Computational Limit of Life May Be Billion Times Higher Than Assumed | Hacker News Reader

Computational Limit of Life May Be Billion Times Higher Than Assumed

popularmechanics.com

22 pointsvirtualritz17 comments

I think we are vastly underestimating how optimised life on Earth is on the axes of energy efficiency vs computational capacity. Current state of robotics and AI together is showing us that life and biological neural networks are capable of doing things we still cannot fully replicate, all the while using orders of magnitude less energy. A honey bee or an ant still remains incomprehensible, and the most basic worm insimulable.

I posit an extrapolation that once we figure out how life has managed to come up with its incredible computational capabilities, our silicon AI is going to become not only a lot smarter, but also a lot cheaper too.

> our silicon AI is going to become not only a lot smarter, but also a lot cheaper too.

It is entirely possible that silicon is just strictly not capable of reaching the computational capacity of meat

It is almost certain that silicon cannot reach the complexity of a human brain in the same "fits inside a human skull" footprint that brains have to adhere to

On the other hand, mechanical devices are massively stronger than the strongest living animal. Metal and composite materials are wildly stronger than bone. Wheeled vehicles and planes move faster than the fastest cheetah or falcon. There are a small number of exceptions where biology-derived materials strictly outperform our best synthetic alternatives, but generally speaking we’ve outperformed in every other area. It wouldn’t surprise me if non-biological materials ultimately beat out biological neurons.

This seems trivially obvious and also "so what?". Of course non-biological materials beat out biological ones in a lot of ways. They don't get tired, they are impossibly strong, whatever else.

We already have machines that do the jobs of many people, with only a single human operator. Think of a simple excavator. Using it, a single human can move more dirt per hour than dozens of people with shovels possibly could. Arguably the human pilot is operating as a stand-in for the machines "brain"

The "so what" is that we already have hundreds of examples of places where biological materials find an amazing balance within whatever energy/cost/resource-availability envelope they were evolved in -- but where artificial materials (perhaps outside that energy/resource envelope) have proven to be hugely superior. I'm just hypothesizing that there's no reason to believe that biological computation will prove to be the exception.

In fact, only two elements - carbon and silicon - can form stable, tetrahedrally bonded 3D networks of sufficient complexity and flexibility to support recursive molecular systems at planetary temperatures

So if there’s going to be another system with equivalent complexity it’s going to either be another carbon based thing or a silicon based thing

THIS is it. People are positing silly extrapolations ("We can make steel stronger than bones, so we can make silicon computers smarter/volume than organic ones").

It eventually comes down to organic chemistry versus non-organic chemistry. Organic chemistry has far greater dimensions of freedom (ways to combine/exchange energy) to exploit unique nonbinary computational systems. For a macro instance, eyes process "lines" and "spots" by combining neural patterns before the signals even reach the brain, and from this can identify vertical obstacles (like trees in a forest) and threats (like pairs of predator eyes) more quickly at lower energy.

But likewise, neurons are not limited to exchanging electrons via a single modality (voltage exceeding threshold). Neurotransmitter chemicals can alter local or systemwide thresholds for certain cells; this amounts to massive preprocessing that silicon can only roughly imitate with massive design efforts (which were effectively done by evolution in the organisms).

I think this is the wishful thinking of people who want to believe that consciousness is a deep mystery that won't be solved in their lifetimes. All while the capabilities of simulated neural networks are eclipsing humans in relatively short period of time.

It's possible human thought leverages quantum processes. That doesn't mean it's likely.

Eh, you’re giving technology too much credit and not considering what biology has developed to do naturally and far more efficiently. It’s also postulated that computational technology is nearing its current limits. Humanity’s hubris around technology is going to be our downfall, like how we think we can “stop” climate change with theorized technology even after we let the climate go too far into the red. We are greatly overestimating and exaggerating what tech can do.

DDG’s bot’s summary of the linked article and one other:

“The computational power of the human brain is often estimated in floating-point operations per second (FLOPS), with estimates ranging from 10^12 to 10^28 FLOPS, depending on the level of detail considered. In comparison, modern supercomputers can perform up to several hundred petaflops, but the brain is believed to operate at a similar or even higher efficiency due to its unique structure and processing capabilities.”

https://foglets.com/supercomputer-vs-human-brain/

No idea how they got 10^28. There are only 10^11 neurons and their firing rate is 2*10^2. Even if you assume they do fp16 accumulation (which I doubt, it is unlikely they are that precise), that adds maybe 10^5 to the total of 2*10^18. That's a very optimistic estimate.

I assume it was a typo, and they meant 10^18.

The 'capabilities' of simulated neural networks have nothing to do with consciousness.

I just don't find that convincing. If every aspect of human planning, skill, and knowledge are reproducable via classical computing then it seems unlikely that consciousness is some special case that requires quantum processes.

> If every aspect of human planning, skill, and knowledge are reproducable...

Neither the article nor the paper discuss consciousness, only biological computing capacity considering quantum states.

https://www.science.org/doi/10.1126/sciadv.adt4623

Today's classical computing accomplishments are impressive, but nowhere near what a human can do. Perhaps someday soon a single AI could (in the words of Heinlein) "change a diaper, plan an invasion, butcher a hog, conn a ship, design a building, write a sonnet, balance accounts, build a wall, set a bone, comfort the dying, take orders, give orders, cooperate, act alone, solve equations, analyze a new problem, pitch manure, program a computer, cook a tasty meal, fight efficiently, die gallantly" but that day is not yet here. If we struggle to build such a device, we might consider that our computers are too limited.

Define 'consciousness' before proceeding.

Meh, the quantum computation capacity of a cell may well be vastly larger than the 'classical' estimates - just like chemistry itself.

But the results boil down to what we observe - signal spikes between neurons, which seem to be not that mysterious.

I don't think anyone's disputing that classical neural activity is primary. But the question quickly becomes: what decides and coordinates that neural activity? It is it just an emergent effect of each neuron firing on instinct, producing an output for every input? Or is there some non-local coordination (quantum coherence) that calculates and ensures a correct firing sequence?

Last I checked the evidence was pretty slim for the later. But it's not zero. (Check out Orchestrated Objective Reduction theory). We've seen quantum coherence in microtubules, during photosynthesis, and superradiance in proteins. But no direct evidence of non-local effects in a living brain. Yet.

Reply on news.ycombinator.com