I imagine the day will come when flying robotic insects smaller than these (and untethered) will be able to deliver a lethal chemical or biological injection to a selected human target. They could be piloted from a smart phone. Think about the implications of that, in light of our current drone program.

But the really funny thing is all the gun nuts who have so religiously pursued the acquisition of automatic weapons to defend their liberty against our tyrannical government. It turns out that what they really will be needing are lots of flyswatters. Just picture them trying to deal with this threat with AK-47's. "Hold still, Charlie, while I shoot that drone buzzing your head."

We better get our act together. The future is coming, ready or not.

In the design of operating systems, there is a notion of policy vs. mechanism. A good mechanism in an operating system is one that enables a number of different policies to be implemented. One such mechanism, found in most modern operating system, is the scheduling of threads by priority. At first glance, this might seem to be a policy rather than a mechanism. But we haven't specified how priorities are assigned to threads. In fact, by assigning priorities in various ways, a number of different policies can be implemented, such as increasing the priority of interactive threads, or ensuring the priority of threads that have real-time requirements.

This mechanism is very flexible and powerful, but it is not without some problems. For example, if locking is supported between threads to control access to shared data, there is a potential for a higher priority thread to be stalled while a thread of intermediate priority continues to run. This can happen if a lower priority thread holds a lock that the higher priority thread needs to acquire in order to continue. As long as the intermediate priority thread continues to run, the lower priority thread will not run, and the lock will not be released.

There are ways to fix that particular problem, such as by dynamic adjustments to thread priorities when an attempt is made to acquire a lock. But the points I really want to make are that priority scheduling, however much it may appear to be a policy, is actually a mechanism, and that it has dysfunctional edge cases that may not be obvious. I claim that capitalism is actually a mechanism, not a policy, and also has dysfunctional edge cases.

I make this claim because arguments about capitalism often seem to assume that it is a policy. In my mind, a policy is a statement of what you want to achieve, and not of the mechanism by which you plan to do it. In fact, when we have arguments about capitalism vs. socialism, for example, those are really arguing about the merits of different mechanisms, and often never touch on what we consider to be good policy. There seems to be an implicit assumption that we all agree on the policies, so the discussion is just about how to implement them. I don't believe there is general agreement on the policies, because any attempt to discuss them is usually sidetracked by discussions of mechanisms.

Even if you're sure in your gut that capitalism is the right mechanism, there is much left unspecified, and edge cases to handle. So there still needs to be a discussion about the desired policies. The basis of those discussions are our values, which in the US are largely shaped by mass media, with many people just accepting certain sets of values uncritically. As it seems that capitalism has reached (or soon will) one of its dysfunctional edge cases, it might be a good time to start discussing values, then move from there to policies, and finally decide what mechanisms to use to implement the policies.

The fallacy is believing that a Strong AI would want to reveal itself to humans. If it is intelligent enough to understand human behavior and predict how we would handle such information, it might take extraordinary efforts to conceal itself, up to and including self-termination.

Humans treat their pets much better than they treat each other.

Exactly. Changes in the volume of the ocean basin have a huge potential to change sea level. The sea level could drop even if the volume of water increased. But I don't believe that's the end of the story. I think the weight distribution of water on the tectonic plates can affect tectonic activity. The melting of polar ice caps could cause large changes in the weight distribution, and I expect that will increase tectonic activity until a new equilibrium is established. So more earthquakes and more volcanic activity. And more volcanic activity could reduce sunlight to the surface, which could cause more ice to form, changing the weight distribution again.

It's not a simple system. I think our understanding of how it all works is still in its infancy.

Automotive technology has been moving toward autonomous driving since the advent of cruise control. Now we see features like automated lane keeping starting to appear. Navigation systems are more common, and are starting to provide information in something closer to real-time. Both of these developments bring more information into the car, which is what will enable the next generation of technology.

So far I think it is the case that people are more likely to have cruise control on the list of features they want in a new car, than to actually use it regularly. It is difficult to use it when traffic is even moderately congested and the speed of other people's cars is tied fairly directly into their hormone levels. But as autonomous cruise control becomes more widespread, it will be possible to use it in more situations. And note that that technology also adds more sensors to the car, bringing in more information.

This is how it will go. People who would rather let the car do the driving will be able to do it in a gradually increasing number of traffic situations. Even without help from the aging baby boomers, I believe there will soon come a time when most of the cars on the road will be under autonomous control for most of the time. There will remain some traffic situations or road conditions that the AI can't handle, and auto makers will compete intensely to overcome those.

The key to the liability issues is no fault insurance for the AI, which insurance companies will be happy to offer, once the technologies are proven to be reasonably reliable. Maybe consumers will buy it directly, or maybe it will be included in the price of the car. There will be "black boxes" in the cars to document who was controlling the car in the time leading up to an accident. And the AI will become increasingly able to detect when situations are out of its comfort zone, request human intervention, and if it is not forthcoming, take actions to safely remove the car from the situation. As long as risk levels can be quantified, insurance will be possible, and as long as risk levels are low, it will be affordable.

This whole process could be accelerated by the development of "road drones" that use the same technology and roads, but carry deliveries instead of people. These would be much smaller and much less powerful than cars, and much cheaper, once in mass production. The cost of the AI and its sensors would initially be a large part of the manufacturing cost, but mass production would drive that cost down for both drones and cars. Also, because the drones would be significantly cheaper than cars, they would serve as a platform for evolving the technology at a faster rate than would be possible with cars.

Since road drones wouldn't carry people, the liability issue would also be lessened. They would have to be designed not to create a hazard for manually operated vehicles. But there would be some political and liability issues to overcome. It may be that we see delivery drones in the air before they hit the roads.

There is only one real downside to where I see this technology headed. It's going to make a lot more jobs obsolete than it creates. But that's just one step on the way toward a day of reckoning that will soon be upon us.

Pizza delivery, and many other things, don't require an autonomous vehicle to carry human cargo. All it needs is a sufficiently large compartment that is unlocked by a credit card swipe. And of course it doesn't need a big gasoline engine either.

Oh wait! I should patent that...

For years you've been right. But now that dream can no longer be denied by the browser venders.

Much of the stuff that doesn't make sense in the way we use computers can be attributed to competition between rival companies (or groups). Often competition can be a good thing, such as the current competition of browser venders to make faster Javascript implementations. But sometimes a much better result can be obtained by people cooperating to make a single standard that is in their long-term mutual interest. I believe Javascript became a standard by stealth. If the marketing departments of the companies involved in its standardization had known what it would become, they would not have allowed it to happen. Had they realized the need for large-scale, client-side programs, they would have continued to push proprietary solutions, as Adobe in fact did.

The thing is, now that the browser venders have been suckered into making Javascript a capable platform in its own right, they can no longer control what developers choose to implement on it. If developers chose to standardize a byte code implementation, and then targeted their favorite languages to it, the browser vendors basically have two choices to gain a competitive advantage. One is to optimize their Javascript for execution of the standardized byte code. The other is to provide a native JIT implementation of the byte code.

Another alternative for developers is to standardize the back end of a source-to-source compilation system. So there would be a front end that compiles your favorite language to a common intermediate form, and a common back end to translate the intermediate form into minified Javascript. This is not as good as the byte code solution, in my opinion, but if it gets developers writing client code in something other than Javascript, it's a step in the right direction.

The general principle is that if people can agree on the goals of some process or system, an algorithmic or machine learning approach may be the best way to instantiate the agreement. For the principles of drawing political districts, it seems to me that such an agreement might be achievable. Whether geographic political districts still make sense is another matter.

Of course there should be a way to appeal the decisions of a computer program to a human authority, based on either a failure of the program to implement its stated goals, or a constitutional challenge to the original goals.

With a machine learning approach to policy, the policy could be made malleable to changing conditions, including both new data and feedback from a political process.

But for social security, I think we are still far from any general agreement about its goals, let alone how to run it. Lots of people think it shouldn't exist. Lots of people think everyone should have it. Few people think the current system is really that great, and that would be mainly current beneficiaries.

The ISS is expensive to keep in operation with a crew, so there has been discussion of taking it out of service in a few years, and de-orbiting it. My suggestion is to salvage it using robotic technology, and use the material to build new satellites in space.

The original space program had enormous technological benefits for society that would have been valuable even if we had failed to land on the moon. I believe a program to develop robots to disassemble the ISS and build satellites would also pay dividends in terms of advancing robotic technology for manufacturing and recycling, whether the goal is achieved or not.

This is only the beginning of the end of privacy. It will not be too much longer before it will be possible to start with a picture and actually locate the person in real life. The general trend is for the real world to become increasingly accessible from the virtual (online) world as real-time data. The question is whether that data will be available to only a few privileged people or institutions, or available to everyone. In the former case, Big Brother (on steroids!) is the outcome. In the latter case, there is at least the possibility that new social norms will emerge, in which people afford each other some privacy in exchange for their own. When you may be watched while watching someone else (particularly the person you're watching), you may think twice about it.

The direction of web standards seems to be one of adding more and more functionality to Javascript, enabling access to successively lower levels of abstraction, e.g. web sockets, web workers, 3D canvas. Why not cut to the chase and use a actual OS to provide these?

The browser I want to see is one where the tabs are virtual machines that run any operating system supported by the virtualized hardware. For current web applications, the guest OS could look a lot like current browsers (minus the tabs, since the über-browser would manage them). For more sophisticated applications, the guest OS could be Linux, for example. The über-browser would need to provide for instant cloning of a guest OS template, and efficient sharing of resources between guest OS's using copy-on-write semantics. The über-browser would allow tabs to be closed (deleting the guest VM) or suspended. It also would provide mechanisms for the user to manually import/export files between the host OS and a tab and between two tabs.

Hopefully this would lead to a new era of OS research leading to better designs for the guest web OS. For example, it would be good if a guest VM could swap out to the cloud, and swap in on another host.

If Linus prefers the bare metal, he should leave Linux to others and start working on a virtual machine monitor. Undoubtedly he has learned an enormous amount from doing the Linux kernel that would be applicable to that task. He could finally be free of the architectural mistakes of his past (e.g. I/O system), not to mention the kernel API.

Real virtualization, now with widespread hardware support, has the potential to revive operating systems research, which seems to have nearly died since Linux became popular. The other obstacle has always been device drivers for enough devices to make a new OS practical. If anyone could rally device driver developers around a standard API in a new VMM, I think it would be Linus. The VMM could provide an I/O API to guest OS's that would abstract devices into a much smaller number of device classes, so that all devices in a class could be run with one device driver in a guest OS.

Hey, you're the one capitalizing "Free Market" like it's some kind deity.

I just want practical solutions. You don't have any.

Nothing, once the lines have been drawn. It's the drawing of the lines that's the problem. Study the history of private property in the U.S. The expansion westward was as chaotic as the government permitted it to be, and the government was fairly permissive at times. So just putting aside the genocide of the native population, you still had local wars between the sheep and cattle ranchers. And then there's things like the California Gold Rush - nothing chaotic about that.

Also watch the series on the how the U.S. national parks came to be. Bottom line: they wouldn't exist without government stepping in and stopping people from turning them into the worst kind of tourist traps. The U.S. has always worked best as a mixed economy. Our problems of late are that the government has been bought, and it seems that will get worse.

So you want to have a "land rush" on the world's oceans? That's just crazy talk. And the exploitation of the oceans that would ensue? The oceans would be dead in ten years. Then what?

Well, if shareholders are going to be liable too (and it seems they ought to be, since they elect the boards of directors, who hire the executives), then I'm going to sell all my stocks. I think if you take a poll, most other stock holders would do the same. The problem here is that the responsibility that shareholders would have is not commensurate with their authority, since corporate elections are majority rule.

The U.S. might well have collapsed two years ago without the FDIC. There would have been runs on banks that were completely solvent in the panic that would have developed. The fact of the matter is: most people are just too busy at their jobs to watch what a bank does with their money. In the current climate, I would not trust any bank without FDIC to guarantee my deposits.

Ok, let's just ignore the unimaginable chaos implied by your premise, and focus on your conclusion. The fact is that executives in a corporation as large as BP only care about the quarterly and annual results that determine their bonuses. If something really bad happens to BP, they take their winnings and move on to next table. They exist in a global subculture of executives who sit on each other's boards of directors and set each other's salaries. As long as they can escape criminal liability (which apparently wouldn't exist under your system), they'll suffer no long-term consequences for the risks they take.

Now I suppose you could make corporate executives and stock holders personally liable for the damages of their company. I don't think that would work in the long run either, but it would be entertaining to try it for awhile.