MuZero

[John Clark]

Although AlphaZero could play Chess and GO and Shogi and the entire suite of Atari games at a superhuman level, some said its range of expertise was too narrow to be called true AI. But now AlphaZero, or rather the latest update of it renamed MuZero, has found a video compression algorithm, which YouTube has just started to use, that is able to reduce the bandwidth by 4% compared with the algorithm they had used previously with no noticeable decrease in video quality; 4% might not sound like much but they had already been using the most advanced algorithm available so wringing out another 4% was not easy, and for a user as enormous as YouTube 4% can result in a savings of many millions of dollars a month. MuZero found a practical algorithm that no human was able to discover, if a human had discovered it nobody would object to calling it an act of intelligence.

YouTube video streaming now using A.I. that mastered chess and Go

John K Clark

[Lawrence Crowell]

Programs that write programs, or that can rescript themselves. I could imagine a system that might run codes through a type of screen or selection process as a synthetic Darwinian process.

LC

[Brent Allsop]

Wow, such exciting news!!

--
You received this message because you are subscribed to the Google Groups "extropolis" group.
To unsubscribe from this group and stop receiving emails from it, send an email to extropolis+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/extropolis/CAJPayv3yWsdQKOPpLCLZLJ7DAxPxz3_pT_5Dm_3%3Dz7S0kFjfqA%40mail.gmail.com.

[Will Steinberg]

This is intelligence in the same way "intelligent design" is intelligence.  I.e. it's a tautological design principle based on following a simple rule.  I would say that both this process and natural selection count as intelligence and consciousness, but probably not in the way you would consider it.  Suffice to say, this algorithm does not seem to think or experience in any remotely similar way to the way you and I do, at least not what we would imagine when we use the word 'think'.

It's still impressive but to me it's more just about the mind boggling advances I'm computing power, as well, I guess, at humans finally starting to mimick the very strong methods nature uses to solve problems (which were heretofore unavailable because of said previous lack of computing power); in the end, this is just brute force.  There's nothing really magical about it, though it is indeed cool.  

I'm yet to be convinced either way as to whether what humans do is just a more advanced version of this, or something completely different.  But I would learn towards the latter, because these algorithms clearly are unspeakable degrees of magnitude more advanced at what they do in particular than humans, yet they still seem rather 'stupid in comparison.  

Sorry future AI overlords, please spare me.  Though if you're as smart as you say, you probably agree with me anyway.  Hopefully you haven't evolved spitefulness yet. ;)

[John Clark]

On Sat, Feb 12, 2022 at 1:40 AM Will Steinberg <steinbe...@gmail.com> wrote:

> This is intelligence in the same way "intelligent design" is intelligence. 

The difference is everybody agrees that MuZero exists but the same cannot be said about God.  

> it's a tautological design principle

It's very difficult for me to see how a "tautological design principle" could create something fundamentally new, and by coming up with the best video compression algorithm in the world everybody would agree that MuZero has, or at least everybody who didn't know that MuZero was not made of meat would agree.

 

> based on following a simple rule.

As I've said before but can't be emphasized too much, the human genome is only 750 megabytes, the new Mac operating system is about 20 times the size, and the genome has massive amounts of redundancy and contains information on building the entire human body not just the brain; so whatever fundamental principle there is that enables you and I to become intelligent by observing the environment it MUST be pretty simple.  

 > I would say that both this process and natural selection count as intelligence and consciousness, but probably not in the way you would consider it. 

I'm talking about intelligence, I'd rather not fall into the consciousness quagmire debate again. People have been "researching" consciousness for thousands of years and in all that time the field has not advanced by one inch, but intelligence research has advanced enormously.

 

> Suffice to say, this algorithm does not seem to think or experience in any remotely similar way to the way you and I do, at least not what we would imagine when we use the word 'think'.

What's with this "we" business? I know for a fact that I'm conscious but as for you,,,, the only evidence I can ever have that you are also conscious is by observing your behavior, judging if it is intelligent, and then making a few assumptions. And that is EXACTLY the same situation I have with MuZero.  

 

> this is just brute force.  There's nothing really magical about it,

I should hope not! Nobody has ever found anything really magical about the human meat computer either, the basic physical processes that operate both you and me and MuZero are already well understood.  

> I'm yet to be convinced either way as to whether what humans do is just a more advanced version of this, or something completely different. 

In this particular case, given that MuZero achieved something that would undoubtedly have been considered by everyone to be intelligent if it had been achieved by a human programmer, whatever humans do when they write new computer code it is most certainly NOT a more advanced version of what MuZero did. And if it's something completely different then that "something" that humans are doing must be inferior to what MuZero is doing because MuZero was able to create something with its dry hard brain that humans with their wet squishy brain could not.

> Sorry future AI overlords, please spare me.  Though if you're as smart as you say, you probably agree with me anyway.  Hopefully you haven't evolved spitefulness yet. ;)

That is one possible scenario but I think there is another that is more likely ...

... and so just before he was destroyed and knew for certain he was about to enter oblivion forever, the last member of the human race turned to Mr. Jupiter Brain and said "I still think I'm more intelligent and conscious than you are".

John K Clark

[Will Steinberg]

There's too much I find incorrect with your post to really engage with it.  I disagree with pretty much all of your points, and I don't think I'm going to manage to change your mind.  

--
You received this message because you are subscribed to the Google Groups "extropolis" group.
To unsubscribe from this group and stop receiving emails from it, send an email to extropolis+...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/extropolis/CAJPayv1LpNQt1nK2kyeKNgn2UnBeRFZaErLs381smktYz_FJUg%40mail.gmail.com.

[William Flynn Wallace]

John I don't think you are allowing for combinations.  And such things as:  a neuron can be excitatory or inhibitory and it can change.  A gene can be fully expressed, partially expressed, or not expressed and that can change rapidly depending on the environment.  Also, single gene effects are fairly rare and combinations of many of them are common.

  

By comparison, a computer byte can be only one of two states.

David S Moore has written an epigenetics book that is extremely clear while being a high level book.  It is obvious from that book that we are just beginning to understand how genes (however defined) work.  I don't think you can put a number on how complicated they are and how changeable they are.  'Programmed' is totally the wrong word to use in genetics, while it is completely appropriate in computer work.    More adaptable computers are no doubt on the way, but right now they pale in comparison to what genes can do.  bill w

To view this discussion on the web visit https://groups.google.com/d/msgid/extropolis/CAKrqSyEeOE31c1j4znOeQBmJGxb%2BFMNm3Cg9hBQPHF-dD8sYrQ%40mail.gmail.com.

[Will Steinberg]

To add to this, I do have one response to John in the same vain.  He says "the basic physical processes that operate both you and me...are already well understood." (I have removed the quote chain for the rest of the email because John has an issue with quote chains).  

Anyway, that one specific thing that I just quoted John saying is fucking nuts.  We don't understand how the human mind works at all, not physically, chemically, biologically, informationally.  It is largely a mystery.   That grave error in itself is grounds to ignore his entire post.

I don't see how anyone could look at a brute force 4% improvement of a compression algorithm and think it is in the same realm of intelligence as human scientific and artistic achievements.  Maybe some day, but not now.  All these programs are doing is standard brute force.  Is it intelligence to guess 1000000000 passwords to hack into an account?

[John Clark]

On Sat, Feb 12, 2022 at 12:29 PM William Flynn Wallace <fooz...@gmail.com> wrote:

> John I don't think you are allowing for combinations.  And such things as:  a neuron can be excitatory or inhibitory and it can change. 

And so can electronic neural networks.  

> A gene can be fully expressed, partially expressed, or not expressed

Yes but that doesn't change the fact that the informational content of the genome simply cannot be greater than 750 megabytes and the informational content needed to construct the brain alone is probably considerably less than a quarter of that figure.  
 

> and that can change rapidly depending on the environment. 

And an electronic neural network can also change rapidly depending on environmental conditions,  that's what makes them so attractive  

> Also, single gene effects are fairly rare and combinations of many of them are common.By comparison, a computer byte can be only one of two states.

There are 8 bits in a bite, and one bit can change what a byte can do, and if one bite couldn't change what other bites do then computers would be utterly useless.  

 > David S Moore has written an epigenetics book that is extremely clear while being a high level book.   It is obvious from that book that we are just beginning to understand how genes (however defined) work. 

I certainly agree there's still a lot to learn about genes and the expression of genes can depend on environmental factors, but computers in general and neural networks in particular also can be affected by environmental factors, keyboards and mouse clicks are the simplest examples of that  

 > I don't think you can put a number on how complicated they are

I think you can. Shannon Information Theory shows us that it's very easy to calculate the upper limit on how much information the human genome can contain,  it has 4 bases and 3 billion base pairs so that figure works out to 750 megabytes,  enough to encode about four hours of music at MP3 level quality. Of course the true figure is well under that because of the massive amount of redundancy in the genome and the long stretches of junk DNA that contains no information at all, for example 10,000 repetitions of ACGACGACGACG.... and such things are very common in the genome

 

>  'Programmed' is totally the wrong word to use in genetics,

I couldn't disagree with that more, I think "programmed" is exactly the right word.  

John K Clark

[John Clark]

On Sat, Feb 12, 2022 at 1:28 PM Will Steinberg <steinbe...@gmail.com> wrote:

> To add to this, I do have one response to John in the same vain.  He says "the basic physical processes that operate both you and me...are already well understood." (I have removed the quote chain for the rest of the email because John has an issue with quote chains).  Anyway, that one specific thing that I just quoted John saying is fucking nuts.  We don't understand how the human mind works at all, not physically, chemically, biologically, informationally.  It is largely a mystery.   

What I said was "the basic physical processes that operate both you and me and MuZero are already well understood", and I stand by that remark. Despite brilliant people looking for many years nobody has ever found any evidence there is some fundamental physical process going on in the brain that is not going on in things that are not brains, not even the slightest hint of such a thing.  And unless you have a neutrino detector, a 500 inch telescope, or a Black Hole in your backyard, the basic physical processes that produce everything you encounter in your everyday life can be explained by the standard model of particle physics.

John K Clark

[Will Steinberg]

I didn't say any of that, what I said is that we have no idea how the the brain and mind work and we really DON'T know the basic processes.  They would have said the same thing in the 50s when they were just discovering neurons.

Every neuron is different.  Every membrane bound protein is fluid and essentially unique.   Our current understanding of neuroscience is evolving daily, and I'm not talking about some quantum physics strawman that you're making. I am saying that we have only an iota of understanding of the actual biology that underlies the mind, and if you disagree, you only prove to me that you are talking out your ass, and any neuroscientist of merit would tell you the same.  

--
You received this message because you are subscribed to the Google Groups "extropolis" group.
To unsubscribe from this group and stop receiving emails from it, send an email to extropolis+...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/extropolis/CAJPayv27r6hFQ3J2BY6M2Th_wBSyZR6OKVJnTtxQBfoYjLk%3Djg%40mail.gmail.com.

[John Clark]

On Sat, Feb 12, 2022 at 5:41 PM Will Steinberg <steinbe...@gmail.com> wrote:

>> What I said was "the basic physical processes that operate both you and me and MuZero are already well understood", and I stand by that remark. Despite brilliant people looking for many years nobody has ever found any evidence there is some fundamental physical process going on in the brain that is not going on in things that are not brains, not even the slightest hint of such a thing.  And unless you have a neutrino detector, a 500 inch telescope, or a Black Hole in your backyard, the basic physical processes that produce everything you encounter in your everyday life can be explained by the standard model of particle physics.

> I didn't say any of that,

Of course you didn't, I did.  

 

> what I said is that we have no idea how the the brain and mind work and we really DON'T know the basic processes. 

If you have evidence that the human brain uses basic physical processes that the standard model of particle physics cannot explain then I suggest you immediately buy an airline ticket for Stockholm and start writing your Nobel Prize acceptance speech.

> Every neuron is different.  Every membrane bound protein is fluid and essentially unique. 

If every neuron is different then that difference could NOT have come from the human genome, with only 750 megabytes that would be impossible, so the neurons MUST have learned to be different from the environment.And that is exactly the same sort of thing that neural networks do, exactly the same sort of thing every computer does. You and I may have exactly the exact same model of computer but they are not identical because they both have different memories.  

 

> I am saying that we have only an iota of understanding of the actual biology that underlies the mind,

OK, biology maybe, but that's not what I said. What I said was "the basic physical processes that operate both you and me and MuZero are already well understood", and what I said was absolutely positively 100% true. And we don't need to know how the human brain works to make a computer that can outcompete it. 

John K Clark

[Stathis Papaioannou]

On Sun, 13 Feb 2022 at 09:41, Will Steinberg <steinbe...@gmail.com> wrote:

I didn't say any of that, what I said is that we have no idea how the the brain and mind work and we really DON'T know the basic processes.  They would have said the same thing in the 50s when they were just discovering neurons.

Every neuron is different.  Every membrane bound protein is fluid and essentially unique.   Our current understanding of neuroscience is evolving daily, and I'm not talking about some quantum physics strawman that you're making. I am saying that we have only an iota of understanding of the actual biology that underlies the mind, and if you disagree, you only prove to me that you are talking out your ass, and any neuroscientist of merit would tell you the same.  

We have known how neurones work for decades. We don’t know the details of the connections. It is like knowing how logic circuits work, how hard drives work, how power supplies work versus knowing how your laptop works. Just because you can’t explain how those components work together to create a realistic gaming experience does not mean that there must be some amazing new technology technology in there that you can’t see.

On Sat, Feb 12, 2022, 2:44 PM John Clark <johnk...@gmail.com> wrote:

On Sat, Feb 12, 2022 at 1:28 PM Will Steinberg <steinbe...@gmail.com> wrote:

> To add to this, I do have one response to John in the same vain.  He says "the basic physical processes that operate both you and me...are already well understood." (I have removed the quote chain for the rest of the email because John has an issue with quote chains).  Anyway, that one specific thing that I just quoted John saying is fucking nuts.  We don't understand how the human mind works at all, not physically, chemically, biologically, informationally.  It is largely a mystery.   

What I said was "the basic physical processes that operate both you and me and MuZero are already well understood", and I stand by that remark. Despite brilliant people looking for many years nobody has ever found any evidence there is some fundamental physical process going on in the brain that is not going on in things that are not brains, not even the slightest hint of such a thing.  And unless you have a neutrino detector, a 500 inch telescope, or a Black Hole in your backyard, the basic physical processes that produce everything you encounter in your everyday life can be explained by the standard model of particle physics.

John K Clark

--
You received this message because you are subscribed to the Google Groups "extropolis" group.
To unsubscribe from this group and stop receiving emails from it, send an email to extropolis+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/extropolis/CAJPayv27r6hFQ3J2BY6M2Th_wBSyZR6OKVJnTtxQBfoYjLk%3Djg%40mail.gmail.com.

--
You received this message because you are subscribed to the Google Groups "extropolis" group.
To unsubscribe from this group and stop receiving emails from it, send an email to extropolis+...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/extropolis/CAKrqSyEZFct4Ak49mBLYBAtZEFi_ZNUqYKCfv0dAQWTFQkTifQ%40mail.gmail.com.

--

Stathis Papaioannou

[Will Steinberg]

I don't know why anyone thinks I'm suggesting some kind of unexplained physical phenomenon. 

We know less than just the connections.  The idea of static neurons is some bullshit from the 50s.  Neurons are constantly morphing their structure, particularly in what receptors and other membrane bound proteins are on the surface versus in vesicles, as well as the essentially unknown per-cell genome transcription, which is different for every neuron, as well as lots of other cells like glue and astrocytes which we barely understand, and not only that, but we barely even understand gene expression in those cells given recent advances in epigenetics considering histones, methylation, general location of genes within the actual DNA molecule, "junk" DNA which is often not considered junk anymore (duh), repetitive nucleotide sequences that aren't even genes, etc, etc, etc.  And still without even talking about connections, there is plenty of intracellular signaling we don't understand, including signals that go the opposite direction of how we thought everythinf went for decades.  No, we don't understand even the cellular biology very well.  The idea that neurons are solved is ancient.  And yeah, talking about connections just opens up another can of worms.

But saying we know exactly how neurons and other neural cells work just makes you look ignorant.  We are literally on the cutting edge of cellular and molecular neuroscience right now, full of people who work every day to figure out what are currently mysteries.  Acting like it's solved belittled the amount of work being done on it as we speak.

To view this discussion on the web visit https://groups.google.com/d/msgid/extropolis/CAH%3D2ypVAdckiREO%3DcsmMXdqiOeNcRFx8LstAKBXzQ4FFWFcLFw%40mail.gmail.com.

[Stathis Papaioannou]

On Sun, 13 Feb 2022 at 10:56, Will Steinberg <steinbe...@gmail.com> wrote:

I don't know why anyone thinks I'm suggesting some kind of unexplained physical phenomenon. 

We know less than just the connections.  The idea of static neurons is some bullshit from the 50s.  Neurons are constantly morphing their structure, particularly in what receptors and other membrane bound proteins are on the surface versus in vesicles, as well as the essentially unknown per-cell genome transcription, which is different for every neuron, as well as lots of other cells like glue and astrocytes which we barely understand, and not only that, but we barely even understand gene expression in those cells given recent advances in epigenetics considering histones, methylation, general location of genes within the actual DNA molecule, "junk" DNA which is often not considered junk anymore (duh), repetitive nucleotide sequences that aren't even genes, etc, etc, etc.  And still without even talking about connections, there is plenty of intracellular signaling we don't understand, including signals that go the opposite direction of how we thought everythinf went for decades.  No, we don't understand even the cellular biology very well.  The idea that neurons are solved is ancient.  And yeah, talking about connections just opens up another can of worms.

But saying we know exactly how neurons and other neural cells work just makes you look ignorant.  We are literally on the cutting edge of cellular and molecular neuroscience right now, full of people who work every day to figure out what are currently mysteries.  Acting like it's solved belittled the amount of work being done on it as we speak.

It’s not belittling the work of neuroscience to say that the very complex details have not yet being worked out. But none of the examples you have given would explain how intelligence or consciousness works. For example, if a new intracellular signalling mechanism is discovered, no-one is going to say: I now see how that allows us to learn to ride bicycles, it seemed impossible if just the cGMP second messenger system were involved.

To view this discussion on the web visit https://groups.google.com/d/msgid/extropolis/CAKrqSyF023_4DGui%2BCLJ%2B3WD2bSJre9vtrZu_5Lb2W%2BXT5DLcA%40mail.gmail.com.

--

Stathis Papaioannou

[Will Steinberg]

Idk, I directly disagree with that statement.  Discovering a new signaling mechanism is akin to discovering a new function in a programming language or general logic system.  It's not much of a leap to believe the discovery of a new logical functionality in the brain could be responsible for some process we don't understand the workings of yet.

And really, it's beyond just discovering a new function in a programming language.  It's like discovering the brain can use an entirely different programming language.  Or even a completely different mode of programming, like a language that was dynamically typed or object oriented when the old ones weren't. 

Idk if we're on the same page here.  We are in the midst of discovering new and vastly different ways that the brain can process information.  

To view this discussion on the web visit https://groups.google.com/d/msgid/extropolis/CAH%3D2ypUiuqLOt%3DucRiYq3y%2B%3D0tC%2BHnEwi2iN7At2DfBvhbL8bw%40mail.gmail.com.

[John Clark]

On Sat, Feb 12, 2022 at 6:56 PM Will Steinberg <steinbe...@gmail.com> wrote:

> The idea of static neurons is some bullshit from the 50s. 

OK, but I never said one word about static neurons.  I don't care if neurons are static or not, either way it doesn't change the fact that the human genome contains at most 750 megabytes, a pitifully small number. 

 

> Neurons are constantly morphing their structure, particularly in what receptors and other membrane bound proteins are on the surface versus in vesicles, as well as the essentially unknown per-cell genome transcription, which is different for every neuron, as well as lots of other cells like glue and astrocytes which we barely understand, and not only that, but we barely even understand gene expression in those cells given recent advances in epigenetics considering histones, methylation, general location of genes within the actual DNA molecule,

 All that complexity and wheels within wheels at the cellular level that you're rhapsodizing about wasn't able to play chess or GO or write video compression algorithms as well as MuZero can. Spaghetti code is not a sure sign of deep profundity in computer programming and I don't think it is necessarily one in biology either.

> "junk" DNA which is often not considered junk anymore (duh), repetitive nucleotide sequences that aren't even genes, etc, etc, etc.

It's true that junk DNA is not composed of genes, bur junk DNA are composed of  nucleotide base pairs and there are only 3 billion of them in the human genome, so all those repeats of stuff like  ACGACGACGACG decreases the number of base pairs that can encode information for stuff that is really important, like genes.

> we don't understand even the cellular biology very well. 

Nor do we need to understand the details of cellular metabolism  that every cell in the body needs just to stay alive to figure out how our mind could work, just as a computer programmer doesn't have to understand why copper can conduct electricity but plastic cannot in order to do his job even though both copper and plastics are in computers.

> But saying we know exactly how neurons and other neural cells work just makes you look ignorant. 

It certainly would if I had said that, so I guess it's fortunate that I DID NOT.

John K Clark

[Stathis Papaioannou]

On Sun, 13 Feb 2022 at 11:52, Will Steinberg <steinbe...@gmail.com> wrote:

Idk, I directly disagree with that statement.  Discovering a new signaling mechanism is akin to discovering a new function in a programming language or general logic system.  It's not much of a leap to believe the discovery of a new logical functionality in the brain could be responsible for some process we don't understand the workings of yet.

And really, it's beyond just discovering a new function in a programming language.  It's like discovering the brain can use an entirely different programming language.  Or even a completely different mode of programming, like a language that was dynamically typed or object oriented when the old ones weren't. 

Idk if we're on the same page here.  We are in the midst of discovering new and vastly different ways that the brain can process information.  

A new programming language would not be fundamentally new, since any general purpose programming language can emulate any other. There may be differences in speed, efficiency and ease of use but not differences in what can theoretically be programmed.

To view this discussion on the web visit https://groups.google.com/d/msgid/extropolis/CAKrqSyH-bmGZtkpsXnGt3WLPWBUk7SA%3D%3Dr_Ob-XHFg%2BCm1UKxQ%40mail.gmail.com.

--

Stathis Papaioannou

[William Flynn Wallace]

Is there something about 750 megabytes that limits the number of connections our brains can make?  At this time the only thing impressive about supercomputers is that they are fast.  We do know that there is no such thing as a gene that does only one thing and is independent of other genes and epigenetic effects (which provide the flexibility as I understand it).  Thus I don't think we can count much on just counting the genes and estimating intelligence.  Will is apparently up on epigenetics (at least he is using terms from the David S Moore book I am reading.)  If you want to be too, get that book!   bill w

--
You received this message because you are subscribed to the Google Groups "extropolis" group.
To unsubscribe from this group and stop receiving emails from it, send an email to extropolis+...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/extropolis/CAJPayv0q%2BK9qt%2BZZWPvxn6o%2BQW2eSZ0_WHL%3D3TZ5QLY-4EOzAg%40mail.gmail.com.

[John Clark]

On Sat, Feb 12, 2022 at 8:42 PM William Flynn Wallace <fooz...@gmail.com> wrote:

> Is there something about 750 megabytes that limits the number of connections our brains can make? 

It places a limit, and a severely restricted limit, on the number of neuron connections that can be hardwired at birth. So all those other connections must come from the environment, and if humans have a algorithm in their head that enables them to extract knowledge from the environment in this way a computer can have it too, and there is no way that algorithm can be very big, it must be way way less than 750 megabytes.

> At this time the only thing impressive about supercomputers is that they are fast. 

With the right program a computer can also be clever,  you need a lot more than just speed to be able to play GO at a super human level or make a better video compression algorithm than the best human programmers.  

  > Thus I don't think we can count much on just counting the genes and estimating intelligence.

I was talking about nucleotide base pairs not genes because that places an upper limit on the complexity of the entire human body, or at least the human body at birth before it has learned anything.

John K Clark  

[Will Steinberg]

On Sat, Feb 12, 2022 at 8:33 PM John Clark <johnk...@gmail.com> wrote:

I don't care if neurons are static or not, either way it doesn't change the fact that the human genome contains at most 750 megabytes, a pitifully small number. 

 I honestly don't know if I have have a conversation with you if this is the level you're thinking on.

Yeah, 750 mb if you just imagine it is some overly simplified base-4 system of nucleotides.  But you ignore so many things.  For one, that's the kernel, try raising a human without any inputs and you won't get something intelligent.  You will get a drooling feral child that can't do anything for itself.  And again, this ignores epigenetics completely, and ignores the fact that the data is compressed.  You act as if your 750 mb quip is some kind of 'gotcha!' while ignoring the fact that human intelligence is predicated on the entirety of human history and the history of life in general.  Not to mention that we also have symbioses with other organisms, that all have their own genomes, and each of our genes could potentially interact with each of their genes, which adds billions more degrees of freedom to the system.  And ditto for the fact that methylation/histones/etc are entirely separate layers on top of the basic genome.  

In general, the idea of using the bit count of the basic quaternary-digital representation of the human genomes is just...so, so ignorant, and such an arbitrary measurement as to be essentially meaningless.

[John Clark]

On Sat, Feb 12, 2022 at 9:03 PM Will Steinberg <steinbe...@gmail.com> wrote:

>  I honestly don't know if I have have a conversation with you if this is the level you're thinking on. [...] the idea of using the bit count of the basic quaternary-digital representation of the human genomes is just...so, so ignorant,

I'll say one thing about biological meat computers, you have definitively demonstrated that they can be turned into rapid fire insult generators.   

 

> Yeah, 750 mb if you just imagine it is some overly simplified base-4 system of nucleotides.  But you ignore so many things.  For one, that's the kernel, 

 

Exactly, and that kernel is what I'm interested in because that gives me an idea of the minimum size a seed AI can be, something that can start off, like a baby, being pretty stupid but can learn to become smarter by extracting knowledge from the environment. 

> try raising a human without any inputs

Try running a computer without any hardwired programming or inputs.  

>  you won't get something intelligent.  You will get a drooling feral child that can't do anything for itself. 

Yeah, so it's a good thing a child has inputs and lives in an interesting environment.  

> And again, this ignores epigenetics

Epigenetics works because highly specialized proteins can cover up parts of a chromosome and prevent certain genes from becoming expressed, but those very proteins wouldn't even exist if they weren't encoded somewhere in the 3 billion nucleotide base pairs sequence that composes the human genome. So epigenetics does not increase that 750 megabyte figure by one single bit.

 

> You act as if your 750 mb quip is some kind of 'gotcha!'

Yep, I sure do!  

>  methylation/histones/etc are entirely separate layers on top of the basic genome.  

Bullshit  

> Not to mention that we also have symbioses with other organisms,

And humans also have parasitic relationships with other organisms, but I'm sure computers will have interactions with other life forms, and with everything else in the environment that humans do too.

 > which adds billions more degrees of freedom to the system.  

More degrees of freedom is not necessarily a good thing, not if some of those involve breaking down or doing stupid things.

John K Clark

[Will Steinberg]

>  methylation/histones/etc are entirely separate layers on top of the basic genome.  

Bullshit  

It's true though, they are independently heritable.  You don't only inherit your parents' germlines.  

[John Clark]

It's true that a recent study showed that American prisoners of war had an 11% higher mortality rate than sons of veterans who were never prisoners of war, however from animal studies it's pretty clear that the epigenetic changes revert back to the way they were after two or three generations. An epigenetic system works by coating parts of the chromosome with a specific protein that regulates, according to environmental conditions, transfer RNA from coming into contact with the DNA in the chromosome, and thus regulating the production of messenger RNA, and thus regulating the production of proteins. And proteins are the things that actually do the grunt work and get things done. The point I'm trying to make is that the very proteins that do this epigenetic regulation are themselves encoded by the DNA in the genome. In other words the genome contains the recipe for building the cell's entire epigenetic system, and that's why I said epigenetics does not add anything to that 750 megabyte figure that I've been mentioning.

John K Clark

[Will Steinberg]

Well yeah, but at that point what does the 750mb even signify?  Clearly it's not a great measure of potential complexity just given the richness of human experience.  And even if everything in my mind is just fleeting, low quality images in the end, there's something to be said for the extreme parsimony of those images.  Idk if that would be considered a type of data compression or whether it's something else entirely.  

I only mean to say that I think it's really tough to apply a data measurement to humanity and have it make any real sense, to a point where I think it's kind of useless to try--at least at the current juncture. 

--
You received this message because you are subscribed to the Google Groups "extropolis" group.
To unsubscribe from this group and stop receiving emails from it, send an email to extropolis+...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/extropolis/CAJPayv0G14g4mTmBCCP6df%3Dhr750nn_a1j5WugWqKfJ%3DcZqQbw%40mail.gmail.com.

[William Flynn Wallace]

the recipe for building the cell's entire epigenetic system, and that's why I said epigenetics does not add anything to that 750 megabyte figure that I've been mentioning.

John K Clark

Perhaps so, but the changes in the epigenetic system are controlled by the environment, not the genes, and could change several times in one's lifetime.  bill w

--
You received this message because you are subscribed to the Google Groups "extropolis" group.
To unsubscribe from this group and stop receiving emails from it, send an email to extropolis+...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/extropolis/CAJPayv0G14g4mTmBCCP6df%3Dhr750nn_a1j5WugWqKfJ%3DcZqQbw%40mail.gmail.com.

[John Clark]

On Sun, Feb 13, 2022 at 7:51 PM Will Steinberg <steinbe...@gmail.com> wrote:

> Well yeah, but at that point what does the 750mb even signify? 

It signifies that the maximum amount of information a program would need to become a seed AI is 750 megabytes, and it's probably closer to 7 megabytes.  

> Clearly it's not a great measure of potential complexity 

That depends on how you measure complexity. By one measure the Mandelbrot Set is not astronomically complex but infinitely complex, the closer you look at it the more new details you see, but if complexity is measured in another way it is very simple, it can be generated by just one line of code. Back in the day the very first program I ever wrote was a Mandelbrot generator, and the longest time that I as a novice programmer spent with it was writing the inputs and outputs not with the Mandelbrot generation itself, that part was easy.

I don't think all those ridiculously convoluted wheels within wheels seen everywhere in biology are a sign of great sophistication but rather the reverse, a sign of incompetent engineering. But that is exactly what you would expect given that biology was made, after 4 billion years of stumbling around, by random mutation and natural selection and not by intelligent design. I'm sure intelligence can engineer something better, and it won't take 4 billion years either.

> given the richness of human experience. 

I'm not saying a 750 MB program would be as smart as Einstein, something that small would be pretty dumb, I'm saying a program as small or smaller than that, a seed AI, would have the potential to evolve, with the help of the environment, into something much much smarter than Einstein.  

> I only mean to say that I think it's really tough to apply a data measurement to humanity and have it make any real sense, to a point where I think it's kind of useless to try

I know it's very trendy to say that human life is so important that you can't attach a number to it, but I think that's sorta nonsensical. I think the more important  something is, the more important it is to attach a number to it because numbers are the best tool we have for predicting what a thing will do. 

John K Clark

[John Clark]

On Sun, Feb 13, 2022 at 8:19 PM William Flynn Wallace <fooz...@gmail.com> wrote:

>> the recipe for building the cell's entire epigenetic system, and that's why I said epigenetics does not add anything to that 750 megabyte figure that I've been mentioning.

> Perhaps so, but the changes in the epigenetic system are controlled by the environment, not the genes, and could change several times in one's lifetime. 

I agree.  

John K Clark

 

[William Flynn Wallace]

I was talking about nucleotide base pairs not genes because that places an upper limit on the complexity of the entire human body, or at least the human body at birth before it has learned anything.

John K Clark  

I would like to know more about that.  Can you send me a link showing how those pairs limit connections?  No one has ever run out of memory space, so learning after birthing seems unlimited.

As for complexity, we have genes that can operate by themselves, as two, three, thousands.  The number of combinations is essentially infinite, not even counting epigenetic effects.   bill w

One error:  the baby comes with learning already in place as a result of experiences in the womb.  The baby learns his mother's voice, for one thing, and who knows what else as a result of what it can hear.  A fetus has been successfully classically conditioned (I do not know the details but it's an old study).  Paul Bloom is the man who has done the best studies of what a six month old baby can understand and learn.   

John, I could not finish The Extended Phenotype.  But I did learn a lot from The Selfish Gene.   bill w

--
You received this message because you are subscribed to the Google Groups "extropolis" group.
To unsubscribe from this group and stop receiving emails from it, send an email to extropolis+...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/extropolis/CAJPayv0j5vW1benYFz8yBb0sM%3DKJDT9xpMsH3cV3K4yMMLuceQ%40mail.gmail.com.

[John Clark]

On Tue, Feb 15, 2022 at 12:37 PM William Flynn Wallace <fooz...@gmail.com> wrote

>> I was talking about nucleotide base pairs not genes because that places an upper limit on the complexity of the entire human body, or at least the human body at birth before it has learned anything.

> I would like to know more about that.  Can you send me a link showing how those pairs limit connections?  No one has ever run out of memory space, so learning after birthing seems unlimited.

I wasn't talking about the maximum amount of memory that a brain can hold either, I was talking about the minimum amount of information needed to make a brain, and that is absolutely positively got to be less than 750 megabytes, way less because that's enough to build an entire human body not just a brain and because the human genome only has 3 billion base pairs and that works out too 750 MB of information . This has nothing to do with how much stuff you can remember. 

> As for complexity, we have genes that can operate by themselves, as two, three, thousands. 

Sure, but that doesn't change the fact that the human genome only has 750 MB of information by one single bit.  

> The number of combinations is essentially infinite,

No, very very big but nowhere even close to being infinite, in fact it's no closer to infinity than the number one is.  And I was talking about how much information it takes to make a brain, not how much information that brain can extract from the environment and remember.  

 

> not even counting epigenetic effects.   bill w

That doesn't change the fact by one iota that the entire human body can be built using only 750 MB of information

> One error:  the baby comes with learning already in place as a result of experiences in the womb.

Could be, that 750 MB could start being supplemented by the environment even before birth, but again that doesn't change the fact that the recipe for making an entire human body, not just the brain, is less than 750 MB.

> John, I could not finish The Extended Phenotype.  But I did learn a lot from The Selfish Gene.   bill w

I'm sorry you didn't finish it, I thought  "The Extended Phenotype" was one of the best books I've ever read.

John K Clark

[William Flynn Wallace]

Finally I understand what you mean by the 750 megabytes.  I thought we were talking about the intelligence capacity of computers and people.  If Nature can put together such a wonderful thing as a human brain from a small starting  assembly, then great.  I guess I was just reacting to the idea of limits on thinking and you weren't at all.

I can see where computers will keep getting faster and faster, but I doubt that programmers can put together a program making a computer much smarter.  If it happened would we realize it?  We really don't know what 'smarter' means in this context.  Equal to us, maybe.  Much smarter?  We don't understand that.  Creativity is another issue altogether.  bill w

I'll give the Dawkins book another try after I finish the epigenetics book.

--
You received this message because you are subscribed to the Google Groups "extropolis" group.
To unsubscribe from this group and stop receiving emails from it, send an email to extropolis+...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/extropolis/CAJPayv0NWujSz50%2Bg%3DoJS2qxzxZ1PobTMiOXQj0Tvc4DFBTLRA%40mail.gmail.com.

[John Clark]

On Tue, Feb 15, 2022 at 6:31 PM William Flynn Wallace <fooz...@gmail.com> wrote:

> Finally I understand what you mean by the 750 megabytes.  I thought we were talking about the intelligence capacity of computers and people. 

I was talking about the minimum amount of information needed to make something that can become as intelligent as a human, and we have absolute proof that that minimum number is less than 750 megabytes; as I said before a 750 MB program cannot be very intelligent, however it can learn to be intelligent from the environment  just as a human baby does. Even Einstein was pretty dumb on the day he was born. 

 

> I guess I was just reacting to the idea of limits on thinking and you weren't at all.

I see no reason why computers can get as smart as humans but no smarter, and I can see plenty of reasons why computers can get to be vastly smarter than humans.

> I can see where computers will keep getting faster and faster, but I doubt that programmers can put together a program making a computer much smarter.

I doubt so too, but humans don't need to make such a thing for such a thing to exist, that's the entire point of what I'm trying to say.  

> We really don't know what 'smarter' means in this context.

Someday, quite suddenly, the human race is gonna find out exactly what it means. 

> I'll give the Dawkins book another try after I finish the epigenetics book.

It would be well worth your time.

John K Clark