It is all simulated
Not necessarily. I see no indication in the paper that they didn't breadboard their best result in order to verify it in the "real world". You're jumping to conclusions.
But even if they didn't and it was "all simulated", so what? Are you not aware that using circuit simulators like SPICE is how *human* designers formulate their designs as well? Are you under the impression that electronic devices like transistors and resistors somehow aren't so well understood and predictable that circuit simulators *aren't* highly accurate? If transistors et all were as unpredictable as you imply, they couldn't even be used to make reliable circuits at all.
Finally, even if evolved circuits allegedly outperform human-designed circuits "on the simulator" (a trivial point, since the behavior of an electronic circuit is just as complex on the simulator as it is in "real life", even if it's allegedly "off" by a few percent), how does that somehow invalidate the point being made, that evolution can outdo human-designed efforts, in whatever realm? The arena is irrelevant when the rules are applied equally to both competitors.
In short, you're really splitting hairs again.
The evolved circuit in the paper has 18 transistors not 17 as in the SciAm article(it is not the same circuit)
True, it's not the same circuit. The researchers appear to have topped their prior "personal best" and produced an even better circuit, which they presented in their later paper. Your point?
The patented circuit has 5 transistors and 4 diodes not 9 transistors(pointed out previously)
Oh, puh-leaze... If you know this little about circuit design, maybe you shouldn't be trying to critique it.
Hint for the newbies: If you wire the base and collector of a transistor together, it acts as a diode.
Now, how many transistors in the Scientific American picture are wired this way. Four, right? How many diodes are in the Patent application Figure 6. Four you say? This leaves 9-4=5 transistors operating as 3-lead transistors, right? How many transistors are there in Figure 6 of the patent application? Can you say 5? I knew you could.
Bigger hint: Replace the input-looped transistors in the Sci.Am. article with diodes, and *ta daa*, you get exactly the circuit in Figure 6 of the patent application.
Thus endeth Circuit Design For Dummies.
The evolved circuit has a larger maximum error
Whoop de do. It's only on the very top end, for a tiny fraction of the domain, and it's 1.3% deviation versus 0.87% deviation for the patented circuit. At almost all other portions of the domain it has significantly less error than the patented circuit, including during the critical first 90%. Furthermore, the patented circuit shows some real ugly deviation from 75% upwards, while the evolved circuit only starts going into significant deviation upwards of 95+%.
You're grasping at straws again. For any real application I'd choose the evolved circuit over the patented one any day.
There is no evidence that the evolved circuit will operate suitably even in simulation at 1Ghz. This is especially suspected since the paper states -- We used the commercially common 2N3904 (npn) and 2N3906 (pnp) transistor modlels unless the patent document called for a different model. This is from the specs for 2N3904 The useful dynamic range extends to 100 mA as a switch and to 100 MHz as an amplifier.. The 2N3906 has similar characteristics.
And since the patent *did* "call for a different model" by specifying a frequency domain up into the gigahertz range, your point is moot, isn't it?
Also previously noted --> I suspect that the performance edge is a paper product. Something that the emulating program has produced. Why do I surmise that? Because in evolving the circuit I doubt that each individual circuit was constructed in order to measure its performance of the cubic function. That would be impractical.
And your belief that circuit simulation (a process so well refined due to the inherent physical predictiability of electronic components that human designers consistently rely upon it when doing *their* designs) is somehow a huge departure from actual performance is based on... what?
Again, you're grasping at any straw you can find to try to avoid having to deal with the undeniable fact that despite creationist "proofs" that evolution can't produce any solution of complexity or "increasing information content" or at a level to compete with "intelligent design", when evolution is actually put to work, time and time again it *does*.
Deal with it squarely, and stop squirming.
I see you have a low self-esteem. Don't be so hard on yourself.
First, I deal with what you write, not with what you wish you had written. Which was --->
In this case, it made a cubic function generator circuit which outperforms the best that all electronic engineers were capable of producing in all the history of electronics.
The circuit at the top was patented in 2000, and is the current state of the art. The circuit at the bottom was produced by pure unaided evolution, and outperforms the human version.
Clearly, you now admit there is still no evidence for your claim.
True, it's not the same circuit.
Second, you have no problems jumping to conclusions
The researchers appear to have topped their prior "personal best" and produced an even better circuit, which they presented in their later paper.
But fail to see the facts.-->Circuits are simulated using SPICE (Quarles, Pederson, Newton, Sangiovanni-Vincentelli 1994).
Third, simulations are not the real thing. We lost several Mars probes, all of which were run through simulations.
Fourth, you use hyperbole erroneously again.
If transistors et all were as unpredictable as you imply, they couldn't even be used to make reliable circuits at all.
I did not imply anything about transistors. I did imply that simulations are not real.
Fifth, the circuit designers and I know that a junction transistor can be used as a diode. They did not choose to do so. The circuit was changed.
Sixth, unterminated runs act as antennae and at high frequencies provide signal loss. I have no doubt that you would select a just-so story over the real thing, but that does not qualify as evidence for the embodiment of the just-so story to perform the required tasks.
For any real application I'd choose the evolved circuit over the patented one any day.
At this point, your assertions are still hanging.