I think this is an important paper. Endres tries to estimate the
probability of abiogenesis, the chance of a chemical soup forming the first
molecule capable of storing the instructions for copying itself and
evolving. If it is more than 10^-24 then it is likely that Earth is not the
only planet in the universe with life. If it is less, then it implies the
universe is vastly larger than we can see beyond the event horizon 13.8
billion light years away.

Unfortunately he doesn't say. In order for life to exist, the information
gain from evolution (at most 1 bit per population doubling) has to exceed
the loss from copy errors. That's about 10^-9 per DNA base in humans and
10^-2 in bacteria. The simplest life form that exists today independent of
other species is about 10^6 bits. Life might have begun as single strand
self replicating RNA driven by cycling the temperature in a bath of bases
like with PCR in DNA sequencing machines. But we don't know. Nobody has
made this work in a lab.

The paper references Stuart Kauffman, who has done important work in self
organized criticality. All complex systems with feedback evolve to the
boundary between stable (perturbations decay exponentially) and chaotic
(perturbations grow exponentially). This includes sets of random logic
gates, gene regulation networks (genes turning other genes on or off,
equivalent to logic gates), neural networks, and software. Systems become
chaotic when interconnectivity increases beyond a threshold. For logic
gates, the transition to chaotic, crossing the narrow window between
uninteresting and unpredictable, occurs somewhere between 2 and 3 average
inputs per gate.

I don't know if this is the same phase transition that Dorian Aur claims
the human brain crosses between 2 and 3 years old, which is why we have no
episodic memory before that. My explanation would be that the hippocampus
is not fully developed. Episodic memories are associations with places, a
map represented by a 2-D grid of neurons there.

-- Matt Mahoney, [email protected]

On Sun, Sep 7, 2025, 8:42 PM John Rose via AGI <[email protected]> wrote:

>
> Interesting minimal protocell K complexity estimate and self-assembly
> probabilities, references Landauer :)
>
> "Taken together, sufficiently complex chemical reaction networks can mimic
> the dynamics of recurrent
> neural networks, or make the emergence of complex logic functions
> inevitable. This raises the intriguing
> possibility that biological systems—even at the molecular scale—approach
> universal computation in their
> expressive power. Indeed, chemical kinetics, when properly structured, can
> implement universal computation
> [66]. More concretely, chemical reaction networks governed by mass-action
> kinetics can simulate
> arbitrary computations and are thus Turing complete [67]."
>
> https://arxiv.org/abs/2507.18545
> *Artificial General Intelligence List <https://agi.topicbox.com/latest>*
> / AGI / see discussions <https://agi.topicbox.com/groups/agi> +
> participants <https://agi.topicbox.com/groups/agi/members> +
> delivery options <https://agi.topicbox.com/groups/agi/subscription>
> Permalink
> <https://agi.topicbox.com/groups/agi/T1ff34dc69e862fc1-M8e00942795ef24d31d4a582a>
>

------------------------------------------
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T1ff34dc69e862fc1-Md9c32857ae082b28847bd7b4
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Reply via email to