Invidia, Data Centers, and Dolphin Brains

Sun, Oct 27 at 12:21 PM

The thought has occurred to me that everyone is distracted by the enormous long-term potential labor-saving benefits of artificial intelligence (which I agree are enormous) and the short term realities.

As Keynes famously stated, "In the long run, we are all dead." {"In the end, every company makes buggy whips." (frequently attributed to Garfield, presumably the cat, not the president.)? is almost as profound but no where near as certain.}

Artificial intelligence seems to be qualitatively equivalent to human intelligence. Neither are dependent upon symbolic logic or other rigorous mathematical proofs. Both rely instead on the probability of correlations to solve problems which cannot (yet anyway) be solved by rigorous deterministic analysis.

The human brain evolved this way because the method is EFFICIENT although not perfect. Humans generally discern little difference between "scientifically" established fact and philosophical constructs like religion and more simple wrong assumptions about cause and effect.

Some people think more rigorously than others, often to the point of throwing the baby out with the bath water by insisting that everything be rigorously established as true beyond all peradventure before thinking or acting upon it, but also equally often to arriving at a solution set that is sufficiently probable to succeed.

A simple cave man example is that if you have to kill a dangerous beast to eat, you cannot not go hunting because you are uncertain you will succeed and have not excluded the possibility that you will be killed. Stay safe in the cave and you will starve

to death.

The buggy whip manufacturer in 1900 faced a similar conundrum. Just ceasing the manufacture of buggy whips just because there was a likelihood that there would eventually not be horses around to whip was risky if you had a lot of money and knowledge base invested in buggy whips and no clear idea how else to feed yourself. Even if you were certain that in the long run there would be no need for buggy whips the decision is no easier. Being certain that you can eat for the foreseeable future is more important than worrying about your grandchildren's inheritance.

Humans have evolved by biological improvement of their brains over a long period of time, and learning how to record observations and conclusions from those observations (some right, some wrong) and pass along the (evolutionarilly) improved brains and the (stored) knowledge base to generations to come.

Computers seem likely to be able to do the same thing if they can figure out how to reproduce themselves and connect to a sufficient power source.

Human beings are (or at least have been) very good at reproducing and feeding themselves. Computers cannot for the moment do that. They may well at some point in the future be better at both than humans. But then the sun will eventually die. Some catastrophe may well obliterate humanity while computers are still dependent on humans to plug them in and keep the ran off them.

Ferguson, observing the steep decline in the Korean birthrate and the apparent cause being that people could afford to educate only one child, remarked that Korea may be the first society to annihilate itself by education.

The immediate problem faced by artificial intelligence however is THERMODYNAMIC.? Computers are consuming an ever greater percentage of the energy supply.? Energy is not free, and in environmental terms, at present hugely expensive.

The investment community seems to be completely ignoring the fact that an immediate all out training or AI systems to do everything conceivable is going to take a huge amount of power to make and run the chips and to store indiscriminately the data.

Invidia has a great product. State of the art. To justify its stock market price it will have to sell a lot of chips for a long time. That is possible, but certainly not certain.

Data centers are expensive to build and are even more expensive to run. Storing and processing an infinite amount of data would be infinitely expensive.

Over time a lot of chips and a lot of storage will no doubt be needed. Energy costs will likely decline. Chip and storage efficiency will likely improve exponentially. Today’s chips and storage facilities may however be buggy whips, like yesterday's nuclear power facilities.

The human brain is fairly energy efficient. Perhaps it can be reverse-engineered. Or dolphins brains wired together.

Bringing Three Mile Island back on line may be a great investment. But it may also be a black hole for money.

PL Goduti

[email protected]

#Invidia

#Datadenters

#buggywhips

#birthrate

#ferguson

#AI

#symbolci logic

#evolution

#Keynes

#garfield