for the military is a direct effect. 'Game-changing potential', whatever that means for this kind of work, is not a first order effect.
startupdiscuss 9y ago
It was common lore in CS grad school that there were self driving cars, funded by DARPA, driving around the CMU campus back in the 90s. It was not as easy to look things up. I am glad this story is being told, but I'd love to know about the history and the early days.
- Animats 9y ago
  There was one at CMU back in the 1980s, the NavLab. Autonomous vehicle with a crew of five. Three racks of workstations. Really, really slow.[1] [1] https://www.youtube.com/watch?v=ntIczNQKfjQ https://www.youtube.com/watch?v=ntIczNQKfjQ
  - jacquesm 9y ago
    Compared to the cars in the background it is doing just fine speed wise!
philfreo 9y ago
StartUp's most recent podcast episode is related: https://gimletmedia.com/episode/grand-challenge-season-6-episode-7/ https://gimletmedia.com/episode/grand-challenge-season-6-epi...
vladislav 9y ago
The DARPA grand challenges were an important step for autonomy, by demonstrating basic capabilities and bringing attention and industry funding to the field, but overhyping the expertise gained there may have held the field back over the years. The hardest part of building a self-driving car has proven to be robust perception (processing sensor data to create a very accurate representation of the dynamic environment), which is an aspect that was not heavily emphasized in the challenges: for instance, the winning entry in 2005 had $500K worth of Lidar on it, and the cars were given a GPS map, which allows very accurate localization with yet another expensive sensor (differential GPS). By all accounts I'm aware of, the perception software involved in the challenges was rudimentary, and a lot of the effort was spent on mechanical engineering and path planning. Most importantly, robustness cannot be demonstrated in a one-off competition. Building robust perception is a hard AI problem, and the expertise required to tackle this aspect is orthogonal to the robotics aspects demonstrated in the challenges (which are also obviously very important).
- matt4077 9y ago
  I think this graph from the article supports its thesis pretty well: https://assets.bwbx.io/images/users/iqjWHBFdfxIU/iUXzVqS5lbvg/v9/-1x-1.png https://assets.bwbx.io/images/users/iqjWHBFdfxIU/iUXzVqS5lbv... And that thesis is not, obviously, that the most important problems, or even any problems, were solved by the DARPA challenges. > Building robust perception is a hard AI problem, Oh, I predict, now and here, that something like the following will be posted to HN in 2025 at the latest: "Yes, it's great that my Tesla Model Z lets me browse the internet while commuting, But lets not pretend that this is AI. It's just basic statistics. Also, I really like cozy warmth of Tesla's new line of coal-powered vehicles."
matt4077 9y ago
This video of the 2007 Urban Challenge has some excellent footage and questionable fashion choices: https://www.youtube.com/watch?v=-xibwwNVLgg https://www.youtube.com/watch?v=-xibwwNVLgg (I'd also like to mention my appreciation for Bloomberg for not only publishing such articles as this, but also for actually linking to other interesting content not published by them) Edit: Another find down the youtube rabbit hole: https://youtu.be/DkdESqML41g?t=3h47m30s https://youtu.be/DkdESqML41g?t=3h47m30s has 9 hours(!) of slow-motion robot tragedies.
659087 9y ago
Always-connected self-driving vehicles in many ways seem like something that would be a dream come true for the government and intelligence/law enforcement agencies (not to mention the advertising companies currently developing some of those vehicles).
abalone 9y ago
Ahhh DARPA, that state controlled, taxpayer funded, centrally planned bedrock of free market capitalism.
- hueving 9y ago
  Even if the government entity is taxpayer funded and centrally planned, this is a competition structured like a market. They set out a bid (the winning amount for the goal) and people can compete however they see fit and spend what they want. Similarly, the Army doing RFPs to buy a bunch of toilets is still a free market to compete in as a toilet supplier.
  - abalone 9y ago
    Didn't say it wasn't "like a market" in some ways. But DARPA is very much not free market capitalism. It is state capitalism. Silicon Valley is the product of massive government investment.

195 points by TrickyRick 9y ago

27 years. The first famous "proof of concept" autonomous driving is Navlab (Thorpe, 1990). (Are there any prior projects?)

Retric 9y ago
A little more than that. Navlab 1 was built in 1986 using a Chevrolet panel van.[2] The van had 5 racks of computer hardware, including 3 Sun workstations, video hardware and GPS receiver, and a Warp supercomputer.[2] The vehicle suffered from software limitations and was not fully functional until the late 80s, when it achieved its top speed of 20 mph (32 km/h).[2] https://en.wikipedia.org/wiki/Navlab https://en.wikipedia.org/wiki/Navlab
- grkvlt 9y ago
  The Warp supercomputer is interesting [0] as it's apparently a 'systolic array' computer. Never heard of that architecture before...? 0. https://en.wikipedia.org/wiki/WARP_(systolic_array) https://en.wikipedia.org/wiki/WARP_(systolic_array)
s-macke 9y ago
1989, ALVINN, an autonomous land vehicle in a neural network https://www.youtube.com/watch?v=ilP4aPDTBPE&t=32s https://www.youtube.com/watch?v=ilP4aPDTBPE&t=32s
- f00_ 9y ago
  the paper, for those interested: http://repository.cmu.edu/cgi/viewcontent.cgi?article=2874&context=compsci http://repository.cmu.edu/cgi/viewcontent.cgi?article=2874&c...
dahauns 9y ago
There's VaMoRs from Ernst Dickmanns (1986) https://en.wikipedia.org/wiki/Ernst_Dickmanns https://en.wikipedia.org/wiki/Ernst_Dickmanns Already driving on the Autobahn in '87: https://www.youtube.com/watch?v=2EH3R6c7Ufg https://www.youtube.com/watch?v=2EH3R6c7Ufg
startupdiscuss 9y ago
Good thread. What was the school affiliation?

cschmidt 9y ago

I would argue that the 2004 DARPA challenge was a failure with the best car, Sandstorm, only going 7 miles before getting stuck on a rock. That led to a do-over in 2005 that was really the first success. https://en.wikipedia.org/wiki/DARPA_Grand_Challenge#2004_Grand_Challenge https://en.wikipedia.org/wiki/DARPA_Grand_Challenge#2004_Gra...

Fricken 9y ago
I would argue that no one who has commented in this thread so far has actually read the article.
- jessriedel 9y ago
  Ha, yes it's especially obvious in this case since the HN-chosen title is opposite the general flow of the article. I think people would be more likely to read the bulk of articles if they actually followed the inverted pyramid structure. Instead, nearly every one begins with a meandering human-interest story to draw the reader in. Whatever happened to those proposals to annotate the web?

bmcusick 9y ago

It was 13 years from the Wright Flyer to the first commercial fixed-wing aircraft. And that was still very dangerous. Jet engines (which really introduced the safety we are used to in airlines) took 50 years.

byebyetech 9y ago
This analogy can't work in 21st century. First this is linear thinking, it does not take exponential progress into account. Second, in the case of Autonomous vehicles, the problem is purely software related and does not have any limitations in terms of hardware.
- tyingq 9y ago
  >does not have any limitations in terms of hardware. That seems a little strong. Lidar tech, for example, seems to still be a differentiator between competitors.
  - mturmon 9y ago
    Good observation. Compounding progress in other sensors (high-resolution stereo cameras, GPS, radar/proximity) has also been important, but none more so than lidar. Also, increased on-board processing power for computer vision/machine learning. In the early 2000s we had nowhere near the onboard compute that we have now.
- Spooky23 9y ago
  The problems are more than hardware and software. Hardware and software is designed against ethical, legal and liability standards. This is illustrated by the high number of rear end collisions that robot cars are prone to vs human drivers. Experience in the field doesn’t happen exponentially. Development of law relating to these matters doesn’t happen exponentially.
- Verdex_3 9y ago
  I don't know, it looks like it's holding true so far. If it was 13 years from the Wright brothers to the first commercial flight and now we're seeing 10 years from the "Wright brothers" of self driving cars to so far not having a commercial product. Why do you think we are experiencing exponential progress in the area of self driving cars? Why do you think that the Wright brothers era weren't experiencing exponential progress with respect to the technology they needed to make commercial flight? Finally, with hardware we normally only have to worry with 3 dimensions plus time. I guess fluid dynamics is kind of complicated, but we can at least put the thing into a wind tunnel. With software things can get more complicated because they're not constrained by physics anymore. But at least we have debuggers. With self driving cars we're going to have a support vector machine or a neuro net or something like that running. Some of these techniques can't even be debugged ... they just work or they dont. I'm not sure being free of the limitations of hardware is an asset.

djrogers 9y ago

Proving that one (or something) can drive is a far cry from proving that one can safely drive in a wide variety of changing conditions around other drivers. I would however argue that many people don't actually possess the latter ability, and I think 'robots' will get there before we as a species do.

barnabee 9y ago

In 2007 (when the first iPhone was released), it had been (at least) 15 years since "phones proved they could run apps" [1]. Knowing a limited version of something is possible doesn't tell you a lot about the road to commercialisation and widespread acceptance. [1] https://en.wikipedia.org/wiki/Smartphone#cite_ref-schneidawind_8-0 https://en.wikipedia.org/wiki/Smartphone#cite_ref-schneidawi...

sandworm101 9y ago
And proving that something is possible is far from proving that it is actually safe. 15 years after kittyhawk aircraft were widespread and common, but still deathtraps by our modern standards.

tw1010 9y ago

Putting what everyone is saying more abstractly; solving a constraint satisfaction problem in a nice and closed domain should not make us surprised that it isn't as quick to deal with when lifted to an unbounded and messy space.

virgil_disgr4ce 9y ago

I was just telling someone about this the other day. In college at UF, I produced dozens of video segments on engineering student projects, and the DARPA challenge was among the biggest deals, and that was now a decade ago. It wasn't on anybody's radars (so to speak). It's funny how even a little bit of progress suddenly flipped the concept into the popular imagination.

wyldfire 9y ago

There was a great "Nova" episode on the Grand Challenge [1]. [1] http://www.pbs.org/wgbh/nova/darpa/about.html http://www.pbs.org/wgbh/nova/darpa/about.html

ChuckMcM 9y ago

It certainly seems to validate the strategy of "grand challenges" to focus research. One could argue that the X prize helped spawn today's commercial space companies. So does this mean a good way for benevolent billionaires to improve the world would be to sponsor things like this? Certainly Gate's sponsoring the toilets prize was influential in bringing out better sanitation ideas. I wonder if anyone would sponsor a prize for an 'off grid' sustainable living space. People create slums and homeless encampments out of a variety of cast off materials, is there a way to create housing for people which would be safer and enhance their quality of life without also creating a public burden? Perhaps that is a pipe dream but it is something I wonder about.

rory096 9y ago
>One could argue that the X prize helped spawn today's commercial space companies. That'd be a stretch. The Ansari X Prize led to a bunch of companies pursuing a goal with little commercial value and few designs with orbital aspirations. The winner led to a costly, dead-end design that has still yet to fly again for suborbital tourism, killed four people along the way, and will never service orbit. The rest of the companies are either dead or vaporware. That said, I'm generally in favor of prize-based challenges to spur development at lower cost, as long as you aren't forced to set a goal that's entirely tangential to what you want out of industry. Newt Gingrich is a notable supporter of the idea, but I'm not aware of any other prominent politicians who support it.
- ChuckMcM 9y ago
  Agreed it is a stretch, but the argument would be that Virgin Galactic is exactly the follow-on to the rocket that won the x-prize and that Blue Origin's effort came out of their work to win the x-prize but didn't get it done in time. That said, the challenge is a goal that, if accomplished, results in some foundation capability (like vehicle navigation in the 'real' world). I was hoping the robotics challenge that DARPA has been running would have similar benefits but those efforts don't seem nearly as applicable to everyday stuff.
- jkimmel 9y ago
  Sanders proposed a prize system for pharmaceuticals. It's an interesting alternative to patents for rewarding innovation in that space particularly, which is entirely patent dependent. https://www.govtrack.us/congress/bills/113/s627/text https://www.govtrack.us/congress/bills/113/s627/text

aresant 9y ago

Keeping up with Darpa's challenges is a good way to see into the future a couple of their current projects: 1) A machine-learning competition to overcome scarcity in the radio frequency which is INSANELY fascinating and, if they pull it off, hugely impactful -> https://spectrumcollaborationchallenge.com https://spectrumcollaborationchallenge.com 2) A program to build technology to drive “swarm sprint” exercises to inform tactics and technologies for large groups of unmanned air and ground robots in urban environments. Yes, like how do we build a collaborative air drone army. And PS they are having a hacker fest in the bay in a couple of weeks, registration is closed but bet you can get in if you email via - https://darpahackfest.com https://darpahackfest.com

vkou 9y ago
The technologies that will be developed in #2 will be absolutely, unquestionably used by governments to repress their own people.
- JumpCrisscross 9y ago
  > The technologies that will be developed in #2 will be absolutely, unquestionably used by governments to repress their own people One can go to the dawn of self-driving cars, the Internet, really any potentially game-changing technology, and throw this out. It's an inconsequential assertion that provokes no follow-on thought. Just emotion. Graduating this seed from FUD to a chain of critical thought involves asking how the technology in question might evolve detrimentally and what we can do, today and specifically, preferably in a technical capacity, to mitigate that risk.
  - vkou 9y ago
    Every technology has a spectrum of usefulness, between applications that make people's lives better, and worse. It's not FUD to point out that a particular technology will be primarily used for the latter. We generally don't hold people responsible for direct effects of their work. Working on drones that will kill people for the military is a direct effect. 'Game-changing potential', whatever that means for this kind of work, is not a first order effect.

startupdiscuss 9y ago

It was common lore in CS grad school that there were self driving cars, funded by DARPA, driving around the CMU campus back in the 90s. It was not as easy to look things up. I am glad this story is being told, but I'd love to know about the history and the early days.

Animats 9y ago
There was one at CMU back in the 1980s, the NavLab. Autonomous vehicle with a crew of five. Three racks of workstations. Really, really slow.[1] [1] https://www.youtube.com/watch?v=ntIczNQKfjQ https://www.youtube.com/watch?v=ntIczNQKfjQ
- jacquesm 9y ago
  Compared to the cars in the background it is doing just fine speed wise!

philfreo 9y ago

StartUp's most recent podcast episode is related: https://gimletmedia.com/episode/grand-challenge-season-6-episode-7/ https://gimletmedia.com/episode/grand-challenge-season-6-epi...

vladislav 9y ago

The DARPA grand challenges were an important step for autonomy, by demonstrating basic capabilities and bringing attention and industry funding to the field, but overhyping the expertise gained there may have held the field back over the years. The hardest part of building a self-driving car has proven to be robust perception (processing sensor data to create a very accurate representation of the dynamic environment), which is an aspect that was not heavily emphasized in the challenges: for instance, the winning entry in 2005 had $500K worth of Lidar on it, and the cars were given a GPS map, which allows very accurate localization with yet another expensive sensor (differential GPS). By all accounts I'm aware of, the perception software involved in the challenges was rudimentary, and a lot of the effort was spent on mechanical engineering and path planning. Most importantly, robustness cannot be demonstrated in a one-off competition. Building robust perception is a hard AI problem, and the expertise required to tackle this aspect is orthogonal to the robotics aspects demonstrated in the challenges (which are also obviously very important).

matt4077 9y ago
I think this graph from the article supports its thesis pretty well: https://assets.bwbx.io/images/users/iqjWHBFdfxIU/iUXzVqS5lbvg/v9/-1x-1.png https://assets.bwbx.io/images/users/iqjWHBFdfxIU/iUXzVqS5lbv... And that thesis is not, obviously, that the most important problems, or even any problems, were solved by the DARPA challenges. > Building robust perception is a hard AI problem, Oh, I predict, now and here, that something like the following will be posted to HN in 2025 at the latest: "Yes, it's great that my Tesla Model Z lets me browse the internet while commuting, But lets not pretend that this is AI. It's just basic statistics. Also, I really like cozy warmth of Tesla's new line of coal-powered vehicles."

matt4077 9y ago

This video of the 2007 Urban Challenge has some excellent footage and questionable fashion choices: https://www.youtube.com/watch?v=-xibwwNVLgg https://www.youtube.com/watch?v=-xibwwNVLgg (I'd also like to mention my appreciation for Bloomberg for not only publishing such articles as this, but also for actually linking to other interesting content not published by them) Edit: Another find down the youtube rabbit hole: https://youtu.be/DkdESqML41g?t=3h47m30s https://youtu.be/DkdESqML41g?t=3h47m30s has 9 hours(!) of slow-motion robot tragedies.

659087 9y ago

Always-connected self-driving vehicles in many ways seem like something that would be a dream come true for the government and intelligence/law enforcement agencies (not to mention the advertising companies currently developing some of those vehicles).

abalone 9y ago

Ahhh DARPA, that state controlled, taxpayer funded, centrally planned bedrock of free market capitalism.

hueving 9y ago
Even if the government entity is taxpayer funded and centrally planned, this is a competition structured like a market. They set out a bid (the winning amount for the goal) and people can compete however they see fit and spend what they want. Similarly, the Army doing RFPs to buy a bunch of toilets is still a free market to compete in as a toilet supplier.
- abalone 9y ago
  Didn't say it wasn't "like a market" in some ways. But DARPA is very much not free market capitalism. It is state capitalism. Silicon Valley is the product of massive government investment.