Comments by RBH

Comment #71286 by RBH on September 18, 2007 at 9:30 am

I've been using genetic algorithms to evolve artificial agents that do real-world tasks in a commercial context for 15 years. The video is at roughly a 1st grade level, barely out of kindergarten, and badly over-simplifies and caricatures the technology, so much so that a naive viewer would literally have a false idea of how genetic algorithms are actually used.

For example, the talking head who started by describing domestication of corn speaks of genetic algorithms as merely devices for generating variants, with somebody sitting at a display selecting which variants will survive and propagate. That's about as inefficient a way to use genetic algorithms as I can imagine. I use populations of 1,000 variants, and a fitness function that incorporates half a dozen properties of the critters. The selection process (where selection is defined as mating probability as a function of relative fitness) and is wholly automatic -- no human sits there picking out this or that variant to propagate. That would be terribly inefficient and in fact would throw away the real strength of GAs. No human knows which nascent lineage might lead to a successful (in terms of the task set) descendant, and so cannot 'select' variants for propagation in any useful way. We use GAs precisely because humans can't do it! The GA doesn't 'know' which nascent lineage might be fruitful, either, but with a decent-sized population it will run multiple lineages in parallel, 'testing' multiple approaches simultaneously and letting them fight it out over generations. Humans aren't very good at that kind of parallel search, either.

So the video basically misrepresents the real strength of evolutionary algorithms in applied contexts.

2. The God of the Bible is No Delusion!

Comment #11700 by RBH on December 6, 2006 at 3:31 pm

Tom B: "What genetic mutation is beneficial?"
The Milano mutation (see http://tinyurl.com/ylsnd2) is but one of millions of examples.

Tom B: "Why havent all monkeys evolved into human like crreatures?"
Because they're well adapted to their current selective environment as they are (or at least as their selective environment was before we humans have been screwing up their habitat). And since we humans already occupy the selective niche for allegedly intelligent primates, we'd slaughter them in a heartbeat as we have slaughtered most of the top-of-the-line predators.

Tom B: "How can an a living thing evolve from inorganic chemicals?"
That's not a question in evolutionary biology, but in geochemistry and biochemistry. Go hassle the chemists.

Tom B: Does organic substances such as alcohols have the potential to become alive?
No. Not all carbon-centric ("organic") chemicals are capable of self-replication with heritable variation.

Tom B: "Who set off the big bang?"
No one.

Tom B: "If we have all evolved from amoeba at as slow a rate as evolution appears to take place in creatures today how long would it have to take for the billions and billions of small steps that would have to have to be taken to create a human?"
About 3.5 billion years.