Wesley R. Elsberry
Posts: 4991
Joined: May 2002
|
| Quote (cryptoguru @ Feb. 17 2015,11:15) | Wesley R. Elsberry: sorry, you are correct ... it was 17 .. which is still trivial from a cumulative search point of view.
In fact it's always a trivial problem if you set targets at that level.
|
26^17 ~= 10^24, which looks like a pretty non-trivial problem space. Even Cryptoguru's fake number of 9 instructions for EQU leads to 26^9 ~= 10^12, which would be an excessively large for a lottery. Powerball uses a base space of a bit more than 10^8.
Nor is "cumulative search" appropriate as a dismissal for what is happening in Avida. I've already noted multiple times that Avida is awarding merit on behavior, not on any examination of Avidian genome content.
| Quote (cryptoguru @ Feb. 17 2015,11:15) |
I do understand how it works
|
Then why is it that pretty much everything Cryptoguru has said about Avida other than that it is a computer program is wrong?
| Quote (cryptoguru @ Feb. 17 2015,11:15) |
Avida selects by using complete logic gates plus IO as symbols,
|
Wrong. Avidian instructions in the 2003 paper do not correspond to digital circuit logic gates. The functions being awarded merit in the 2003 paper correspond to typical logic functions (AND, OR, NOT, etc.). Cryptoguru is making this into a persistently repeated error.
| Quote (cryptoguru @ Feb. 17 2015,11:15) |
this could be compared to bacterial "evolution", which is accomplished by swapping entire operons (multiple genes) on one plasmid symbol.
|
What a confused statement. Between the utter ignorance of anything to do with Avida and the confusions about the biology, saying the above declaration is worthless insults rubbish. "Plasmid symbol" is apparently jargon from the "Bioshock" game. Nor is Avida limited to broad-brush analysis as Cryptoguru claims, which he should now by now because I told him before:
| Quote |
This is not in the original challenge, and is thus irrelevant to answering the original challenge. The Avida instruction set includes "mov-head", "jump-head", and "set-flow", which can and do change expression dramatically. Avida itself has been used to perform in silico experimentation on overlapping genes:
| Quote |
One consequence of overlapping genes is to reduce the tolerance for mutation. Virtual experiments conducted within the past several years using a software system called Avida have indicated that overlapping reduces the probability of accumulating so-called neutral mutations in a gene (mutations that have no effect). Neutral mutations are unlikely with overlapping genes, because the mutation must have no effect on two genes with different reading frames.
|
|
You can't do experimentation about overlapping genes if your model doesn't permit changes within its gene model. QED.
But wait, there's more... some time back, Cryptoguru made a different claim about level of Avidian genome:
| Quote |
Section D:
comparison of AVIDA to biological evolution
1) an analogy must be drawn between AVIDA commands and the genome
2) it possibly makes most sense (I concede) to assume that AVIDA commands are analogous to codons (and not proteins), so that any mutation will always create a set of valid codons.
3) the level AVIDA is selecting at is therefore analogous to a folded functional protein.
|
That this is entirely incompatible with the new claim that Avida only acts on multi-genic blocks is evident.
| Quote (cryptoguru @ Feb. 17 2015,11:15) |
I agree that swapping out an entire chromosome is large enough to select on. But, a nucleotide is waaay too small a symbol to select for universally in multi-cellular organisms, except of course for a few edge cases where a point mutation happens to have the effect of switching on/off a whole gene ... but this cannot explain the arrival of new information.
|
The premise was false, thus whatever conclusion was drawn has no validity.
| Quote (cryptoguru @ Feb. 17 2015,11:15) |
A single improvement needs to have an increase in fitness of greater than 10% (probably more like 50%) to be selected over environmental noise.
|
Um, BS. Effectiveness of selection over drift is a function of a variety of factors conveniently omitted from the above. Nor do I see any authority stating any such thing, nor do I recall any such large change in fitness being mentioned as necessary in my classes or further reading.
| Quote (cryptoguru @ Feb. 17 2015,11:15) |
Also, you would need less than 1 average symbol mutation per organism per generation to keep any advantages you have accrued. There are thousands more "neutral" or bad mutations compared to advantages, so there's no way to hold any good ground you may have covered in your previous mutations.
|
Cryptoguru also is confused about the potential of where mutations happen and the actual extent to which they happen. Yes, there are lots more ways things can go wrong than ways they can go right. But selection can both eliminate the bad and preserve the good (assuming those turn up in separate organisms). So, what about Cryptoguru's claim of 1 average base mutation per organism?
From here:
| Quote |
In particular, examination of sequence conservation between humans and primates implies that ~ 38% of coding sites are maintained by selection, and that the net mutation rate is high: U ~ 4.2 per generation. (46)
|
That's an average of 4.2 SNP mutations per human individual. The notion that 1 is a limit is false. If it were true, the human race would be in a world of hurt from mutational load already. One has to look at the likelihood that a mutation will effect the gene that has accrued an advantage, not just the general mutational load. I discuss the general form of that calculation with respect to the "weasel" program here.
The quoted article covers a variety of limiting factors for selection. Unlike Cryptoguru, it sets things out in a reasonable way and takes notice of actual evidence.
| Quote (cryptoguru @ Feb. 17 2015,11:15) |
Also, only bacteria have hundreds or thousands of offspring. These evolutionary algorithms only work in large populations with small mutation rates. Mammals particularly don't have hundreds of children.
|
Bacteria don't have hundreds or thousands of offspring, at least not at one time.
Since we were talking about Avida in particular, we once again come up with the simple observation that Cryptoguru is completely clueless on the topic. The Avida environment of the 2003 paper had a limit of 3,600 Avidians. While that is a sizable number, it isn't a large effective population relative to the sorts of biological populations whose members do have hundreds or more offspring at a whack. Nor is the Avida mutation rate "small" by comparison to biological exemplars.
Finally, Cryptoguru knows bugger-all about mammalian biology. It took seconds to find this discussing naked mole rats, a mammal:
| Quote |
Many breeding females in captivity have reared offspring for more than 15 years, and our most fecund female reared >900 pups over her 11-year reign (10).
|
| Quote (cryptoguru @ Feb. 17 2015,11:15) |
So as I was saying, you may be able to demonstrate a simple shuffling on the order of bacterial operons, with large populations and very short generation lengths. But anything more complex cannot be modelled by Avida or anything else, and to a mathematician like me ... seems utterly dumb!
|
What seems to me to be utterly dumb is to continue to spout fabricated nonsense about topics one doesn't have a clue about, but Cryptoguru seems eminently comfortable doing just that. Avida is not doing shuffling on the order of bacterial operons. The notion that Avida cannot model anything more complex than what we have been discussing is wrong, too. I've mentioned that before, so one wonders exactly how slow Cryptoguru is on the uptake:
| Quote |
Avida has been used to generate and evaluate UML models and for the generation of firmware for wireless sensors to be deployed in wireless sensor networks, and those are a couple of applications that I knew about before I left MSU back in 2009. MSU got the BEACON grant shortly thereafter, and things have been hot in the lab since then.
|
| Quote (cryptoguru @ Feb. 17 2015,11:15) |
Maybe others are fooled when you use really long time scales ... but the complexity ramps up exponentially from these simple toy problems you're solving with programs like Avida. You don't have anywhere near enough time .. and even if you did, you wouldn't be able to keep hold of any benefits anyway. This is what we observe. |
Wrong again. Cryptoguru hasn't been right often enough to be able to substantiate a claim to have observed anything. I've noted Cryptoguru's complete lack of addressing the math of retaining genetic benefits. Nor would I characterize what Avida has been applied to as being all toy problems. (And that's a list from just one of my former colleagues.)
--------------
"You can't teach an old dogma new tricks." - Dorothy Parker
|
