The interesting part of this article (to me) is not that they made bacteria solve sudoku. What I find interesting is how they solved it:

1) Unlike most sudoku solvers, which use a centralized algorithm. The bacteria use a distributed algorithm: Each individual bacteria cell only knows the contents of cells in their row or column. It's actually a lot more complicated than this though, since there are many bacteria cells for each sudoku square and cells only respond to the first signal they hear from a given position. Given enough bacteria (or time to grow them), the bacteria could brute force a solution (though there appear to be some inherent heuristics that would make a solution probable without the bacteria differentiating into all possible types).

2) The way logic is implemented. They use, what they call a 4C3 leak-switch. This basically is a piece of RNA that codes for 4 different proteins. This piece of RNA can only be transcribed to proteins when there is only one protein left. When the signal is received from another cell, it removes the part of the RNA corresponding to that protein.

3) The communication infrastructure. The bacteria communicate by releasing simple viruses (coded for using the 4C3 leak-switch). These viruses are specialized to only infect bacteria in a certain row or column. When the viruses infect a bacteria they remove the part of the RNA in the 4C3 leak-switch. The viruses are specialized to only infect cells in the corresponding row or column.

The amount of biological power employed in this case is actually rather frightening. This requires the creation of (at least) 16 unique viruses and 16 unique bacteria. Specific receptors for the viruses to bind to the bacteria must have been designed and the protein for both the virus coat and payload transcription need to be tweaked and introduced to the bacteria. A sufficient quantity of each bacteria must have been created.

Usually when you say that the bacteria 'likes' acidity it means that at least one of the proteins it depends on requires the acidity to function. If there are several proteins that are essential for the bacteria to live, the probability that all of the required mutations would occur becomes reasonably small. Additionally, even if the bacteria are able to mutate in such a way to live outside the concrete, they would be poorly adapted to that environment, and would most likely become food for something else. That's not even considering the likelihood that the food source the bacteria uses in its concrete environment may not be available elsewhere.

tl;dr:
The amount of change necessary to go from a bacteria that thrives in concrete to a bacteria that thrives in the lungs is large enough (under the expected conditions) to be considered insurmountable.

Read parent again. Your AVI/WMV/MPGs are all compressed.

If the stenographic data is added after compression it can still be detected. It is likely the encoder that created them left a unique "fingerprint" in the way it encoded the files. Any deviation from this fingerprint between multiple parts of the same file will tell a determined adversary that there is stenographic data hidden in them.

If the stenographic data is added before compression, it must survive the lossy compression. Since these lossy compression algorithms are designed to keep only the subjectively relevant information, the stenographic data will be subjectively relevant (i.e. clearly visible).

I fail to see how rendering a scene at a high framerate would be any more challenging than rendering a complex scene at a lower frame rate. Remember that the hardware either is or is not in use. The ROPs, the shaders, etc. It isn't like there is some magic thing about a simple scene that makes a card work extra hard or something.

The difference is that with complex scenes, the framerate is limited either by the CPU (to calculate AI and physics) or IO (to send commands, textures and meshes to the card). With a single static screen, each frame the program only has to upload a single texture and render a single rectangle. If there is no framerate limit, the CPU will the render command as fast as possible, and the only IO is a command to the card to render. The graphics pipeline has only to render a single texture. The texture reads are spaced in a common access pattern, so there is no contention among processing units for access to GPU memory. This means they can run at full speed. Overall this means that the usual limitations on the GPU pipeline speed don't apply.

My own take on the Fermi paradox comes from the observation that modern radio communication systems - spread spectrum and ODFM - approach the Shannon limit of the bandwidth's information carrying capacity. As they do that, they approach the appearance of pure noise.

Even if the communications look like noise, they will raise the noise floor over the normal cosmic background radiation. No matter how the signal is encoded, the RF power will be detectable. All we would need to look for is bright spots in the sky where the "noise" floor is higher than normal.

I first heard about this site through an email from work. We handle a lot of government contracts, some of which are probably secret (though I'm not involved in any of that). The email was instructing us not to visit the site. That way we could more convincingly "neither confirm nor deny" anything from that site.

Could you define "reason"? The AI field has worked for a while (until the 80s) to build machines that reason. There were some successes with expert systems and things like that (which, given sufficient data could "reason" according to standard definition). The problem with them is that they need complete information, so they never really made it out of the lab. Current works have turned instead of "reasoning" systems to Bayesian inference engines which use complicated statistical methods and approximations to find the most likely answer. They build estimates of probability distributions based on training (equivalent to experience) and then can use them to make decisions or predictions. They are much more flexible than the reasoning machines that were build before and handle incomplete data appropriately.

These probably don't "reason" by your definition, but then again, neither does the human brain: the current understanding of cognition suggests that it is also a inference engine, making probabilistic judgments based on experience in spite of limited information.

If these methods count as AI, then AI already exists. It is used in everything from handwriting and speech recognition to ranking players on XBox Live. If you look at the tasks that AI researchers hoped to solve when research in AI began, a large number of them have been solved. So if a computer can solve a problem that was previously considered an AI problem, wouldn't it be "moving the goal posts" to say that we don't have any AI today?

I think there's a third way that should allow the two groups to coexist. Just give up on science as "the way things are" and instead define it as "the way things seem to be". Science can never tell us the way things work. It can't tell us whether the world is 6000 years old or if there was a man some 2000 years ago that rose from the grave. All science can do is tell us "the world seems to behave according to this model". The model in and of itself is useful for predicting certain things (past, present or future), but can never be guaranteed to provide the facts of the matter (particularly if you assume the existence of an omnipotent being).

That's a really bad idea. Sure that may help in this particular case, but then you're just creating more problems down the road. As long as people are not making a rational evaluation of the issue on an individual level, then these mindless "stampedes" will occur every time something hits an emotional chord in the population. It would be better to simply analyze each issue independently and try to encourage the promotion rational thought instead of emotional reaction.

OT: Speaking of which, maybe we /.'ers should form a political organization to reform education to promote rational thought and critical thinking. If the young earth ID people can infiltrate school boards and rewrite education, then we should be able to too.

That may be right, but it shouldn't be that way. If you have a cell phone and cancel the service, you can still use it to call 911 in emergencies. Why shouldn't OnStar be handled the same way?

Gold is a soft metal. It can neither be eaten or drunk. It is poor for constructing shelters. In and of itself, it provides no net increase in production of useful goods. In essence, gold is only as valuable as people think it is. You might be better off buying diamonds, since at least they can be made into useful tools. Regardless such an investment is not likely to increase in effective value without soft external factors.

Real investments should have some use or purpose that can directly be used to increase wealth. A good example of this is land. Until interstellar travel picks up some, it really is a limited resource. It can be used directly to create food. Water can also be harvested. Shelter can be constructed. If you have more than you need, you can rent it out to others in exchange for something else.
Other less direct example are goods that directly improve productivity such as plows, tractors, fertilizers, power plants, etc. If such items are too expensive for an individual, groups can collectively fund these investments and share the benefits.

The idea of stocks are to collectively fund such investments. The problem is that investments are sometimes difficult to liquidate. That's where stocks come in. The problem is that people are treating stocks as the investment "with buy low and sell high" instead of simply as a tool to liquidate shared investments. The primary gain from investment should (and theoretically must) come from the improved productivity of the investment and increased value of infrastructure. Unfortunately, this effect appears to be lost in the noise of the stock market, only being visible in the long term.

Ah, but if you're the criminal you can get paid twice:
1) find vulnerability
2) sell vulnerability to fraudsters ($$)
3) report vulnerability to google for $$
4) google patches vulnerability so fraudsters can't use it anymore
5) goto 1
6) profit!

No, 768-bit RSA is not broken. they just found the factors for a single number. It only took them 2.5 years, and over 5 terabytes of data too. I don't consider this making 768-bit RSA "broken" any more than 56bit DES is "broken", because they didn't find a way to solve it faster than brute force. The point is that it is now possible to solve this kind of problem. And if they can do it in 2.5 years with 3 supercomputers, a dedicated adversary could probably do it in a few months with a couple dozen.

Factorization is simply finding the prime factors of a number:
For example, the factors of n=21 are p=3 and q=7.

In RSA, I would take these factors and use them to calculate some other things:
I choose e=11 to be my public key exponent (since it is less than and shares no common factors with (7-1)*(3-1)). Then I would calculate my private key exponent "d" such that d*e=1 mod n: for example, d=2 would work.

So you announce n and e publicly. Someone can break your key if they can find d, which is equivalent to finding the two factors of n: p and q. So if I were to tell you 21, you have basically broken my private key once you know 3 and 7 are p and q.
In the article, it's not much different. Basically they found p and q for a 768-bit n.