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
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
> 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.
Anything that increases the destructive capacity of armed forces while lowering the manpower requirements is going to make it easier for politicians to misuse those armed forces.
Most obvious example would be the Vietnam disaster. It was eventually shut down because of "bring our boys home". Not because of international reputation, human rights violations or the capability to win or lose the war.
Another example would be how Germany used mechanized forces and air power in the start of WWII as force multipliers. Without sophisticated war machines, attacking France and Great Britain would have been suicidal. Simply because they combined matched German population.
The best bet to mitigate such technology? In those examples it was probably MANPAD that killed helicopter cavalry after Vietnam war. And evolutionary steps of Bazooka helped to keep west Europe from soviet tank invasion during cold war. Those specific inventions increased the manpower requirement while adding relatively small and very specific firepower advantage.
> Anything that increases the destructive capacity of armed forces while lowering the manpower requirements is going to make it easier for politicians to misuse those armed forces
Anything that (a) improves peoples' lives and (b) has a logistical dependency, e.g. requires or becomes better with electricity, an Internet connection, package delivery, et cetera, improves the standing of a centralized power. This will be true as long as those logistical dependencies are subject to economies of scale.
Leaving society in the 18th century, before electricity, antibiotics or daily disposable contact lenses, wasn't easy. But you'd have to give up fewer luxuries to do that then that you would today because there are more luxuries today.
This is why saying "this will be used for evil things that centralized powers do" is so inane. It's a corollary of economic progress.
Except the thing is that geopolitics is the dictating form of politics. Everything else comes after that.
Let's say US government needs lots of manpower to stay as global hegemon. Then say that there is culture in US where people will gladly "defend the land of the free". Now if you mimic a degree of freedom, you get that manpower relatively easily.
Then you can innovate things that connect to the grid as much as you want. As long as the DoD needs grunts, they will give you civil rights and freedom simply because they have to. They probably would do that most of the time because they want to. But given enough time, you will see clusters of bad apples to get office.
It's kinda ironic that US has to be "free" and "powerfull" compared to the enemies. If China opens up their internet bit more, US citizens will get net neutrality without fighting. Also if China produces weapons that take away the technological edge or US, then America needs more manpower to the armed forces. That would lessen political polarization, corruption etc. Nothing unites like common enemy. Except capable common enemy.
All this and you couldn't see that the fault in your logic was that sufficiently enforced rules and regulations could mitigate the usage of evil things that centralized powers do. Obviously that infrastructure minimally exists now, but should we keep running as quickly in a direction that makes it more difficult to create such infrastructure?
I mean there's so much more to life than new technology and the corresponding progress in economies. But beyond that,do you genuinely believe that lawmakers are going to keep up with the necessary rules and regulations to keep sufficiently advanced AI under control in more hedonistic acting entities? Obviously that's the extreme case but it's a good thought experiment to test these models of innovation just for the sake of innovation and money.
As someone whose job is innovating where the current state of the art is, it's getting beyond ridiculous that everyone is ignoring the bloated elephant in the room.
Your bit about Germany isn't correct and is a common misconception. Britain was more powerful than Germany, had a considerably larger GDP, had a vast empire which could be asked (and told) what to do in ways its own population couldn't. The population of the empire is what mattered, not Britain’s. By this measure it was roughly 60 million versus somewhere north of 450 million.
Britain had a colosal and distributed industrial base at the start of the war and both countries had a very similar level of mechanisation early on. Britain only fell behind in armoured vehicle numbers around the time of the fall of France. This loss was very swiftly made up and Britain had roughly 25% more armoured vehicles in service by 1941. In no war year did Germany produce more aircraft than Britain.
Attacking Britain was suicidal and the myth of Britain, alone and standing firm is quite intentional, but untrue. The force multiplier wasn’t the tools, it was how they were used, and it took the allies a while to get the hang of things. The most surprising part is how far Germany got with such a weak position.
Britain’s War Machine by David Edgerton covers this well and somehow avoids it being a dry pile of charts.
I meant that attacking Britain in any way would have been obviously suicidal to the point of being complete deterrent without certain war machines.
If Polad would have been taken with only infantry and field artillery and supplied by railway and horses (like WWI was fought), the conflict would have lasted way way more than one month. If the Polish campaign would have lasted long enough, Germany would not have even attempted to attack France.
Landing on British isles was correctly identified as suicidal. So what did Germany do...
That's partly because railroads in the Soviet Union were of a different gauge then those in Europe - and because the Soviets, as much as they could, pursued a Russian classic - scorched-earth policy.
Mechanized armies needed enormous amounts of supplies. If you can't transport them by rail, and you don't have the diesel to run trucks (the only source of oil the Germans had access to was Romania,) you have to do it by horse.
Horse-drawn armies need enormous amounts of supplies, the energy density of animal feed is a lot lower than that of oil. The British Army learned this lesson in WWI and was totally mechanized by 1939.
Depends where you are located and how large formations of troops you have concentrated on single location. Finnish Defense forces pretty much completely relied on horses during WWII and with great success. There are lot's of tales how Finns were able to supply very hard to get places and also had successful practice to bring back all the dead to be buried in their home towns.
They had to give up horses in the 60's because you could no longer find young men who could drive horse carts. The good mobility of horse in boggy and snowy terrain was only matched in the 80's with Bandvagn 206 carrier.
Yeah, but you can grow animal feed in the fields of occupied France, or take it from Ukranian peasants. You can't really grow petrol on those fields. (Although you can produce synthoil out of coal - which the Germans did.) The UK, the USSR, and the USA were, on the other hand, controlled something like 90% of all of the world's known oil reserves.
Fuel shortages for the fascists in WWII were so bad, in the later years of the war, they used horses to pull fighter planes from their hangars, to their runways. They did not have enough fuel to taxi under their own power.
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.
the same arguments against the drone fleet is the same against the atomic bomb. What something is designed for (death in both cases) is not always their effect. For example, the atomic bombs were used to incredible devastation and death when used but it could be argued they prevented the next war with Russia since a war between two nuclear powers meant total destruction on both sides.
I'll venture into prediction here. Do you think a regional warlord or rogue state ( ISIS or whatever comes after it ) would enter a conflict with another state that was backed by the US? They might, knowing that the US would be unlikely to do more than give airstrike and strategic support. But what if the US had a drone army/fleet that could do close air-support, block to block firefighting, differentiate combatants and non-combatants? It could mean a tougher fight for a group like that since little but money is risked on the side of the US.
Seems to me that at the moment there are a number of state and non state actors who seem happy to pull that tail of the US tiger, knowing that the consequences of the US reacting badly is the significant destruction of their country. Doesn't seem to deter them.
Expecting terrorist orginisations or rogue states to act in a rational way is the same as expecting the US to back down when their pride or national interest is threatened.
Anyone who thinks deeply on the ramifications of the computer and related technology should come to the conclusion that they will inevitably lead to the destruction of individual liberty and total subjugation or annihilation of a significant portion of humanity.
This is not gunpowder, the printing press, or the cotton gin and to compare them you're either naive or being disingenuous.
There is little hope for us to deviate from this path barring significant technological regression due to cataclysmic events or mass enlightenment of the human character. The former is a problem in and of itself and the latter is comically unlikely.
I think that's a very pessimistic view of the world. So far technology empowered as much as it destroyed individual liberties. There is still hope. With the upcoming decentralized technologies, I hope that the power returns to individuals again.
What is the solution? I would rather have automated drone armies in the hands of the US government than anyone else.
The same could be said of the atomic bomb. A weapon made purely for mass scale indiscriminate destruction of humanity. But if not us, then who would we trust to develop such a technology?
Permit my inquiry but whenever i see comments like this, I can't help but think that your main reason for believing in this train of thought is majorly because the negative effects of U.s military industrial complex has pretty much never affected you. because if you were perhaps an Arab who watched his/her home town bombed to shit by a us drone and had the us write off the human casualties that resulted there in from this act as collateral damage you most likely will not feel this way about a single country trying to amass absolute power, particularly when it is a specific country that seem to have a very itchy war finger, which also just elected a very racist man as president.
If the table were turned will you still feel this way, i wonder.
And p.s i am not an Arab, I am African so yes i do sympathize somewhat with what they go through. I also do not hate the U.S but i do feel it that it was time the rest of the world closed the military and technological gap the U.S and the west really had over the rest of the world as this will stop a lot of their bullshit.
I am American, and I hear this from Americans all the time. The reality is far different, though. Does the Russian military perform more humanely in war than the US military? I don't think so. Would the Chinese? Things like My Lai and Abu Ghraib are the exception with the US military, not the rule.
I obviously don't want any single powerful entity to have access to this technology, but I can see the reality of it, which is that someone will have it. Who will it be?
I imagine knowledge over whether those are the exception or the rule for the US military is classified. they don't have to tell you when they've been doing evil
Yet it always comes out. Not because of a bug in code, or a document that fell off of a truck. It’s humans involved that know something is wrong and decide to report it.
The My Lai was reported by the American close air support crew, not a Vietnamese journalist.
The British, Russians and French have these weapons and have never used them. The South Africans developed them and gave them up. The Israelis probably have them but don’t admit to it.
Of these people, only the Americans have used them, and they used them on population centers.
I’m not taking a position on whether this was right or wrong. I’m just pointing out that you are trusting the only government that has ever killed thousands with these weapons and distrusting several governments that have not.
The atomic bomb is a blunt instrument with no finesse. The ability to vaporize large groups of people in a flashy infrastructure destroying display is not particularly conducive to control.
Atomic bombs don't let you watch, record, and analyze the movements of millions of people. They don't let you record and analyze the personal lives, conversations, and secrets of millions of people. They don't eliminate the need for human labor.
Nearly all technology can/will be used for that purpose. I think swarm AI is likely to help people more than it will harm. It's a critical step in big tasks like terraforming Mars.
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.
>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.
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.
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.
Capitalism is a huge motivator by itself. Come out ahead of the competition and reap the benefits, it's a permanent contest. It's not without its downsides (externalization) but it definitely seems to move the needle.
I think some (e.g., Koch Bros) would argue that we don't really have capitalism. That (too often?) Uncle Sam steps in and unbalances the market, often in favor of current winners.
A contest, equally open to all, certainly feels like an appropriate proxy.
yes, but that doesn’t mean it’s the best solution for all problems. Investors/founders want to have their exit strategy. We get a lot of “uber of”, etc, for example.
The goal might not be a marketable solution but a step in the desired direction. I don’t know the limits of this type of prize but it’s worth pursuing:
I do really enjoy the ideas that have come out of the tiny house theme. I don't like the politics it brings out[1]. One of the on going issues is sanitation (Bill Gate's efforts not withstanding), can you provide sanitation in a reliable and difficult to screw up way?
In my notebook I sketched out a possible windmill / pump idea that would pump water to a reservoir which, when full would 'flush' for all of the commodes it served. Challenges are odors and blockages, and a reliable sluice system.
And what would it be like if you let people cook with fire in their space? (lots of small spaces in Brazil and India with simple three stone type cookfires. Is it possible to accommodate something like that without creating an undue fire risk?
They are high-efficiency, low-smoke wood stoves and the company actually tries to make a difference to people that rely on fire for cooking: https://www.bioliteenergy.com/pages/mission
I've had the camp stove for a couple of years and it works really well. I wouldn't recommend using it indoors (the monoxide risk if nothing else), but it definitely fits in with the off-grid movement.
The trouble with your particular scenario is that when it comes to improved efficiency, newer stuff will always win, so that excludes cast-off materials. If being off-grid were cheaper than being on-grid, no one would be on the grid.
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.
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?
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).
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."
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.
Jet travel brings improved speed and comfort, not really improved safety. Prop planes, as a technology, are very safe. It is a lot easier to land a prop with a failed engine or another part -- due to is low speed one can land small props with a high chance of no injuries on unprepared surfaces such as grass fields. For jets, if you have a failed engine and no landing strip you are likely toast (although there are some well publicized exceptions).
The great flight safety numbers of modern jet travel are due to the fact that commercial traffic is heavily regulated end to end, from production to operation, licensing and maintenance. None of this is specific to jet travel, it is just that speed and comfort pushed props mostly out of the business.
You're not really looking at an apples to apples comparison. The last generation of piston engine airliners (the only ones really comparable to the jets that replaced them) used massive turbocharged radial engines that were notoriously complex and unreliable because of all the power being squeezed out of them. Today the vast majority of "prop planes" used in commercial airlines are actually turboprops. Piston engines are basically relegated to small private planes.
It is very counter-intuitive, but speed increases safety in an aircraft type both as a function of hours and as a function of miles. This is an empirical fact that's true for both military and civilian aircraft, I was looking for the US Air Force chart on this, but couldn't seem to find it. Not to discount other factors, but the OP is correct that speed increases safety. Maybe someone with more Google skills can find this chart.
What you describe has nothing to do with jet versus prop, it's just big versus small. Small planes can land in small, unprepared fields, which are common in many parts of the world, whereas large planes can't. You could make a small jet-powered Cessna that would be just as capable of landing in such small places, while offering more safety due to better reliability of the engine. And an airliner-sized prop plane would be no more capable of landing in a hayfield than a 737 would be.
Sure, big vs small is a better way to slice it. In fact fast vs slow is likely a better distinction still. But jets are more practical for fast(er) planes.
Why do you say that jet engines introduced safety? If anything, jet engines were very unsafe for a long time. I would argue better manufacturing, redundant parts, better flight training, and quality assurance that led to the safety I expect when flying today.
Every screw on a Boeing aircraft has a documented specification for the torque used to install it. That’s why I feel comfortable to fly as much as I do.
Pre-Jet flight wasn't exactly a suicide run. A turboprop plane with dual engines was as safe as a jet, fixed for a number of factors.
The larger benefit with jets is being able to fly much higher and faster and being able to hit the 30,000-35,000 sweet spot where the air is thinner, the plane is above storms, and jet fuel efficiency for typical planes is pretty good. The larger 'innovation' on the success of jets is the lowered cost per seat. Props at 10k altitude burned a lot of fuel and were slow which hurt their ability to get more fares. Safety didn't hold back the industry. Cost and speed did.
In fact, turboprop (and turboshaft which is the same thing, but for applications other than fixed-wing aircraft propulsion) is jet engine designed such that it produces mainly mechanical power instead of thrust.
Most commercial airliners use turbofans which are similar in that majority of their thrust does not come from the engine's jet exhaust.
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.
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.
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.
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.
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.
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.
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
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.
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.
(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)
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).
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.
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.
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.
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
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