Its a straw man if the ID argument you are challenging is about functional information and you use Shannon information the try and refute it.
That would depend on the claim, and here we have no claim other than Information theory is a straw man.
Information Measures have a lot in common (I’ll skip the mathematical definition, but it exists). Statistical theory and Shannon theory have considerable overlap, so I understand Shannon very well. Algorithmic Information is a different Measure, but the theory behind it parallels that of statistical/Shannon theory. It’s not identical, but there are analogs to concepts and theorems. I spent a year reading up on Algorithmic theory a while back to make sure I was understanding it properly.
Hazen’s Functional Information is also an Information Measure. I’m not up to speed on applications, but the theory behind it should also parallel that of SI and AI.
Of note: CSI as defined in Dembski (2005) is technically not an Information Measure (because it is not defined on a probability distribution), though it could be corrected to be one. The quantity it measure is not useful in Biology (See Joe Felsenstein’s comments above).
Defined how, and by whom–exactly?
Are we now agreed that @gpuccio’s definition should be discarded?
Hello DaveB!
May I ask if you are the same DaveB that posted briefly in TheologyWeb science forum back in 2019?
Never mind - your posts confirm that you are - and that you didn’t learn anything from your time there.
You’re right here, I am not agreeing with the idea that “Directional processes such as natural selection show CONSIDERABLE potential for supplying the information needed for substantial connection between structure and function.” However, I am happy to agree with the following statement : “Directional processes such as natural selection show SOME potential for supplying the information needed for optimizing, within a given island of function, the connection between structure and function ».
OK, so in a case where the entire space of sequences was one island, natural selection would be able to put functional information into the genome? Somehow I doubt that Giltil will say even that …
One assumes that starting with a random sequence with an identified function and subjecting it to selection for that function, would result in a sequence improved for that function. Which, unless I am mistaken, would be an increase in the functional information of that sequence according to Hazen regardless of how rough the fitness landscape might be.
@Giltil is still yet to provide a mathematical (or any other sort of) framework by which we could judge if the number of globally optimized traits, relative to the total number of traits and when adjusted for which have multiple islands of functionality in the first place, for known valley crossing methods, and so forth, is problematic.
How do you think islands of function work for multi gene/protein systems like the ubiquitin system or developmental pathways like WNT or Sonic hedgehog?
They would be larger, because they involve so much protein-protein binding. Binding is simpler. Proteins are incredibly sticky (which you’d know if you ever purified any), and binding is more selective than specific.
And generally speaking the larger a protein, the more surface area to bind, meaning there’s more real-estate to exhibit some affinity, meaning you’ll need less specificity. Large patches of weaker affinity can have the same effect as a very narrow spot with stronger affinity. The net binding affinity can be equal.
Don’t be too hard on Gil; SOME potential is a big step from NO potential.
Just like the Hawaiian chain, new islands are arising all the time. This is supported by a mathematical model of the genetics involved:
AFAICT ubiquitin is a single protein, not a multiprotein system, and Sonic Hedgehog is a protein, not a developmental pathway.
So this is akin to asking how skis work for camels like the Eiffel tower.
I think he meant the ubiquitin-proteasome complex.
And that thing about valley crossing methods is pretty important, as numerous mechanisms and causes of valley crossing are known from evolutionary biology. Fitness landscapes are also dynamic not static, almost always have more than one contributing cause, and there are other types of mutation than just point substitutions. Heck, you can even cross valleys just by drift through fixation of suboptimal mutants in small populations.
To just note that because fitness landscapes often times have a rugged feature - therefore evolution is impossible - is a complete non-sequitur.
I think that if a functional island can be found in some ways, then some new abilities can probably in some case be acquired by divergence of gene duplicates through random mutation and natural selection. But one cannot generalize this possibility for it must be experimentally investigated on a case by case basis.
It seems absurd to demand that any and all cases of diverged duplicate genes be experimentally demonstrated to be feasible, particularly when the fact that one can construct consilient phylogenies from totally independent genes and have these nevertheless strongly corroborate each other. There is simply no other good explanation for this reality than that they really did evolve and diverge through sequence, structure, and function, from common ancestral genes.
The fact that biochemists have been able to use said phylogenies to reconstruct functional ancestors and show which historical changes in sequence were causal for such functional transitions should effectively put any doubts about the general reality of such a phenomenon to rest, even if this has only been done for a small fraction of all known protein coding genes. There just is no basis in physics or chemistry for thinking some specific example that has yet to be experimentally characterized in that way is somehow an exception.
And just to be sure, this is not meant to support a claim that any and all genes can be evolved gradually and incrementally into each other while retaining biologically useful functions, just that where we can detect significant phylogenetic signal in sequence and structure, we already there have very strong evidence that the genes in question share an evolutionary history, and hence whatever functional transition is implied by this phylogeny must de facto be possible, otherwise there is (again) no good explanation for why we should have such evidence in the first place.