| General > General Technical Chat |
| Is Arduino killing the electronic hobby? |
| << < (31/31) |
| fourfathom:
--- Quote from: themadhippy on July 07, 2021, 01:21:04 pm ---How do we know other lifeforms aren't trying to communicate with us,just with a method we haven't discovered yet,or even by something that we see regularly and put down to nature,at the moment its raining,we assume thats nature, but could each rain drop be a 1 and the space between drops a 0 and billions of light years away some little alien is sitting there with his water pistol tapping out a message. --- End quote --- The SETI people occasionally get suggestions that they try listening for signals that use modes unknown to present physics (scalar waves, for example). Well golly, we would if we could! Just let me fire up my perpetual zero-point energy power supply, and I'll get right on it. Again, the problem with detecting leakage from TV transmitters, city lights, etc. is that we can do some of that within our solar system, but at interstellar distances we don't have the capability to detect such low levels against the background noise. It's the Shannon limit thing, so unless we already know the characteristics of the to-be-detected signal, and it's amenable to spectral compression, there's not much hope. Frank Drake (and others) would like to have detectors positioned to use the gravitational lensing of our Sun to create a very high-gain, tightly-focused antenna. I don't know, maybe we could aim the lens at one of the near-by stars with known "Goldilocks-zone" planets, and finally pick up "I Love Xyrty" re-runs? I haven't done the gain/noise math. |
| Nominal Animal:
--- Quote from: Zero999 on July 07, 2021, 12:52:35 pm ---Look at it, from the opposite perspective: is it possible to look at the earth's antropogenic EM spectrum and plausibly attribute it to natural causes? --- End quote --- Oh! I misunderstood you, sorry. I find these concepts difficult to convey, so perhaps I can suggest terminology we could use here? (If anyone has better ones, I'm all for adopting those instead.) Signal The presence, absense, or varying of something we can measure. Signal payload Information encoded in the varying aspects of a signal. Any signal, regardless of its spectrum or other features, that we cannot explain originating in natural physical processes, can reasonably be supposed to have an intelligent origin, but the more likely one is that we just don't know enough natural physical processes yet. Mathematically, it is impossible to determine whether noise-like signal payload is uniformly random or simply extremely well compressed. In some signal modulation techniques, we ourselves are coming closer and closer to this, not because we want to be sneaky, but because it gives more bandwidth for the same energy expenditure, with the cost being just some extra computation. At low bandwidths, we are currently not at all good at compression: as mentioned, we still rely on framing and such – the start and stop bits of serial UART communications being a perfect example. At high bandwidths, for example video compression schemes, we're much further on, although there you can still see the effect of our computational limits (including simple binary concatenation of binary tokens, with the size of each unit either inferred from the context or set directly or indirectly by a previous token) – and those do not really use proper compression techniques, they just quantize the data in a lossy manner, and describe those in minimally-small tokens. When the bandwidth is high enough, or there is enough data, we already use compression techniques right now that make it hard to distinguish whether the data is really random or conveys information. (Side note on randomness: One of my favourite pseudo-random number generators is Xorshift64*. It uses three binary shifts (12 right, 25 left, 27 right) and three exclusive-or operations to generate the next 64-bit unsigned integer value in the sequence. This sequence does not contain zero; there are exactly 2⁶⁴-1 valid states, and this is the period of this PRNG. When emitted, the value in the sequence is multiplied by a 62-bit constant (2685821657736338717), with only the 64 least significant bits of the result kept, just to shuffle the output bits somewhat; and this is important to make the sequence look random. Yet, among the standard tests we have to determine whether something is random or not (BigCrush suite of test), that sequence is only detected as being non-random by the MatrixRank test. You do need a somewhat long sequence to meaningfully test it with MatrixRank, though. Compare to e.g. the "industry standard" Mersenne Twister and its MT19937 and MT19937-64 implementations, which fail both MatrixRank and LinearComp tests.) While you can "hide" the signal payload by simply using efficient compression techniques – and you want to do that if you can, because it expands the effective bandwidth; you can either spend less energy to send the same data, or send more data using the same amount of energy –, you cannot hide the signal itself. Thus, it is the interpretation of a signal payload that can be mathematically impossible to determine. If well enough compressed, even a superintelligence cannot determine whether it contains data or is just white noise, is what the math says. Moreover, without being there to see the signal being generated, it may not be possible to rule out intelligent source or a natural physical process as a source, unless we can decode the signal payload to a meaningful message. (And even then it could be pareidolia, or accidentally mistaking a "data generator" for a decompressor. See e.g. how little information current computer games store per character face.) Which gives a very good reason why concentrating on looking for such "message beacons" as fourfathom mentioned, would make sense: on everything else, we need to speculate and choose based on statistics or squishy human emotional reasons. And for now, our dataset is too small for statistics to meaningfully apply. |
| jh15:
So long, and thanks for all the fish. |
| Zero999:
--- Quote from: Nominal Animal on July 09, 2021, 12:25:35 am --- --- Quote from: Zero999 on July 07, 2021, 12:52:35 pm ---Look at it, from the opposite perspective: is it possible to look at the earth's antropogenic EM spectrum and plausibly attribute it to natural causes? --- End quote --- Oh! I misunderstood you, sorry. I find these concepts difficult to convey, so perhaps I can suggest terminology we could use here? (If anyone has better ones, I'm all for adopting those instead.) Signal The presence, absense, or varying of something we can measure. Signal payload Information encoded in the varying aspects of a signal. Any signal, regardless of its spectrum or other features, that we cannot explain originating in natural physical processes, can reasonably be supposed to have an intelligent origin, but the more likely one is that we just don't know enough natural physical processes yet. Mathematically, it is impossible to determine whether noise-like signal payload is uniformly random or simply extremely well compressed. In some signal modulation techniques, we ourselves are coming closer and closer to this, not because we want to be sneaky, but because it gives more bandwidth for the same energy expenditure, with the cost being just some extra computation. At low bandwidths, we are currently not at all good at compression: as mentioned, we still rely on framing and such – the start and stop bits of serial UART communications being a perfect example. At high bandwidths, for example video compression schemes, we're much further on, although there you can still see the effect of our computational limits (including simple binary concatenation of binary tokens, with the size of each unit either inferred from the context or set directly or indirectly by a previous token) – and those do not really use proper compression techniques, they just quantize the data in a lossy manner, and describe those in minimally-small tokens. When the bandwidth is high enough, or there is enough data, we already use compression techniques right now that make it hard to distinguish whether the data is really random or conveys information. (Side note on randomness: One of my favourite pseudo-random number generators is Xorshift64*. It uses three binary shifts (12 right, 25 left, 27 right) and three exclusive-or operations to generate the next 64-bit unsigned integer value in the sequence. This sequence does not contain zero; there are exactly 2⁶⁴-1 valid states, and this is the period of this PRNG. When emitted, the value in the sequence is multiplied by a 62-bit constant (2685821657736338717), with only the 64 least significant bits of the result kept, just to shuffle the output bits somewhat; and this is important to make the sequence look random. Yet, among the standard tests we have to determine whether something is random or not (BigCrush suite of test), that sequence is only detected as being non-random by the MatrixRank test. You do need a somewhat long sequence to meaningfully test it with MatrixRank, though. Compare to e.g. the "industry standard" Mersenne Twister and its MT19937 and MT19937-64 implementations, which fail both MatrixRank and LinearComp tests.) While you can "hide" the signal payload by simply using efficient compression techniques – and you want to do that if you can, because it expands the effective bandwidth; you can either spend less energy to send the same data, or send more data using the same amount of energy –, you cannot hide the signal itself. Thus, it is the interpretation of a signal payload that can be mathematically impossible to determine. If well enough compressed, even a superintelligence cannot determine whether it contains data or is just white noise, is what the math says. Moreover, without being there to see the signal being generated, it may not be possible to rule out intelligent source or a natural physical process as a source, unless we can decode the signal payload to a meaningful message. (And even then it could be pareidolia, or accidentally mistaking a "data generator" for a decompressor. See e.g. how little information current computer games store per character face.) Which gives a very good reason why concentrating on looking for such "message beacons" as fourfathom mentioned, would make sense: on everything else, we need to speculate and choose based on statistics or squishy human emotional reasons. And for now, our dataset is too small for statistics to meaningfully apply. --- End quote --- I understood all of that. Well-compressed data is indistinguishable from random noise, but my point still stands: the spectrum emitted due to human activities is far from random. There are peaks at strange frequencies and at times, which can't be explained by natural phenomena. |
| Nominal Animal:
--- Quote from: Zero999 on July 16, 2021, 07:38:32 am ---I understood all of that. Well-compressed data is indistinguishable from random noise, but my point still stands: the spectrum emitted due to human activities is far from random. There are peaks at strange frequencies and at times, which can't be explained by natural phenomena. --- End quote --- Yes, agreed. That is why I earlier wrote that we can likely only detect technological civilizations similar to ourselves right now. The spectrum emitted due to human activities is non-random due to choices and technology, not because of physics or mathematical limitations (compression). The problem occurs if/when we try to extrapolate from that to "if there were civilizations out there, we'd have heard from them already", because that omits the crucial "similar to ourselves right now". Our technology is advancing very rapidly in this domain, with signal strengths reducing and their spectra becoming more noise-like every year. So it's not just "like ourselves in general", but also "similar to our current technology level". Very important factors when considering the Fermi paradox and the Drake equation, and their implications in the light of current results. (I.e., all we know is that there does not seem to be civilizations similar to ourselves with similar technological capabilities in the close stellar neighborhood right now, although we're not exactly sure of even that because we don't actually know if we could detect ourselves from a few light years away.) |
| Navigation |
| Message Index |
| Previous page |