You are certainly correct about the specialization and the use of abstraction, so yeah I would have to agree on that.
However, improving the quality of communication and the quality of reaction must surely be factors. Agreed, we can only handle limited slices of the complexity, which means even perfect communication won't solve all problems. I probably worded my prior post a little too strongly there, as it only deals with data loss/corruption across a group and group management, it does not deal at all with the limitations of individuals - either their ability to obtain information or process it.
I've explored the idea of "neurologically-driven" education - education designed not to teach specific material but rather to produce the most powerful mind the individual's brain is capable of supporting, but although I can see that buying time (it should skew the bell curve a little to the left), it would merely slow down the trend. In fact, both my "solutions" merely allow you to progress further than you could otherwise before hitting a wall.
It is certainly possible that eugenics will be required to break through said wall. If so, we'd surely want such a decision made by social institutions with the greatest possible ability to understand the consequences of their actions. Here's three options out of the many that could probably be implemented:
Eugenics via DNA option 1: Since sea sponges can organically grow optic fibres, perhaps it would be possible to replace the incredibly slow and inefficient neural connections we currently have with something that has thousands of times the bandwidth and a millionth of the latency
Eugenics via DNA option 2: Octopi arms are "intelligent devices" - they have special-purpose neurological networks that handle how something is to be done, once the central brain decides on what it wants to do. A major limiting factor in the human brain is that it has very limited space to work with and all the autonomous functions are located in the same place as the higher functions AND all the sensory data processing. Alter DNA to create a second brain just for autonomous functions, so that the primary brain can dedicate more neurons to memory and the processing of ideas.
Eugenics via DNA option 3: Combine #1 and #2. The faster networking means you can move the processing of sensory data (a huge chunk of the brain) elsewhere with no loss of performance and no significant latency. Aside from now being a three-brained Cylon, the parts of the brain dealing with executive functions can be considerably larger, perhaps even triple the size.
Ok, two of those are based on improving bandwidth and two are based on offloading functions. There are non-eugenics ways of achieving the latter. Here's two possibilities:
Non-Eugenics option 1: Each person has a computer (protein-based, electronic, doesn't matter) to which they have a wired and/or wireless connection in their brain. Instead of trying to store and process everything in the human brain, thoughts and memories can be offloaded onto the computer and then re-loaded as needed. The computer provides you with virtual memory (so you don't need to have everything in your brain) and with some data management tools (reducing how much you need to process your thoughts in order to comprehend something).
Non-Eugenics option 2: Since space is your limiting factor, move the brain itself into an external enclosure and have the computer in the head. The latency of transmitting even to the other side of the planet is far smaller than the latency inside the brain, so the transmission delays are simply not significant. Bandwidth might be a problem, though. This is not strictly eugenics, since you aren't selecting for or against anything.
Some of these have been explored in sci-fi (non-eugenics option 2 is the approach used in Fred Hoyle's "A for Andromeda") but science is still way too primitive to say which options would actually work in practice -- or how they might actually be implemented if they would work. Socie
Eugenics is widely practiced, even if we happen to call it "genetic screening", "genetic therapy" or "designer babies". You still end up deciding certain genetic lines should not exist. Forced sterilization is also practiced in many countries (including highly civilized ones).
So the taboo is really only in discussing the ethics of such practices and where the lines should be drawn. It is extremely arguable that allowing a child to be born with a genetic disease that will likely be terminal in a relatively short space of time is unethical, but ANY action to prevent such an event (including changing the genetics involved) is apriori selection of what genetic lines are permitted and which ones are not. That is eugenics, no matter what you call it.
An extremely delicate balance will, some day, need to be drawn between the ethical prevention of genuinely functionless suffering and the unethical prevention of individuals who can utilize some aspect of themselves that others would regard as a handicap or suffering. That cannot happen until the taboo on discussion is removed.
Indeed, if we look at other areas of science in which there is a grey area (be it the creation of viruses, or whatever), the underlying issues tend to revolve around poor communication, poor levels of awareness, poor understanding (by scientists and public alike), poor standards of education leading to incorrect analysis, poorly moderated debates, etc.
This is why I often talk about the need for a balanced society, one where the arts, politics and society have evolved as fast as the sciences. When these are all keeping pace with each other, then and only then informed and intelligent discussions become possible, along with informed and intelligent action based on those discussions. As things stand, science and technology outpace the ability to sanity-check it. Since there is no value in slowing science and technology down, everything else must be sped up.
That makes perfect sense to me. However, precisely because it makes perfect sense, and because logarithmic scales work just fine mathematically, I have to dispute TFA's claim that mathematics is not a universal language. Linear scales may not be universal, but there doesn't seem to be a problem with mathematics itself. Even birds can handle basic mathematics (indeed, Alex the parrot is credited with discovering the concept of zero).
The tribe in TFA uses geometrical constructs, just not the same geometrical constructs we use. I found the conclusions drawn (eg: that mathematics isn't universal) to be suspect at best (using "up" and "down" to represent time rather than "forwards" and "backwards" does not mean that they cannot visualize sequential time - it may mean they use either logarithmic time or proportional time rather than linear time, but it's still a system in which all the usual mathematical rules apply).
In fact, a system in which your fixed quantity covers the maximum range you are interested in at that time has definite benefits. It means you are always dealing in easily-managed quantities as opposed to our modern conventions where it is the units that are easily-managed. Indeed, this seems perfectly natural - in early recorded history, the number of "standard" units was extremely large (and not terribly standard), which can be regarded as a half-way house between a fixed number of values of an infinite number of units and an infinite number of values of a fixed number of units.
I see nothing inconsistent, therefore, with what we historically know, or indeed with what we know about mathematics and visualization. (Basic geometry requires only a means of drawing an arc of fixed radius and a straight line, plus the ability to say if two things are equal or not equal. It does not require the ability to measure or quantify either the radius OR the lines in any absolute sense. It doesn't even require the ability to say if something is greater than or less than. Equal or not equal is sufficient. Since the tribe can conceptualize "higher than" and "lower than", it already has more comparative functions than is needed to do basic geometric construction and to make a number of the key discoveries documented by Euclid.)
Or brought it about. Unless you're really good at reading tea-leaves, you cannot possibly know what the probability of a nuclear confrontation with Iran is now versus what it would have been. So far, every country on the US' naughty list that has lacked WMD has been attacked and those on the list that have had WMD have not been attacked. If Stuxnet was indeed an attack, then Iran has recent experience of the former, which lends itself to the idea that it might prefer to be in the latter group.
Possibly. But, then, maybe not. Not everyone who screams loud is a raving lunatic, hell-bent on mutually assured destruction.
There was no intelligence (military or otherwise) prior to Stuxnet that the Iranians were after the bomb -- or that they were not. There is no intelligence (military or otherwise) today that the Iranians are or are not after the bomb. Drawing conclusions from a state of complete ignorance is, well, ignorant. We simply don't know what the hell is going on and all guesses are just that.
Have the master password database at the manufacturer strongly encrypted, then have the password for that database on a couple of smartcards (one for use in recovery, one held elsewhere as a backup in case the first is rendered unusable). The database is only at risk if the smartcard's contents are intercepted by malware on that machine, up to (but not beyond) the point where the database is re-encrypted under a new key. If the machine is properly secured, the risk of this is close to zero.
OR
Have the master password database at the manufacturer off the corporate network. Passwords must be transferred physically from the master password computer to a networked machine in order to be used. Only the keys being used at that instant are ever at risk, the rest of the database is invisible. If the machine is properly secured, the risk of intercepting even the one or two keys exposed is close to zero.
OR
Use a one-time password system. You call up the manufacturer by phone, you read the challenge to them and they read you back what to type in to reset the administrator password. Since this changes each time a connection attempt is made, even if the call is intercepted the password is useless as a new socket connection by an intruder would have a different challenge even if created before the operator typed the answer to the challenge in.
The problem is that manufacturers are part of the precipitate rather than part of the solution.
Still not hard*. Any sequence of multiplications and divisions can be kept in logarithmic notation. Any parentheses can be handled by an intermediate value buffer. Any barrier operation can be handled by having a 1-bit flag per result you're waiting on, where the barrier is cleared when the total value reaches 0 (assuming all flags start at 1 to mark that that operation is blocking). You can build that as a trigger by ORing the bits together and having the NOT of that as your input to a transistor's collector, where the transistor feeds into your trigger.
Early computers (including some for doing computational fluid dynamics) were all hard-wired. They weren't etched into silicon, they were a rat's nest of wires, thermionic vaccuum tubes, mercury delay lines, relays and other such components. It can be done, for digital, analogue and mixed signal. Although mixed signal is a lot harder with the two top people in the field dying in car crashes.
*Not being hard doesn't make it easy. This is a full spectrum, not a binary state. Furthermore, a compiler that can do the layout is very different from a compiler that can do a good layout. Software optimization is frequently demented and software is far easier to optimize. It is also very different from a compiler that can do a layout within specific constraints. Regular optimization is best-effort and a solution (even if not a good one) exists for any given method, but constraint-driven optimization need have no solution (the constraints may make it impossible to solve the problem given the optimization strategies known to the compiler) and you cannot know if that is definitely the case within finite time. You can use herustics and say that you can't guess at a solution within a time limit, but that's not the same as there being no solution.
The problem with designing ASICs or other chips is that you're almost always working with a constraint-driven problem. Fab plants work with dies of fixed size on a wafer of fixed size, in a 2D space. Heat has to be dissipated as evenly as possible, to avoid hot-spots. Pin count is a major determinant of cost and there will be cost constraints. Die size also impacts costs - and rejection rates. Writing a compiler to generate a hardware description that technically works is trivial in comparison to the problem of writing a compiler that can generate a hardware description that you can build deployable hardware from - particularly at a lower cost than needed to just get a good chip designer to hand-craft the damn thing.
That's the real hold-up. The compiler itself isn't the hard part, it's the writing of a compiler that can produce something worth producing at a cost that makes it worth producing that way.
If the stress causes cell replacement (cell death on its own would be insufficient) then that would be sufficient. No further mechanism is required. If the stress only causes cell death, then you need an additional mechanism to explain the cells being replaced. My post was on the assumption that stress would not account for the increase in cell production, only for the increase in cell death. If that assumption is incorrect, then you are entirely correct and nothing else is needed.
Now, even if I am correct, it is unclear if the epigenome would be considered an additional mechanism since it regulates existing mechanisms. It is entirely possible to contend that heightened stress levels cause cell death because it alters the epigenome, thus allowing your description and mine to simultaneously be entirely correct. That's one of the fun things with these sorts of mechanics - not all explanations are mutually exclusive.
The probability of a mutation should, however, drop. It's also important to distinguish mutations and dangerous mutations. Most mutations won't have any harmful effect and a good number (even with zero cruft) won't have any impact at all. As a result, although your calculation is correct for the probability of a mutation hitting something necessary, it's not enough to get an idea of how dangerous the mutation is.
(For example, it's quite possible that a mutation in one of the live viruses embedded in your DNA will cause the cell to become cancerous. After all, the virus code can be hundreds of times more active than any other part of the DNA and that sort of massively accelerated growth is certainly seen on some of the nastier cancers. That virus is certainly not necessary, although in a few cases they generate some useful proteins, though that code can be relocated. Eliminating the viruses should therefore eliminate all the rapid-growth cancers, assuming those to be the only cell mechanisms that allow that kind of mass overproduction.)
However, you are certainly correct in saying that there's a greater likelihood of mutations hitting something critical. To be "safe" (ish) you'd have to couple it with the other suggestion of genetically modifying/rewriting the DNA repair mechanism. If damage can be repaired with greater reliability, you can ensure that dangerous mutations have a high probability of being removed. (Now, the REALLY clever geeks would also ensure that benign and/or potentially beneficial mutations have a lower-than-normal probability of being corrected, biasing what mutations are retained towards that which is fitter and thus accelerating evolution through technology.)
It's likely related. Telomeres don't shorten on their own. One (of several) environmentally-controlled systems in the cells is the epigenome - a string of proteins that controls how DNA is interpreted. It may well be that emotional stress alters the epigenome in areas affecting the immune system and telomeres.
(There's some evidence that highly stressed adult humans are also more susceptible to cancer, and cancer again is linked to both the immune system and the telomere system.)
I think we're going to find that a number of things we've taken for granted as the "right way" for a society to function will prove to be carcinogenic and/or physically toxic. It will be interesting to see if that results in societies changing or whether they deem subjecting carcinogens and toxins on others to be a fundamental freedom (or that people are expendable anyway, or that the science isn't agreed on by 107.3% of all toothpick manufacturers, etc).
The ideal therapy would involve determining the probability of a dangerous mutation then resizing all the telomeres accordingly. You don't want excessively long telomeres (it's an intentional self-destruct mechanism for preventing a cell damaged over time from becoming malignant) just as you don't want telomeres being too short.
Cancer cells are not necessarily ones with over-long telomeres - typically what happens is that the cell's mechanism for shortening the telomeres breaks so that the cell can replicate forever. That doesn't, however, mean that it will or that the replication will occur in a timeframe that's of any significance. You'd have to have additional damage to cell mechanisms for that. If you can modify telomere length on-the-fly, the easiest one is to shorten all the telomeres in a person to something that'll only allow a few copies, then close to the deadline lengthen them just a little. That way, if a cell goes nuts and replicates excessively prior to the telomere system breaking, it'll suicide before it reaches the point of being able to replicate forever.
A better option, though considerably further into the future, would be to modify the repair mechanism in DNA to be rather more reliable. The better-able DNA is at fixing damage, the longer you can make the telomeres without it causing harm. As it stands, the mechanism has limited value. So much so that mtDNA has no such mechanism at all and can handle such a state just fine.
Of course, it helps that mitochondrial DNA is much shorter. The current nucleic DNA is a combination of the original nucleic DNA plus a lot of DNA from symbiotic organisms that became part of the cell and eventually became part of the nucleus, PLUS a great many retroviruses. Perhaps 8-10% of nucleic DNA is from fossil viruses (some still active) and according to recent studies perhaps another 40% is from other external sources.
It aught to be possible to take a fully-sequenced (and I MEAN fully-sequenced) human genome and optimize it. There'll be plenty of genes that belong to fossil lifeforms that serve no useful purpose as far as the human host and the microflora within the host are concerned. (That's over 5,500 lifeforms, so you've got to be very sure of these things.) Decrufting and compacting the human genome would likely reduce the risk of dangerous mutations. It may be that replacing the central DNA core with an XNA core would also help, but I saw nothing in that article about whether XNA molecules have the capacity to unwind properly and replicate, only that XNA had been constructed and was able to carry the same base pairs. This solution is in the FAR future (Star Trek timeframe at best) but there's nothing there that breaks any known rule. We can already do some of the steps, the main reason I'm putting it 500+ years in the future is that the problem space grows exponentially with the number of genes and even quantum computers aren't going to have sufficient power to handle a space that large for a very very long time. If ever. GM is unpredictable enough when adding/deleting single genes, but compacting DNA would involve wholesale rewrites of the genetic code.
I am using Traditional British nomenclature. Meteors and meteorites are defined according to size (meteorites being larger) and not according to whether they land or not. I can't help it if Americans invented their own definitions rather than using the perfectly good ones everyone else uses.
Ok, let's rephrase the experiment. You have four photons - A, B, C, D. A starts off entangled with B, C starts off entangled with D.
What the experiment appears to show is that if B is then entangled with C, then A is effectively entangled with D. In other words, entanglement is transitive. What it does NOT show is a violation of causality, unless I'm seriously misunderstanding the results.
(There may be other alternative explanations, but I'm satisfied that the results can be explained without resorting to violations of causation.)
However, I am going to throw in another thought -- IF it is established that causation is indeed violated, the Many Worlds theory of quantum mechanics must be false. (The Many Worlds theory says that the universe splits at the event, and that the measurement simply tells you which universe you're in - until then, there's a given probability you're in any of the possible universes. However, the event hasn't taken place at the time of the measurement here, so all probability waves must coexist, so you should observe every possible state. This isn't what's observed. Ergo, one or both of Many Worlds and Violation of Causality must be wrong.)
Loops to prevent parallelization can be substituted for with triggers, provided they can be detected. Triggers are easy to do in hardware.
Dividers can be implemented in four steps - a log-table lookup, an inverter, an adder and an inverse log-table lookup. No repeating steps. If you're programming a general-purpose maths core, you'll need a log table anyway. Since that's the only significant consumer of real-estate and it's already allocated for, you lose nothing by using it.
Having the adder interact with the divider is also not very complicated
Option 1: You have an instruction stack that stores one operation and three operands. The first two operands specify the source buffers* and the third operand specifies the target buffer. All three would probably be done as ring buffers and those are easy to do in hardware. The instruction stack would be triggered if and only if all* source buffers are non-empty. The instruction stack does not care if a previous instruction is complete or not, giving you the capacity for parallel execution where appropriate.
*Where you've one input, the other operand would be null. To simplify logic, you'd specify that a null buffer always has content.
Option 2: Each operation has its own independent logic, with the results going into buffers at the end. This time, it is the operations that have the lock and which wait until all the input buffers have content. So the adder will be called but will block because the second input buffer is currently empty. Once the division is complete, the results are copied into the second input buffer. This triggers the add.
There will be others, but these are the two most basic forms. Incredibly simple, incredibly easy to implement. Triggers, latches and buffers take care of almost all the significant hardware-related issues. They are your friends. Well, except when they are your enemies.
....in where it landed. Meteorites are valuable, especially if linkable to a historic event.
In terms of significance, 100,000 tonnes (110,231 tons) of matter falls into Earth's atmosphere every year. This was 70 tonnes. Not a significant fraction of the total mass per year, but still quite respectable. Besides, you probably wouldn't want the yearly quota in one lump sum.
Presumably, though, you could use a source-to-source compiler to convert C (with certain restrictions) into SystemC.* From there, you could do source-to-source compilation to convert SystemC into Verilog or whatever. You'd end up with crappy hardware, but the claim says nothing about design quality only design capability.
*The obvious restriction is that you can't translate something for which no translation exists, whether that's a function call or a particular class of solution.
Going directly from C to hardware without intermediate steps would indeed be a lot harder. But again that's not what the startup promises. They only promise that they can convert C to hardware, they say nothing about how many steps it takes on their end, only what it seems like from your end.
Having said that, a direct C to hardware compiler is obviously possible. A CPU plus software is just emulating a pure hardware system with the code directly built into the design. Instead of replacing bits of circuitry, you replace the instructions which say what circuitry is to be emulated. Since an OS is just another emulator, this time of a particular computer architecture, there is nothing to stop you from taking a dedicated embedded computer, compiling the software, OS and CPU architecture, and getting a single chip that performs the same task(s) entirely in hardware -- no "processor" per-se at all, a true System on a Chip. Albeit rather more complex than most SoC designs currently going, but hey. There's no fun in the easy.
Although there are uses for direct-to-hardware compilers, direct-to-FPGA for pure C would seem better. Take hard drives as an example. You can already install firmware, so there's programmable logic there. What if you could upload the Linux VFS plus applicable filesystems as well? You would reduce CPU load at the very least. If the drive also supported DMA rather than relying on the CPU to pull-and-forward, you could reduce bus activity as well. That would benefit a lot of people and be worth a lot of money for the manufacturer.
This, though, is not worth nearly as much. New hardware isn't designed that often and the number of people designing it is very limited. Faster conversion times won't impact customers, so won't be a selling point to them, so there's no profit involved. Further, optimizing is still a black art, optimizing C compiled into a hardware description language is simply not going to be as good as hand-coding -- for a long time. Eventually, it'll be comparable, just as C compilers are getting close to hand-turned assembly, but it took 30-odd years to get there. True, cheaper engineers can be used, but cheaper doesn't mean better. The issues in hardware are not simply issues of logic and corporations who try to cut corners via C-to-hardware will put their customers through worlds of hurt for at least the next decade to decade and a half.
I don't buy it either. If the author had claimed that people in SE who overspecialize will be find their careers dead in a decade -- that I could believe. Most competent programmers are multilingual and more than a few are polyglots. Many (especially Linux geeks) are not "just" coders, but have strong skills in network administration, systems administration and computer security as well. A decent number will have coding skills not just in one niche area (such as Android applets or Java Servlets) but will have coded a wide range of software types.
The diverse will always survive, the rigid will die off. That is the way of things.
Yes, many of the hacks who specialized only in JBoss Java Servlets or C#.Net will find they are incapable of learning new skills. (Diversity is itself a skill and if you've not learned it you can't make use of it.) A given programming environment probably does have a shelf-life of 10 years or so. Those on that path are of no more use to mainstream society once those software tools have died than flint knappers are.
But so what? They're useless anyway. It's people like that, who can see no further ahead than the next hour, who have turned the global economy into mush. We could do with their total extinction. If Bloomberg can talk such morons into jumping off a bridge (so long as it includes morons in all fields of endeavor), the world would be a lot nicer to live in, progress would become practical, and the population crisis would be resolved.
...an aquifer was found in Africa it was drained dry due to wastage and abuse of resources. This isn't a miracle cure, guys. If used properly, it might reduce the stress on the land (so allowing it to recover, so increasing rainfall) but it is NOT a substitute for surface reservoirs, it is NOT a substitute for learning how to be efficient with resources, it is NOT infinite and it is NOT going to cure centuries (if not millenia) of neglect of Africa.
Your proposals are based on the false assumption that people are doing their job badly
No, I am not. My proposals are based on the provable fact that there is a high level of fraud in research AND on the provable fact that it is impossible to distinguish good research from bad. Your suggestions have no quality control, add no mechanisms for examining if an experiment has indeed been reproduced, and historically have tended to produce people who do nothing at all. My suggestions may not be perfect, but at least I bother to understand what has been tried and why it failed.
And that's a key difference you'll see between my posts and those of my critics -- I acknowledge that my ideas are a starting point, you and the others argue that it's worthless to start at all.
Scientists MUST be free to say a claim is wrong, obtain negative results or otherwise get results corporations aren't going to like.
I did NOT say corporate funding was bad, I SAID corporate funding should be decoupled from research. I am reading what I write, you clearly are not. Either learn to read or bugger off. I see no point in continuing a discussion where you are aiming to find things to object to, even when they aren't there.
I did NOT say there should be no corporate funding, I SAID that researchers should be free to criticize claims. If you wish to debate what I discuss, fine, but if you wish to moronically jabber then I have no interest.
You're not looking at what I'm writing. That case was because journals won't publish negative results, funding may have been nominally public but corporate interests rule current funding bodies so nominal labels should be ignored. The case was also because there's no validation and 90% of the proposal is on multi-layer validation (results, methodology and self-consistency being the three layers validated), making fake results impossible.
Corporations ask questions about areas they are actually interested in. Those questions need answering. Corporations put money into whatever is politically expedient. There will be overlap, but they are NOT the same things. The CFO and the CTO will typically be in a political war with each other and it doesn't get better or saner anywhere else in most companies. It often gets worse. You are assuming companies are rational, but they are actually split-brained and schizophrenic.
People pick causes, yes, and achieve little or nothing. The typical efficiency of donation-funded projects is under 50%, the popular projects are more often feel-good than effective, and people don't usually look beyond feeling good - they don't want to fix things, they don't want to understand things, they want to look and feel like they're doing stuff. It's all smoke and mirrors. Companies are much the same. It's more profitable to file patents without doing the work, since work costs. The less work you can do, the more money you make. At the limit, you do nothing and get everything. That's where the big money is.
Hence decoupling without removing. The money has to be there, the questions have to be there, but it is bad news whenever the money depends on the results being what the managers want rather than what actually is.
That is precisely why abuses occur, and indeed you gave a wonderful example of an abuse (theft of work). There is no means of checking anything. If checking became fundamental and a requirement to publish, you break that pattern. The challenge is to devise a method of checking that cannot simply break the same way as the current system is broken. Pushing the breakage around isn't helpful. The breakage seems to rely on very tightly-coupled dynamics between economics, politics and science, plus no verification of any stage or connection.
I accept my proposal has flaws, perhaps fatal flaws. It attempts to decouple every level and verify every stage, on the theory that inertia and friction in society will try to move back to the status quo, that means you need a large initial impulse and a dynamic that can act as an engine that can maintain momentum.
You're right that nobody gives a flying (that's your inertia) and I don't see how to fix what is broken until that is a disadvantage in all three metrics.
Slashdot Mirror
You are certainly correct about the specialization and the use of abstraction, so yeah I would have to agree on that.
However, improving the quality of communication and the quality of reaction must surely be factors. Agreed, we can only handle limited slices of the complexity, which means even perfect communication won't solve all problems. I probably worded my prior post a little too strongly there, as it only deals with data loss/corruption across a group and group management, it does not deal at all with the limitations of individuals - either their ability to obtain information or process it.
I've explored the idea of "neurologically-driven" education - education designed not to teach specific material but rather to produce the most powerful mind the individual's brain is capable of supporting, but although I can see that buying time (it should skew the bell curve a little to the left), it would merely slow down the trend. In fact, both my "solutions" merely allow you to progress further than you could otherwise before hitting a wall.
It is certainly possible that eugenics will be required to break through said wall. If so, we'd surely want such a decision made by social institutions with the greatest possible ability to understand the consequences of their actions. Here's three options out of the many that could probably be implemented:
Eugenics via DNA option 1: Since sea sponges can organically grow optic fibres, perhaps it would be possible to replace the incredibly slow and inefficient neural connections we currently have with something that has thousands of times the bandwidth and a millionth of the latency
Eugenics via DNA option 2: Octopi arms are "intelligent devices" - they have special-purpose neurological networks that handle how something is to be done, once the central brain decides on what it wants to do. A major limiting factor in the human brain is that it has very limited space to work with and all the autonomous functions are located in the same place as the higher functions AND all the sensory data processing. Alter DNA to create a second brain just for autonomous functions, so that the primary brain can dedicate more neurons to memory and the processing of ideas.
Eugenics via DNA option 3: Combine #1 and #2. The faster networking means you can move the processing of sensory data (a huge chunk of the brain) elsewhere with no loss of performance and no significant latency. Aside from now being a three-brained Cylon, the parts of the brain dealing with executive functions can be considerably larger, perhaps even triple the size.
Ok, two of those are based on improving bandwidth and two are based on offloading functions. There are non-eugenics ways of achieving the latter. Here's two possibilities:
Non-Eugenics option 1: Each person has a computer (protein-based, electronic, doesn't matter) to which they have a wired and/or wireless connection in their brain. Instead of trying to store and process everything in the human brain, thoughts and memories can be offloaded onto the computer and then re-loaded as needed. The computer provides you with virtual memory (so you don't need to have everything in your brain) and with some data management tools (reducing how much you need to process your thoughts in order to comprehend something).
Non-Eugenics option 2: Since space is your limiting factor, move the brain itself into an external enclosure and have the computer in the head. The latency of transmitting even to the other side of the planet is far smaller than the latency inside the brain, so the transmission delays are simply not significant. Bandwidth might be a problem, though. This is not strictly eugenics, since you aren't selecting for or against anything.
Some of these have been explored in sci-fi (non-eugenics option 2 is the approach used in Fred Hoyle's "A for Andromeda") but science is still way too primitive to say which options would actually work in practice -- or how they might actually be implemented if they would work. Socie
Eugenics is widely practiced, even if we happen to call it "genetic screening", "genetic therapy" or "designer babies". You still end up deciding certain genetic lines should not exist. Forced sterilization is also practiced in many countries (including highly civilized ones).
So the taboo is really only in discussing the ethics of such practices and where the lines should be drawn. It is extremely arguable that allowing a child to be born with a genetic disease that will likely be terminal in a relatively short space of time is unethical, but ANY action to prevent such an event (including changing the genetics involved) is apriori selection of what genetic lines are permitted and which ones are not. That is eugenics, no matter what you call it.
An extremely delicate balance will, some day, need to be drawn between the ethical prevention of genuinely functionless suffering and the unethical prevention of individuals who can utilize some aspect of themselves that others would regard as a handicap or suffering. That cannot happen until the taboo on discussion is removed.
Indeed, if we look at other areas of science in which there is a grey area (be it the creation of viruses, or whatever), the underlying issues tend to revolve around poor communication, poor levels of awareness, poor understanding (by scientists and public alike), poor standards of education leading to incorrect analysis, poorly moderated debates, etc.
This is why I often talk about the need for a balanced society, one where the arts, politics and society have evolved as fast as the sciences. When these are all keeping pace with each other, then and only then informed and intelligent discussions become possible, along with informed and intelligent action based on those discussions. As things stand, science and technology outpace the ability to sanity-check it. Since there is no value in slowing science and technology down, everything else must be sped up.
That makes perfect sense to me. However, precisely because it makes perfect sense, and because logarithmic scales work just fine mathematically, I have to dispute TFA's claim that mathematics is not a universal language. Linear scales may not be universal, but there doesn't seem to be a problem with mathematics itself. Even birds can handle basic mathematics (indeed, Alex the parrot is credited with discovering the concept of zero).
The tribe in TFA uses geometrical constructs, just not the same geometrical constructs we use. I found the conclusions drawn (eg: that mathematics isn't universal) to be suspect at best (using "up" and "down" to represent time rather than "forwards" and "backwards" does not mean that they cannot visualize sequential time - it may mean they use either logarithmic time or proportional time rather than linear time, but it's still a system in which all the usual mathematical rules apply).
In fact, a system in which your fixed quantity covers the maximum range you are interested in at that time has definite benefits. It means you are always dealing in easily-managed quantities as opposed to our modern conventions where it is the units that are easily-managed. Indeed, this seems perfectly natural - in early recorded history, the number of "standard" units was extremely large (and not terribly standard), which can be regarded as a half-way house between a fixed number of values of an infinite number of units and an infinite number of values of a fixed number of units.
I see nothing inconsistent, therefore, with what we historically know, or indeed with what we know about mathematics and visualization. (Basic geometry requires only a means of drawing an arc of fixed radius and a straight line, plus the ability to say if two things are equal or not equal. It does not require the ability to measure or quantify either the radius OR the lines in any absolute sense. It doesn't even require the ability to say if something is greater than or less than. Equal or not equal is sufficient. Since the tribe can conceptualize "higher than" and "lower than", it already has more comparative functions than is needed to do basic geometric construction and to make a number of the key discoveries documented by Euclid.)
Or brought it about. Unless you're really good at reading tea-leaves, you cannot possibly know what the probability of a nuclear confrontation with Iran is now versus what it would have been. So far, every country on the US' naughty list that has lacked WMD has been attacked and those on the list that have had WMD have not been attacked. If Stuxnet was indeed an attack, then Iran has recent experience of the former, which lends itself to the idea that it might prefer to be in the latter group.
Possibly. But, then, maybe not. Not everyone who screams loud is a raving lunatic, hell-bent on mutually assured destruction.
There was no intelligence (military or otherwise) prior to Stuxnet that the Iranians were after the bomb -- or that they were not. There is no intelligence (military or otherwise) today that the Iranians are or are not after the bomb. Drawing conclusions from a state of complete ignorance is, well, ignorant. We simply don't know what the hell is going on and all guesses are just that.
Have the master password database at the manufacturer strongly encrypted, then have the password for that database on a couple of smartcards (one for use in recovery, one held elsewhere as a backup in case the first is rendered unusable). The database is only at risk if the smartcard's contents are intercepted by malware on that machine, up to (but not beyond) the point where the database is re-encrypted under a new key. If the machine is properly secured, the risk of this is close to zero.
OR
Have the master password database at the manufacturer off the corporate network. Passwords must be transferred physically from the master password computer to a networked machine in order to be used. Only the keys being used at that instant are ever at risk, the rest of the database is invisible. If the machine is properly secured, the risk of intercepting even the one or two keys exposed is close to zero.
OR
Use a one-time password system. You call up the manufacturer by phone, you read the challenge to them and they read you back what to type in to reset the administrator password. Since this changes each time a connection attempt is made, even if the call is intercepted the password is useless as a new socket connection by an intruder would have a different challenge even if created before the operator typed the answer to the challenge in.
The problem is that manufacturers are part of the precipitate rather than part of the solution.
Still not hard*. Any sequence of multiplications and divisions can be kept in logarithmic notation. Any parentheses can be handled by an intermediate value buffer. Any barrier operation can be handled by having a 1-bit flag per result you're waiting on, where the barrier is cleared when the total value reaches 0 (assuming all flags start at 1 to mark that that operation is blocking). You can build that as a trigger by ORing the bits together and having the NOT of that as your input to a transistor's collector, where the transistor feeds into your trigger.
Early computers (including some for doing computational fluid dynamics) were all hard-wired. They weren't etched into silicon, they were a rat's nest of wires, thermionic vaccuum tubes, mercury delay lines, relays and other such components. It can be done, for digital, analogue and mixed signal. Although mixed signal is a lot harder with the two top people in the field dying in car crashes.
*Not being hard doesn't make it easy. This is a full spectrum, not a binary state. Furthermore, a compiler that can do the layout is very different from a compiler that can do a good layout. Software optimization is frequently demented and software is far easier to optimize. It is also very different from a compiler that can do a layout within specific constraints. Regular optimization is best-effort and a solution (even if not a good one) exists for any given method, but constraint-driven optimization need have no solution (the constraints may make it impossible to solve the problem given the optimization strategies known to the compiler) and you cannot know if that is definitely the case within finite time. You can use herustics and say that you can't guess at a solution within a time limit, but that's not the same as there being no solution.
The problem with designing ASICs or other chips is that you're almost always working with a constraint-driven problem. Fab plants work with dies of fixed size on a wafer of fixed size, in a 2D space. Heat has to be dissipated as evenly as possible, to avoid hot-spots. Pin count is a major determinant of cost and there will be cost constraints. Die size also impacts costs - and rejection rates. Writing a compiler to generate a hardware description that technically works is trivial in comparison to the problem of writing a compiler that can generate a hardware description that you can build deployable hardware from - particularly at a lower cost than needed to just get a good chip designer to hand-craft the damn thing.
That's the real hold-up. The compiler itself isn't the hard part, it's the writing of a compiler that can produce something worth producing at a cost that makes it worth producing that way.
If the stress causes cell replacement (cell death on its own would be insufficient) then that would be sufficient. No further mechanism is required. If the stress only causes cell death, then you need an additional mechanism to explain the cells being replaced. My post was on the assumption that stress would not account for the increase in cell production, only for the increase in cell death. If that assumption is incorrect, then you are entirely correct and nothing else is needed.
Now, even if I am correct, it is unclear if the epigenome would be considered an additional mechanism since it regulates existing mechanisms. It is entirely possible to contend that heightened stress levels cause cell death because it alters the epigenome, thus allowing your description and mine to simultaneously be entirely correct. That's one of the fun things with these sorts of mechanics - not all explanations are mutually exclusive.
The probability of a mutation should, however, drop. It's also important to distinguish mutations and dangerous mutations. Most mutations won't have any harmful effect and a good number (even with zero cruft) won't have any impact at all. As a result, although your calculation is correct for the probability of a mutation hitting something necessary, it's not enough to get an idea of how dangerous the mutation is.
(For example, it's quite possible that a mutation in one of the live viruses embedded in your DNA will cause the cell to become cancerous. After all, the virus code can be hundreds of times more active than any other part of the DNA and that sort of massively accelerated growth is certainly seen on some of the nastier cancers. That virus is certainly not necessary, although in a few cases they generate some useful proteins, though that code can be relocated. Eliminating the viruses should therefore eliminate all the rapid-growth cancers, assuming those to be the only cell mechanisms that allow that kind of mass overproduction.)
However, you are certainly correct in saying that there's a greater likelihood of mutations hitting something critical. To be "safe" (ish) you'd have to couple it with the other suggestion of genetically modifying/rewriting the DNA repair mechanism. If damage can be repaired with greater reliability, you can ensure that dangerous mutations have a high probability of being removed. (Now, the REALLY clever geeks would also ensure that benign and/or potentially beneficial mutations have a lower-than-normal probability of being corrected, biasing what mutations are retained towards that which is fitter and thus accelerating evolution through technology.)
Kaffe was a clean-room Java, so yes it can be done.
It's likely related. Telomeres don't shorten on their own. One (of several) environmentally-controlled systems in the cells is the epigenome - a string of proteins that controls how DNA is interpreted. It may well be that emotional stress alters the epigenome in areas affecting the immune system and telomeres.
(There's some evidence that highly stressed adult humans are also more susceptible to cancer, and cancer again is linked to both the immune system and the telomere system.)
I think we're going to find that a number of things we've taken for granted as the "right way" for a society to function will prove to be carcinogenic and/or physically toxic. It will be interesting to see if that results in societies changing or whether they deem subjecting carcinogens and toxins on others to be a fundamental freedom (or that people are expendable anyway, or that the science isn't agreed on by 107.3% of all toothpick manufacturers, etc).
The ideal therapy would involve determining the probability of a dangerous mutation then resizing all the telomeres accordingly. You don't want excessively long telomeres (it's an intentional self-destruct mechanism for preventing a cell damaged over time from becoming malignant) just as you don't want telomeres being too short.
Cancer cells are not necessarily ones with over-long telomeres - typically what happens is that the cell's mechanism for shortening the telomeres breaks so that the cell can replicate forever. That doesn't, however, mean that it will or that the replication will occur in a timeframe that's of any significance. You'd have to have additional damage to cell mechanisms for that. If you can modify telomere length on-the-fly, the easiest one is to shorten all the telomeres in a person to something that'll only allow a few copies, then close to the deadline lengthen them just a little. That way, if a cell goes nuts and replicates excessively prior to the telomere system breaking, it'll suicide before it reaches the point of being able to replicate forever.
A better option, though considerably further into the future, would be to modify the repair mechanism in DNA to be rather more reliable. The better-able DNA is at fixing damage, the longer you can make the telomeres without it causing harm. As it stands, the mechanism has limited value. So much so that mtDNA has no such mechanism at all and can handle such a state just fine.
Of course, it helps that mitochondrial DNA is much shorter. The current nucleic DNA is a combination of the original nucleic DNA plus a lot of DNA from symbiotic organisms that became part of the cell and eventually became part of the nucleus, PLUS a great many retroviruses. Perhaps 8-10% of nucleic DNA is from fossil viruses (some still active) and according to recent studies perhaps another 40% is from other external sources.
It aught to be possible to take a fully-sequenced (and I MEAN fully-sequenced) human genome and optimize it. There'll be plenty of genes that belong to fossil lifeforms that serve no useful purpose as far as the human host and the microflora within the host are concerned. (That's over 5,500 lifeforms, so you've got to be very sure of these things.) Decrufting and compacting the human genome would likely reduce the risk of dangerous mutations. It may be that replacing the central DNA core with an XNA core would also help, but I saw nothing in that article about whether XNA molecules have the capacity to unwind properly and replicate, only that XNA had been constructed and was able to carry the same base pairs. This solution is in the FAR future (Star Trek timeframe at best) but there's nothing there that breaks any known rule. We can already do some of the steps, the main reason I'm putting it 500+ years in the future is that the problem space grows exponentially with the number of genes and even quantum computers aren't going to have sufficient power to handle a space that large for a very very long time. If ever. GM is unpredictable enough when adding/deleting single genes, but compacting DNA would involve wholesale rewrites of the genetic code.
I am using Traditional British nomenclature. Meteors and meteorites are defined according to size (meteorites being larger) and not according to whether they land or not. I can't help it if Americans invented their own definitions rather than using the perfectly good ones everyone else uses.
Ok, let's rephrase the experiment. You have four photons - A, B, C, D. A starts off entangled with B, C starts off entangled with D.
What the experiment appears to show is that if B is then entangled with C, then A is effectively entangled with D. In other words, entanglement is transitive. What it does NOT show is a violation of causality, unless I'm seriously misunderstanding the results.
(There may be other alternative explanations, but I'm satisfied that the results can be explained without resorting to violations of causation.)
However, I am going to throw in another thought -- IF it is established that causation is indeed violated, the Many Worlds theory of quantum mechanics must be false. (The Many Worlds theory says that the universe splits at the event, and that the measurement simply tells you which universe you're in - until then, there's a given probability you're in any of the possible universes. However, the event hasn't taken place at the time of the measurement here, so all probability waves must coexist, so you should observe every possible state. This isn't what's observed. Ergo, one or both of Many Worlds and Violation of Causality must be wrong.)
Loops to prevent parallelization can be substituted for with triggers, provided they can be detected. Triggers are easy to do in hardware.
Dividers can be implemented in four steps - a log-table lookup, an inverter, an adder and an inverse log-table lookup. No repeating steps. If you're programming a general-purpose maths core, you'll need a log table anyway. Since that's the only significant consumer of real-estate and it's already allocated for, you lose nothing by using it.
Having the adder interact with the divider is also not very complicated
Option 1: You have an instruction stack that stores one operation and three operands. The first two operands specify the source buffers* and the third operand specifies the target buffer. All three would probably be done as ring buffers and those are easy to do in hardware. The instruction stack would be triggered if and only if all* source buffers are non-empty. The instruction stack does not care if a previous instruction is complete or not, giving you the capacity for parallel execution where appropriate.
*Where you've one input, the other operand would be null. To simplify logic, you'd specify that a null buffer always has content.
Option 2: Each operation has its own independent logic, with the results going into buffers at the end. This time, it is the operations that have the lock and which wait until all the input buffers have content. So the adder will be called but will block because the second input buffer is currently empty. Once the division is complete, the results are copied into the second input buffer. This triggers the add.
There will be others, but these are the two most basic forms. Incredibly simple, incredibly easy to implement. Triggers, latches and buffers take care of almost all the significant hardware-related issues. They are your friends. Well, except when they are your enemies.
....in where it landed. Meteorites are valuable, especially if linkable to a historic event.
In terms of significance, 100,000 tonnes (110,231 tons) of matter falls into Earth's atmosphere every year. This was 70 tonnes. Not a significant fraction of the total mass per year, but still quite respectable. Besides, you probably wouldn't want the yearly quota in one lump sum.
Presumably, though, you could use a source-to-source compiler to convert C (with certain restrictions) into SystemC.* From there, you could do source-to-source compilation to convert SystemC into Verilog or whatever. You'd end up with crappy hardware, but the claim says nothing about design quality only design capability.
*The obvious restriction is that you can't translate something for which no translation exists, whether that's a function call or a particular class of solution.
Going directly from C to hardware without intermediate steps would indeed be a lot harder. But again that's not what the startup promises. They only promise that they can convert C to hardware, they say nothing about how many steps it takes on their end, only what it seems like from your end.
Having said that, a direct C to hardware compiler is obviously possible. A CPU plus software is just emulating a pure hardware system with the code directly built into the design. Instead of replacing bits of circuitry, you replace the instructions which say what circuitry is to be emulated. Since an OS is just another emulator, this time of a particular computer architecture, there is nothing to stop you from taking a dedicated embedded computer, compiling the software, OS and CPU architecture, and getting a single chip that performs the same task(s) entirely in hardware -- no "processor" per-se at all, a true System on a Chip. Albeit rather more complex than most SoC designs currently going, but hey. There's no fun in the easy.
Although there are uses for direct-to-hardware compilers, direct-to-FPGA for pure C would seem better. Take hard drives as an example. You can already install firmware, so there's programmable logic there. What if you could upload the Linux VFS plus applicable filesystems as well? You would reduce CPU load at the very least. If the drive also supported DMA rather than relying on the CPU to pull-and-forward, you could reduce bus activity as well. That would benefit a lot of people and be worth a lot of money for the manufacturer.
This, though, is not worth nearly as much. New hardware isn't designed that often and the number of people designing it is very limited. Faster conversion times won't impact customers, so won't be a selling point to them, so there's no profit involved. Further, optimizing is still a black art, optimizing C compiled into a hardware description language is simply not going to be as good as hand-coding -- for a long time. Eventually, it'll be comparable, just as C compilers are getting close to hand-turned assembly, but it took 30-odd years to get there. True, cheaper engineers can be used, but cheaper doesn't mean better. The issues in hardware are not simply issues of logic and corporations who try to cut corners via C-to-hardware will put their customers through worlds of hurt for at least the next decade to decade and a half.
I don't buy it either. If the author had claimed that people in SE who overspecialize will be find their careers dead in a decade -- that I could believe. Most competent programmers are multilingual and more than a few are polyglots. Many (especially Linux geeks) are not "just" coders, but have strong skills in network administration, systems administration and computer security as well. A decent number will have coding skills not just in one niche area (such as Android applets or Java Servlets) but will have coded a wide range of software types.
The diverse will always survive, the rigid will die off. That is the way of things.
Yes, many of the hacks who specialized only in JBoss Java Servlets or C#.Net will find they are incapable of learning new skills. (Diversity is itself a skill and if you've not learned it you can't make use of it.) A given programming environment probably does have a shelf-life of 10 years or so. Those on that path are of no more use to mainstream society once those software tools have died than flint knappers are.
But so what? They're useless anyway. It's people like that, who can see no further ahead than the next hour, who have turned the global economy into mush. We could do with their total extinction. If Bloomberg can talk such morons into jumping off a bridge (so long as it includes morons in all fields of endeavor), the world would be a lot nicer to live in, progress would become practical, and the population crisis would be resolved.
Give the guy a break. He's written more security holes per hour than the rest of us will code in a lifetime. That's gotta be worth something.
...an aquifer was found in Africa it was drained dry due to wastage and abuse of resources. This isn't a miracle cure, guys. If used properly, it might reduce the stress on the land (so allowing it to recover, so increasing rainfall) but it is NOT a substitute for surface reservoirs, it is NOT a substitute for learning how to be efficient with resources, it is NOT infinite and it is NOT going to cure centuries (if not millenia) of neglect of Africa.
No, I am not. My proposals are based on the provable fact that there is a high level of fraud in research AND on the provable fact that it is impossible to distinguish good research from bad. Your suggestions have no quality control, add no mechanisms for examining if an experiment has indeed been reproduced, and historically have tended to produce people who do nothing at all. My suggestions may not be perfect, but at least I bother to understand what has been tried and why it failed.
And that's a key difference you'll see between my posts and those of my critics -- I acknowledge that my ideas are a starting point, you and the others argue that it's worthless to start at all.
I said:
I did NOT say corporate funding was bad, I SAID corporate funding should be decoupled from research. I am reading what I write, you clearly are not. Either learn to read or bugger off. I see no point in continuing a discussion where you are aiming to find things to object to, even when they aren't there.
I did NOT say there should be no corporate funding, I SAID that researchers should be free to criticize claims. If you wish to debate what I discuss, fine, but if you wish to moronically jabber then I have no interest.
You're not looking at what I'm writing. That case was because journals won't publish negative results, funding may have been nominally public but corporate interests rule current funding bodies so nominal labels should be ignored. The case was also because there's no validation and 90% of the proposal is on multi-layer validation (results, methodology and self-consistency being the three layers validated), making fake results impossible.
Corporations ask questions about areas they are actually interested in. Those questions need answering. Corporations put money into whatever is politically expedient. There will be overlap, but they are NOT the same things. The CFO and the CTO will typically be in a political war with each other and it doesn't get better or saner anywhere else in most companies. It often gets worse. You are assuming companies are rational, but they are actually split-brained and schizophrenic.
People pick causes, yes, and achieve little or nothing. The typical efficiency of donation-funded projects is under 50%, the popular projects are more often feel-good than effective, and people don't usually look beyond feeling good - they don't want to fix things, they don't want to understand things, they want to look and feel like they're doing stuff. It's all smoke and mirrors. Companies are much the same. It's more profitable to file patents without doing the work, since work costs. The less work you can do, the more money you make. At the limit, you do nothing and get everything. That's where the big money is.
Hence decoupling without removing. The money has to be there, the questions have to be there, but it is bad news whenever the money depends on the results being what the managers want rather than what actually is.
That is precisely why abuses occur, and indeed you gave a wonderful example of an abuse (theft of work). There is no means of checking anything. If checking became fundamental and a requirement to publish, you break that pattern. The challenge is to devise a method of checking that cannot simply break the same way as the current system is broken. Pushing the breakage around isn't helpful. The breakage seems to rely on very tightly-coupled dynamics between economics, politics and science, plus no verification of any stage or connection.
I accept my proposal has flaws, perhaps fatal flaws. It attempts to decouple every level and verify every stage, on the theory that inertia and friction in society will try to move back to the status quo, that means you need a large initial impulse and a dynamic that can act as an engine that can maintain momentum.
You're right that nobody gives a flying (that's your inertia) and I don't see how to fix what is broken until that is a disadvantage in all three metrics.