Hacker News Reader

Comment by rhelz | Hacker News Reader

In any case, 6.2 billion bits (interestingly enough, almost exactly as much information which is on an audio CD which you used for your romantic mixtapes) is an upper bound.

This rules out pretty much every nutty theory which evolutionary psychologists propose. Such as we evolved for altruism, we evolved to believe in religion, etc etc. Complete B.S. Exactly how much information would you need to specify a behavior like being predisposed to a belief in religion??? There's less than 80 minutes worth of music's worth of information in our genomes, and most of that is concerned with just keeping us alive.

You are not predisposed to be anything. Go create the kind of person you want to be.

View on HN

> There's less than 80 minutes worth of music's worth of information

Or awful lot of text information (state of art compressors can do up to 1:10 ratio for plain text, decoder itself is rather small, 750MB compressed could potentially contain like 7GB of text data).

Also, look at demoscene. 4k (4 kB is the size of executable) can do crazy things, and 64kB can fit a lot of nice 3D objects, music, text, complex effects etc. weight less than any screenshot of any moment of running demo. In 95kB you can have full game (google kkringer)

P.S. better example: full snake game in 56 BYTES https://github.com/donno2048/snake

For comparation the link above is 34 bytes, whole sentence is 83 bytes. It's possible to do a lot if we're talking about code

There's an interesting implication to this. We assume that evolution happens when random mutations (similar to random bit flips, removal or injection?) occur and when the random result has an advantage, the mutation tends to remain in the gene pool.

Yet at the same time the result of this random code is extremely compressed, to the point we compare it to procedural generative code.

Not sure what we can do with this but it certainly seems like we can once again get inspired by nature on this one.

I'd argue you could even take that one step further. Limiting it to the data encoded by DNA does not take into account what it is interacting with. DNA interacts with an ocean of protein leading to untold numbers of interactions. The DNA could just be the operating system in all this calling upon RNA and other "devices" to execute functions.

To expand upon your compression idea, the index it is using exists outside the DNA encoding itself which means it could be holding an absolute ton of data.

Bonus: https://xkcd.com/3056/

Indeed. Chances are that the DNA itself is but one part of the puzzle. The protein soup the DNA interacts with is partially random and partially a consequence of the DNA itself and that interplay is likely a complexity space several orders of magnitude bigger than the DNA itself

> The DNA could just be the operating system...

I am fond of the analogy of DNA to procedural generation. The "operating system", as I see it, is physics. Everything else is primitives built on top of that.

Our brains can't begin to comprehend the untold multitudes of interactions occurring at a molecular scale over geologic time.

Saying that the information doesn't come from the DNA, but from the environment--well that is just conceding my point. Sure, environment is way, way, more important to the formation of who you are than genetics is.

> Also, look at demoscene. 4k (4 kB is the size of executable) can do crazy things

There are limits to how Kolmogorov complexity scales up. Many of these tricks are exploiting procedural techniques that can be expressed in minimal terms. Once you start feeding in actual information that is not feasible to express procedurally (i.e., is already compressed/high-entropy), you are forced to accumulate bits. An obvious example of this would be incorporating a texture that is multiple megabytes when compressed as a jpeg on disk.

Evolution also uses dirty tricks all the time, for no reason. E.g. the same region gets reused for totally unrelated use-cases.

> An obvious example of this would be incorporating a texture

Some random range of storage data is now the texture. It was used to process formatting logic but now also a texture

> There's less than 80 minutes worth of music's worth of information in our genomes

That’s a very misleading take, this is lossless audio and the majority of the bits are spent encoding noise. You can encode way more audio at perceptually but not technically lossless level in that space.

There is no logic behind your argument

> There's less than 80 minutes worth of music's worth of information in our genomes

What an insanely bad take.

Not only did you not read and/or comprehend the article, the article itself undersells the information content of the genome (I'll post on this at the top level).

> You are not predisposed to be anything.

This does not logically follow your preceding statement, even if we were to accept the foregoing limited information content as factual

no matter how much information you think is stored in DNA, the information stored in your brain is at least 5 orders of magnitude more. In comparison with what you have learned, your DNA is a rounding error.

What you can learn will just swamp any predispositions you have.

This isn't a great argument - simple rules can produce complicated behavior and, at any rate, I don't think any evpsych people believe that evolution inescapably predisposes people to the things you talked about, only that evolution has produced biases in our behavior which manifest (at certain times and in certain circumstances) as those phenomena.

chuckle hello dissenters, downvotes, and doubters :-) Seems like a lot of the objections stem from a few misconceptions:

1. You can dramatically increase the amount of information stored by compression. Uh, no. Information content is measured, as it were, "post-compression". There is a limit to how much information you can store in a gigabyte, and that limit is--a gigabyte.

2. Information content is not just stored in the DNA, but also in all of the ancillary proteins, etc. Well, the the information on how to create the proteins themselves are stored in DNA. Any additional information, then, has to be contributed by the environment. But that is exactly my point--environment matters way more than the information stored in the DNA.

3. "you are wrong", "you are a mathematical ignoramus" etc etc is ad hom. It is not a valid argument, contributes nothing to the conversation, and is not a good look. If you disagree let's see some math.

4. No matter how much information you think is stored in DNA, the amount of information stored in your brain is at least 5 orders of magnitude larger. The information you can learn swamps out any predispositions you might have. Go become the kind of person you want to be.

You are predisposed to acting like your closest social circle.

An audio CD is a very inefficient way of storing information.

I think a more apt comparison would be that of a LLM of that size. qwen:0.5b is about 400MB, its abilities are laughable compared to the likes of ChatGPT, but it can write coherently about general topics. For instance.

  >>> why would people be altruistic
  People are likely to be altruistic because they believe that helping others is better for everyone involved.
  People may also believe in the power of compassion and empathy towards others, which can contribute to greater altruism.
  Overall, people are likely to be altruistic because they believe that helping others is better for everyone involved.

It is not a statement about LLMs, more about what you can achieve with "just" 400MB for storage. The other similarity is that LLMs are also "messy", if you want to see the results of finely crafted work in a really small amount of space, look at what sizecoders can do with a few kB of code or less.

Reply on news.ycombinator.com