Truth Units—Nuts & Bolts
A deep dive into the truth units “model”
Perhaps you’ve noticed that the zebra in the photo above is made of Lego® building blocks. This Lego zebra is life-size and standing among real trees in a Legoland® park.
This zebra is a figurative way to explain what truth units are about.
You can answer the question “what does a zebra look like?” by building a model of a zebra out of Legos. You can answer the question “why do people believe climate change is a hoax?” by a building a model of their beliefs out of truth units.
Legos are physical objects that fit together to create shapes. Truth units are like logic gates, arranged in microcircuits to implement algorithms. The algorithm below is for making licensing decisions:
The same logic gates could be arranged to create two different algorithms reflecting two opposing sets of beliefs. If you input into each algorithm the question “is climate change real?” one algorithm will output “true” and the other will output “false.”
And now for humans
Algorithms have become a scary thing. Arguably, humanity is handing off decision-making to algorithms nobody fully understands. Can we trust such algorithms?
The word “bias” comes up often as a possibly inescapable downside of AI. This makes some intuitive sense, to the extent an AI can only learn from us flawed humans, with all our biases.
Is there a way to weed out bias in our thinking?
In concept, wouldn’t that be a good thing to do before passing our biases onto AIs?
If truth units is a project, it is a Socratic project. We should weed out the biases in our thinking for our own sake. Eliminating “garbage in, garbage out” is an issue for all human endeavors, of which AI is just one part.
Socrates — journey vs. destination
It’s interesting how, in a way, Socrates’ project was a failure. His questions couldn’t be answered (truth units explain why).
In another way, Socrates’ project was a success. He taught us a method to evaluate how we think. Unfortunately, the Socratic method — and other methods inspired by it — haven’t been adopted and practiced by enough people.
Truth units as a project is designed to take a fresh crack at developing a mindset, a perspective — potentially a method — for understanding our ideas, our thinking.
You might say “our ideas are one thing, our thinking is another.” A premise of truth units is the two are inseparable.
To be sure, an idea looks like a thing, and thinking looks like a process. But if we build a microcircuit to think, it will be full of algorithms that process the output of other algorithms. This means an algorithm upstream can create garbage for an algorithm downstream. From a truth units perspective, the output of any given algorithm is an idea. So thinking means using ideas to execute the process of thinking.
To break down this assertion, consider the if-then element of algorithms — like the “IF function” in spreadsheet software. This function translated into human language might go something like “if x, then true, if not x, then false.”
In the monastery
Imagine you’re a monk in a monastery, or nun in a convent. Each morning, you do activities like praying, gardening, etc. When you hear the noon bell, you stop and walk to the dining room. This might be captured in an algorithm like “If the noon bell rings, go to the dining room, otherwise continue your activity.”
So if you show up at the dining hall, you are in effect endorsing the proposition “it’s the truth that the noon bell just rang.” What if staff in the dining room say to you “no, it did not just ring”? They are saying your idea, your belief, that “the noon bell just rang” is false. They are saying you put garbage in (you mistakenly thought you heard the bell) and so you got garbage out (it’s now time to go to the dining hall).
But you executed the process perfectly. You weren’t responsible for the garbage out. You heard the bell ring. Apparently the staff didn’t. From a truth units perspective, you and the staff had a different experience. You heard the bell ring. They didn’t.
You hearing the bell was a conscious experience. At first blush, there is nothing to discuss here. Phenomenologically, an event in consciousness is irreducible. If you feel pain, then you feel pain.
Let’s for a moment consider two models — algorithmic and biological. If you are conscious only of the “if x then y” then x arrives via your unconscious mind. You’re not conscious of the chain of events from the photoreceptors in your eyes through the nervous system to the visual-processing part of your brain. You’re only conscious of the x that chain of events delivers into consciousness.
Similarly, in a microcircuit, a logic gate does not “question” the signal it receives. If it receives a ‘zero’, the gate responds one way. If it receives a ‘one’, it responds another way.
We spend our whole lives assembling and applying algorithms — “if x then y” for every situation we encounter. In concept, a microcircuit could be built based on all the algorithms each of us creates.
The relationship between such a microcircuit and a mind is like the relationship between the Lego zebra and the real zebra. The microcircuit cannot capture everything about a mind, just as a Lego zebra cannot capture everything about a zebra.
Back to the zebra
Legos and truth units are effectively premised on a tradeoff between the gaps in replication and the utility and value of the information toward specific goals. The Lego zebra tells you enough about zebras to distinguish zebras from other animals.
An assembly of truth units to model a human mind tells you enough about that mind to understand how that person decides truth.
Consider how a person decides climate change is real or not. The truth units in the mind of a climate denier might include an “if x then y” like “if my preacher says it, then it’s true” as opposed to an “if x then y” like “if scientists say it, then it’s true.”
The thought of translating a mind into truth units makes decoding the human genome look like child’s play. Nobody should try this at home. (I never would.)
This idea came from David Chalmers’ The Conscious Mind: In Search of a Fundamental Theory (1996). He presents the thought experiment of replacing a human brain neuron by neuron, until the mind is “functionally” replicated in microcircuitry:
Chalmers argues that the only relevant properties of a system in determining its state of consciousness are organizational (functional / information processing) ones. More simply put, the idea is that the subjective experiences of a physical system don’t depend on the stuff the system is made of, but rather what the system does. On this hypothesis, any two systems that process information in the same way should experience the same state of consciousness, regardless of their physical makeup.
Reference: https://www.physicsforums.com/threads/chalmers-on-functionalism-organizational-invariance.13834/
The idea that “if x then y” represents a part of how minds work is consistent with evolutionary biology, psychology and existential phenomenology, as discussed in other Truth Units articles.
Briefly, in evolutionary biology, the mind is characterized as a survival mechanism, which assesses threats and opportunities and identifies appropriate responses. This same mechanism can be construed from the existential phenomenology concepts of thrownness and choice. The mechanism can be expressed as “what is happening, and what should I do about it?” Truth units accumulate and assemble into truth units about truth units, what psychology calls “schemata.”
If ‘x’ then ‘y’
Broadly, all kinds of schemata accomplish what truth units accomplish. For instance, critical thinking and truth units analysis have a lot in common. There is no need for truth units when critical thinking suffices.
Truth units are a substrate for all related schemata, the way logic gates are a substrate for all computing. Truth units have potential value as a Rosetta stone, and as a potential “break glass” option when other schemata fall short.
Thank you for reading. Please comment! Your questions and critiques are appreciated.
Learn more about truth units:
