Again Against AI Hysteria
In recent times the internet has been awash with a lot of outrage and fearmongering about a new chat bot. Such bots have been around for decades of course, but this one has been trained with more data, and it has impressed more people. The initial thrill that a computer program can converse with you is gone, and now the new thrill is that the chat bots can pass standardized college entrance exams, too.
These impressive achievements – in a few very constrained domains of human verbal behavior – have inspired all kinds of Science Fiction fueled doomsday fears. The reasoning from a piece of software which was trained with a lot of material available online passing a relatively hard exam to the complete subjugation of the human race is shady at best, click-bait at worst. I think parts of these fears are coming from the deep psychology of American religiosity, which has a grasp on a lot of people over there even if they are not members of a traditional religion. The “Singularity” is nothing more than the Salvation with better gadgets and apps:
What Will Really Change?
Living in the lovely Philippine countryside also saves me from being too panic-y about the AI takeover. Will AI chop our neighbor’s wood? Catch the other neighbor’s fish? Will AI eat the eggplant in our garden? Will it direct us how to grow bigger eggplants? How will it take over our neighborhood?
Will it take over the island’s power grid? With the math skills shown below? The grid already doesn’t work too well, we frequently have a day without electricity. We survive that just fine, spending a day on the beach. How would the AI even get its hold onto the switches of the grid? Will there be robots guarding the entrance of the power company to make sure no anti-AI rebel enters the building and disconnects the AI overlord from the control boards? There are actually armed communist rebels on our island, but they fight the police, not chatGPT. Why would the great AI overlord even bother with the power grid of an island with the Philippines? So that the people who now raise chickens and grow sugarcane will instead make more AIs? None of that makes sense.
I think a big part of the world lives more like the Philippine countryside, instead of like the folks in the tech bubbles of central Europe and northern California who seem most struck by AI-takeover-fear.
A Few Interesting Takes
Not everybody is full of fear and awe of the new developments in “AI”.
My friend Andy Davis did an interesting little experiment, asking the bot what the correct maximum depth for a certain diving gas mix (nitrox, air enriched with more oxygen, at 36% instead of 21%, as in the atmosphere):
The correct answer is 91 feet – 28 meters – to stay below a partial pressure of oxygen of 1.4. Following our future AI overlord’s superintelligent advice would certainly kill you. This is really basic math!
This is also a reasonable take: “But the causal story that starts with a GPT-5 or GPT-4.5 training run, and ends with the sudden death of my children and of all carbon-based life, still has a few too many gaps for my aging, inadequate brain to fill in.”:
If AI scaling is to be shut down, let it be for a coherent reason
This is also a good take: Humans are still much better at spewing nonsense, I generally like Jacobin mag, old-school, rational working class rights leftism.
Finally, my friend Jay and I just submitted a manuscript that any intelligent structure rivaling the human brain will probably run about the same number of computations per second as a human brain. Given the massively worse energy-efficiency of present-day chip-based computers, this will not happen anytime soon. Funny mimics of some aspects of human verbal behavior? Yes. Great performance in tasks which digital computers are made for (math-like games like chess). Yes! Super-human intelligence and world-takeover? Very unlikely.
Here is the pre-print of our manuscript:
