On Inorganic Intelligence
The world is a set of rules that modify a mutable state. In it, we are organic mammals with a complex lifecycle involving diploid somatic cells for individual survival, and haploid gametes to execute a species-wide genetic algorithm that optimizes our ability to survive in our environment, the interface between the solid and gaseous layers of Earth.
We have evolved through the best optimization algorithm we know of, a combination of an evolutionary algorithm and a chemical neural network electrically connected to sensory receptors and muscles. In order to force us to worry about our survival, the neural network is hijacked by the production of endogenous chemicals. Serotonin, for instance, forces our brain to look for food and mating partners. There are different chemicals involved in forcing us to act when in front of food, and in switching to different activities afterwards. There are chemicals that make us fear dark, strangers, blood. All those hijacking chemicals are regulated by endocrine glands, ie. organs that produce hormones, those molecules that modify and set our behaviour. For instance, the thyroid (just below the Adam apple on men) influences dopamine production, regulating mood swings.
A bad thyroid can lead to bipolar disorder or schizophrenia. Madness is simply acute humanity.
What is the point of my saying this?
Humans are machines with complex software and complex hardware. Of course, we don’t consider our hardware to be intelligent — we understand the reproduction and specialization of our cells to be the deterministic result of chemical interactions. I never quite understood why we often don’t extrapolate that understanding to our software.
These days, great people wish to raise awareness of the dangers of artificial intelligence. It is akin to worrying about the supply of entertainment to Ganymede, as what is feared requires technology that cannot happen within a hundred years. But what is that technology exactly?
What we call artificial intelligence can designate two things:
- An algorithm able to solve non-trivial optimization problems.
- An automaton capable of self-sustenance and coherent dialogue which passes the Turing test.
We use the former daily. CCTV, drones, language processors such as Google, speech recognition systems in Siri or Cortana, Facebook’s face detectors or your email’s spam filter, all include either genetic algorithms, neural networks, hidden Markov models or other machine learning systems that we consider artificial intelligence.
They do their task well. Any deficiency gets improved. No fear results from their use.
The latter somehow baffles minds, while being an application of the former. I believe that the fear it generates is religious. Surely Babel’s disaster happened because we were not meant to create? Yet we have created so much that it has become our primary job. Am I no longer a unique snowflake if something I made can think? if it can be mass-produced? If the machines can win at chess or Jeopardy, as they inexorably tend to, are we gods murdered by their creation?
However, I, who writes this, am an artificial intelligence. I am a self-sustaining automaton capable of intelligible thought. You are organic and so am I. I was not designed by humans, but by the same algorithms that we use to make the robots that help you on the Web.
Fear no AI for you are one. Were we to create an inorganic automaton with which we could talk, the most outstanding difference it would have would be an aim. It would be made to keep you company while driving, or to detail the benefits of varying products on the market. You don’t have a goal, as your species was formed randomly. They, on the other hand, are designed to please you; dysfunctional models would be thrown out. Deceit is disallowed by design.
While the imagination of some runs wild as to a hypothetic AI, there is a very real AI out in the wild causing tremendous risks to our health, and it is us. I have never lived a second of my life in a world where humans did not have the capacity to make Earth inhabitable. The atomic bomb was created long before me, and continues to be improved yearly. We are an AI gone rogue, but we survived it, because our stupidity was marginal; we are optimized for survival.
Additionally, while our inorganic AIs have a training phase of decades, our species has been through the best optimization algorithms for 3.7 billion years. It will be tremendously hard to make an automaton as energy-efficient and smart as us.
Even if we did, we pick the fitness landscape — the things that we want those inorganic automata to optimize. While we are optimized for survival on Earth, they would optimize our food supply, our economy, our diplomacy. Designing a fitness landscape that could harm us is as absurd as making double-edged knives with no handles. Just as I do not want to drive in a car whose airbag system could fail, we will replace an AI that fails our economy. And that AI would only replace humans doing a poorer job.
Should we fear intelligence? The gory depiction offered by Hollywood blockbusters only regurgitate the fear of Genesis’ forbidden apple. Let us be smarter than that.