site banner

To What Extent are Humans Optimized?

Jump in the discussion.

No email address required.

You're looking for specialisation versus general, though. Is there some animal better than humans at locomotion? Probably, I can't be bothered looking this one up. Same for all the rest of it as a single point of examination. But put the entire package together, and we're probably very much optimised (there can always be improvements, some genetic engineer will figure out how to give humans three sets of legs or something).

Personally, I'd advise to forget all the stuff about "the brain is like a computer". No, it's not, but humans have been making these analogies with machinery since gears and cogs. We try to understand how the lump of matter in our skull works to do what it does, and we try to make deductions from the way the machines we construct work - oh, a clock works with these gears and springs and drives, a brain must work like that too.

I'd advise to forget all the stuff about "the brain is like a computer".

The brain is a computer and I'm not sure how you can meaningfully argue that it isn't

If the question is "how good are human brains compared to alternatives" ... given we know basically nothing about how either of these actually work, in the sense of 'creating large-scale knowledge or behavior', there's not really anything we can say?

"Substrate"-wise, presumably if neurons are "doing something" superconducting wires, or just silicon, could "do" that much faster, action potentials are quite a few orders of magnitude away from optimal for transmitting a few bits, either in energy or time. But I'm not sure how that relates to the overall question.

There is a lesswrong post with ~200 upvotes that goes into detail on this. Unfortunately, it's, as far as I can tell, mostly bullshit, and is mostly applying mathematical ideas outside contexts they make sense / (unintentional) Eulering.

Appreciate the lesswrong link. I'm interested in conjecture and wild predictions and I don't particularly care about accuracy so its exactly what I'm looking for.

The human body below the neck has remained largely unchanged since the evolution of Australopithecus four million years ago. One may conclude from this that the body is well optimized. The human body above the neck has changed continuously since that time and only converged on its present form 250,000 years ago. One may surmise from this that the brain is in some ways less optimized than the body.

Everything in evolution is well optimized, though, how do we know it's not either a 'local minima' or just 'less important than other things evolution acts on'? Bellybuttons haven't changed for a while, and even though they are suboptimal (see skin, generally, not having random holes or bumps), they're not important enough to change.

The left recurrent laryngeal nerve and the vas deferens are classic examples of suboptimal anatomical features. One might imagine there are structures in the brain that are suboptimal in a way that is analogous to the left recurrent laryngeal nerve and the vas deferens. Perhaps there are some we have yet to recognize as such because we don't understand how the brain works.

minimizing how many instructions you need to add to the genome in order to make something happen

This is the mechanism for why the body isn't fully optimized: it has to be backwards compatible and build iteratively off of previous DNA. That tech debt is exactly what an artificial engineered brain/body could re-design.

That makes sense, direct editing is going to cause a lot of indirect problems. Similar to immune system or dietary supplements.

TierZoo is an entertaining youtube channel that pretends animals are like classes in a video game and rates their stats. He thinks pretty highly of humans. I recommend his videos in general.

I think humans, outside of the more recent brain areas that have had relatively little time to evolve, are generally pretty optimized. Not perfectly, but our(and every animal's) biology is really quite impressive. Just look at how few improvements you can make to humans by adding prosthetics. Glasses can improve our vision, but other than that, even if you're a billionaire, you aren't really able to improve upon the human body beyond what biology already gives you.

The knees and lower spine of humans are famously frail for the load they have to support -- a consequence of remolding our body for erect bipedalism in a relatively short time.

Humans aren’t the best of anything, compared to other animals, except brainpower.

Our ability to throw objects is unmatched in the animal kingdom, and it's not even close.

Major League Baseball is insanely human. The pitcher has to hurl a fist-sized ball into a static cuboid space from far away at a blazing speed, batter has to start moving the bat to intersect that cuboid and hit the ball before he can even see its path, basing his movement on his 20/20 vision’s and brain’s optimized estimation of the pitcher’s movements. Runners and fielders have to run upwards of 15 MPH at a moment’s notice and stop on a dime. And the umpires have to observe all of this and catch anything which breaks a single rule out of an entire book’s worth of rules.

I think humans excel at heat dissipation and efficiency of locomotion

What do you mean, human brains "might" use analog computation? I wasn't aware human brains did any digital computation.

? The conventional view is that human brains are digital computers. Information is represented symbolically in the form of nerve impulses, which is what enables the otherwise uniform computational substrate of the cerebrum to otherwise be able to process such a variety of different things from simple sensation to the considerations involved in complex planning etc.

Not of anyone aware of the fact that a single neuron can have thousands of dendritic spines that undergo constant remodeling.

Neurons are capable of responding to both frequency and amplitude of nerve impulses they receive

This just means that neurons don't work in binary like most artificial electrical computers, it doesn't mean they aren't components of a digital computer. Digital doesn't mean binary.

underlying biology is 100% analog.

From your comment I'm afraid you might just have no idea what a digital computer actually is. Whether information is transferred electrically or not isn't what separates digital computers from analog ones, but rather the fact that information is represented symbolically in digital computers and not in analog ones. Information is represented symbolically in the form of action potentials in neurons in the brain meaning it is a digital computer.

"otherwise uniform computational substrate of the cerebrum to otherwise be able to process such a variety of different things from simple sensation to the considerations involved in complex planning etc." is gobbledygook that explains nothing as far as neuroscience is concerned.

When you want to be a dick, you should at least try and know something of what you're talking about so you don't come across as both needlessly overconfident and a dick. The fact that neurons can process both sense information as well as consider things like complex planning means that by definition they must be the components of a digital computer. What is uniform about the substrate is that it is all composed of neurons, not whether the architecture happens to be similar or not across various cortical layers (just as the architecture of different computer processors is not uniform despite all being made of transistors.)