The company (Mythic AI) in the article was featured on a Veritasium video[1] about analog computing.
The interesting part for analog is obviously going to be wear and repeatability, because the digital layer auto-corrects the errors until it becomes damaged and unusable clearly, while analog will slide off somewhat gracefully but you need a way to detect it happening multiple layers later.
AI is somewhat tolerant of errors since, the cumulative error across a lot of biases can approach zero if the errors are random.
Mythic shut down recently. Well, technically they still exist on paper, but they're a shell of the former selves. It's a shame, really. Their technology is (was) amazing.
Can you please provide any sort of justification for your (quite ridiculous) statement. We have quite a long history of signal processing migrating to digital precisely because it is better in every regard, INCLUDING POWER. Maaaaaybe, since AI neuromorphic-like circuits are quite fuzzy, they wouldnt require such strong noise resilience and precision, this could be a potential use case, but there is also a point of reliability and testability, which on these required scales for AI will be completely infeasible for analog circuits.
It is fascinating that in almost everything that seeks to emulate nature, eventually we will find that the design used by nature is the most efficient and capable one, likely exceed most things we can come up with.
We can make things "better" in some specific tasks, but for something that do everything a natural design can do, we can't match evolution.
The brain is analog. And it is funny to me that just now we start to think maybe an analog computer is probably better than a digital one to emulate it.
> eventually we will find that the design used by nature is the most efficient and capable one, likely exceed most things we can come up with.
That's a slight exaggeration, but I like the spirit. It shouldn't surprise, as nature has had many more optimization iterations, and doesn't worry about abstraction. The object code is being edited directly; any source code is long gone.
> doesn't worry about abstraction
> The object code is being edited directly; aoy source code is long gone
I was under the impression that there is in fact some abstraction in DNA and RNA. For instance genes, which can be thought of as code that generates proteins (thus, some form of abstraction).
This gets tricky in a way that leads computer folks to their doom in biology.
DNA transcribes to RNA, except DNA is a molecule with molecular interactions and those interactions determine which DNA gets transcribed to RNA (plus there's the whole introns and extrons thing, which, good luck).
RNA transcribes to proteins, except that RNA is also a molecule, and has a tendency to fold into functional forms, including ribosomes which are involved in transcribing RNA.
The amino acid chains that become proteins are molecules - their interactions between each other and the surrounding environment are what cause them to fold into their effective forms (chemical function is determined by molecular form) - except they're not just produced all at once, they're chained one at a time, and it turns out that if you do just create that string of amino acids all in one go, it won't fold into the same shape as it does while being transcribed, so the informatics is interesting but the transcription process is also a factor, and transcription speed is affected by availability of both the amino acids and the TRNA, so even swapping one codon for another that represents the same amino acid can actually affect the protein structure and function.
So, Yes, there's some abstraction, but it'll lead you astray if you think you can ignore the actual physical reality, and it's definitely the case that nature is not ignoring the physical reality and is operating on molecules, not informatics.
Everyone always throws this out there, and I always think, "analogous to what?" But I think it must be in the meaning of continuous signal. But only at a macro or molecular scale could the brain be considered analog, because quantum mechanics is definitely not analog. "Quantum" implies "discrete" and is therefore digital. So the brain is digital at the quantum scale.
Differential equations was the hardest course I've ever taken, I got a B in it but only because it was graded on a curve.. very easy curve..almost the entire class would have failed otherwise because the professor barely spoke English and could not communicate with us in any meaningful way
analog computing is experiencing a surprising comeback?? Article says though being overshadowed by digital computing for decades, analog computing can be useful in solving complex problems that are difficult for digital computers to handle . Scientists are finding that analog computing can help solve problems related to climate change, energy conservation, and artificial intelligence, among other things. This resurgence is surprising to me though, mostly because analog computing is hella expensive (compared to digital computers), difficult to maintain, and requires specialized knowledge to use.
Digital computers are also expensive, require specialized knowledge and are difficult to maintain. In a lot (most?) cases the scales tip in favour of digital solutions, but that's not always the case.
I'm not a specialist but I once saw here an analog computer stimulating a black hole using sound waves. Making a digital simulation would probably be less precise or a lot more expensive.
Zombie? I feel it's always been super alive in the electronic music world, there's tons of people who love analog synths and make really nice music with them. They make great musical instruments, all the qualities like inpreciseness actually shine there.
It's awesome to realize that stuff like cybernetics was ever conceived of before we ever went digital.
I think of analog synthesis - using analog components to create audio waveforms which can be piped into an amplifier and speaker to create sound - as being different from analog computation like it’s described in the article.
Maybe the best way to explain the difference is dedicated purpose-built analog circuits (like synths) vs analog computers that are more general and can be programmed.
Well analog synths were all dominating.
They were a more maintenance.
Tuning and such.
Then over time digital synths became all dominating
for a long time.
The main problem with a vast majority of digital synths
was the lack of button and knobs to create new patches.
Scrolling though a tiny 16x2 screen was a pain in the ass.
Then someone created UIs that ran on the computer and
that was cool
Some digital synths got hardware extensions with knobs.
Then we started to get digital synths with lots of buttons.
The analog or hybrid analog synths started coming out
again.
I personally would not wish to tour with some of those fun
analog monsters. Well perhaps if I had one for each patch
I needed and a tech who changed them out between songs.
Creating sound and sampling it is always an option with
a steep price.
Also noteworthy in this context: BEAM robots, which are mostly analog circuits that implement some behaviour like light-seeking with minimal components:
Any other anecdotes to share regarding the interest in retro technologies from them? I love hearing about this. As a fellow lover of withered technologies, I'm especially interested in hearing the "why"!
I suspect technology for the younger generation who grow up with it from the crib, it gets boring, or too familiar. This may lead to many of them looking to the past for novel or interesting things. Specifically the recent pass. My generation, generation x, did not have this as much. Yeah we did like old cars, probably for same reasons, but not 8-track tapes or laser disc.... because the tech we were being introduced to was still growing and "unfinished".
The recent pass to the new generations, like gen-z, has cool things worth exploring or collecting. As an example I point to vinyl records. Vinyl records are being purchased more and more by the younger folk.
I further suspect, there could be a market for old tech that actually works and can be used, and not just collected. This is a bit hard for a rotary phone, since many houses no longer have land lines. I imagine there might be room for an RJ11 to USB adapter that you can hook up to your phone and connect a real genuine 40 year old rotary phone to it.
The human brain may operate on 20W (?) which is good but doesnt seem relevant.
We have no real idea how it works and we are have not yet invented
the wheel in so far as recreating it artificially or hell even just
harvesting one and programming it.
38 comments
[ 3.8 ms ] story [ 91.9 ms ] threadThe interesting part for analog is obviously going to be wear and repeatability, because the digital layer auto-corrects the errors until it becomes damaged and unusable clearly, while analog will slide off somewhat gracefully but you need a way to detect it happening multiple layers later.
AI is somewhat tolerant of errors since, the cumulative error across a lot of biases can approach zero if the errors are random.
[1] - https://www.youtube.com/watch?v=GVsUOuSjvcg#t=15m44s
Source: https://mythic.ai/whats-new/mythic-raises-13-million-to-brin...
EDIT: I see this was actually mentioned in the article. Well, the new CEO happens to be the former CTO and is quite talented.
We can make things "better" in some specific tasks, but for something that do everything a natural design can do, we can't match evolution.
The brain is analog. And it is funny to me that just now we start to think maybe an analog computer is probably better than a digital one to emulate it.
That's a slight exaggeration, but I like the spirit. It shouldn't surprise, as nature has had many more optimization iterations, and doesn't worry about abstraction. The object code is being edited directly; any source code is long gone.
I was under the impression that there is in fact some abstraction in DNA and RNA. For instance genes, which can be thought of as code that generates proteins (thus, some form of abstraction).
DNA transcribes to RNA, except DNA is a molecule with molecular interactions and those interactions determine which DNA gets transcribed to RNA (plus there's the whole introns and extrons thing, which, good luck).
RNA transcribes to proteins, except that RNA is also a molecule, and has a tendency to fold into functional forms, including ribosomes which are involved in transcribing RNA.
The amino acid chains that become proteins are molecules - their interactions between each other and the surrounding environment are what cause them to fold into their effective forms (chemical function is determined by molecular form) - except they're not just produced all at once, they're chained one at a time, and it turns out that if you do just create that string of amino acids all in one go, it won't fold into the same shape as it does while being transcribed, so the informatics is interesting but the transcription process is also a factor, and transcription speed is affected by availability of both the amino acids and the TRNA, so even swapping one codon for another that represents the same amino acid can actually affect the protein structure and function.
So, Yes, there's some abstraction, but it'll lead you astray if you think you can ignore the actual physical reality, and it's definitely the case that nature is not ignoring the physical reality and is operating on molecules, not informatics.
Everyone always throws this out there, and I always think, "analogous to what?" But I think it must be in the meaning of continuous signal. But only at a macro or molecular scale could the brain be considered analog, because quantum mechanics is definitely not analog. "Quantum" implies "discrete" and is therefore digital. So the brain is digital at the quantum scale.
https://bellmar.medium.com/guess-what-analog-computing-may-b...
https://news.ycombinator.com/item?id=34585958 (99 points | 2 months ago | 84 comments)
---
https://semiengineering.com/can-analog-make-a-comeback/
https://news.ycombinator.com/item?id=32106546 (130 points | 8 months ago | 125 comments)
I'm not a specialist but I once saw here an analog computer stimulating a black hole using sound waves. Making a digital simulation would probably be less precise or a lot more expensive.
It's awesome to realize that stuff like cybernetics was ever conceived of before we ever went digital.
Maybe the best way to explain the difference is dedicated purpose-built analog circuits (like synths) vs analog computers that are more general and can be programmed.
Eno brought in his EMS Synthi AKS, a synthesiser built in a briefcase [4]
[1] Moog Synthesizer : https://en.wikipedia.org/wiki/Moog_synthesizer [1] List of Moog Synthesizer Players : https://en.wikipedia.org/wiki/List_of_Moog_synthesizer_playe... [2] Mellotron : https://en.wikipedia.org/wiki/Mellotron [3] "Heroes" (David Bowie) : https://en.wikipedia.org/wiki/"Heroes"_(David_Bowie_song) [4] EMS (Electronic Music Studios) : https://en.wikipedia.org/wiki/Electronic_Music_Studios [5] Kraftwerk : https://en.wikipedia.org/wiki/KraftwerkThen over time digital synths became all dominating for a long time.
The main problem with a vast majority of digital synths was the lack of button and knobs to create new patches. Scrolling though a tiny 16x2 screen was a pain in the ass.
Then someone created UIs that ran on the computer and that was cool
Some digital synths got hardware extensions with knobs.
Then we started to get digital synths with lots of buttons.
The analog or hybrid analog synths started coming out again.
I personally would not wish to tour with some of those fun analog monsters. Well perhaps if I had one for each patch I needed and a tech who changed them out between songs.
Creating sound and sampling it is always an option with a steep price.
- https://www.computerhistory.org/collections/catalog/X42.79
- https://en.wikipedia.org/wiki/Eurorack
- https://www.analogmuseum.org/english/collection/
- Limited state. That sundial has fixed storage requirements.
- Hard to hack. If it ain't got an IP address, the Elbonians can't tamper with it.
Analog Computer : https://en.wikipedia.org/wiki/Analog_computer
History of Computing Hardware § Analog Computers : https://en.wikipedia.org/wiki/History_of_computing_hardware#...
Timeline of Computing Hardware before 1950 : https://en.wikipedia.org/wiki/Timeline_of_computing_hardware...
(I was an Aegis guy back at the end of the Cold War)
Also noteworthy in this context: BEAM robots, which are mostly analog circuits that implement some behaviour like light-seeking with minimal components:
https://en.wikipedia.org/wiki/BEAM_robotics
https://smfr.org/robots/
https://hackaday.com/tag/beam-robotics/
While BEAM has always been more of a soldering exercise and artform, many interesting principles applicable to general robotics are buried there.
https://en.m.wikipedia.org/wiki/Braitenberg_vehicle
The recent pass to the new generations, like gen-z, has cool things worth exploring or collecting. As an example I point to vinyl records. Vinyl records are being purchased more and more by the younger folk.
I further suspect, there could be a market for old tech that actually works and can be used, and not just collected. This is a bit hard for a rotary phone, since many houses no longer have land lines. I imagine there might be room for an RJ11 to USB adapter that you can hook up to your phone and connect a real genuine 40 year old rotary phone to it.