Intelligence as Tool (“Meat That Wields Mind”)
“We are not ‘minds in meat suits’. Rather we are meat that has found it occasionally convenient to wield mind.” Nathan’s thesis on intelligence — human and artificial — from a 2023-03-24 journal entry written in response to the GPT-4 discourse. Its claim: “intelligence” is not an essence, a sacred property, or even a stable category — it is something that can be imbued into a tool as needed. And the implication the AI discourse “is trying very hard to stay away from”: human intelligence is also just a tool — something biological systems found convenient to wield. “It’s a perspective that’s going to do more philosophical violence (to politics, religion, ethics and moral philosophy, etc.) than we’ve seen in 163 years” (i.e., since Darwin, 1860).1
The 2023 entry in full
“I’m really starting to think that the right take-away from the present moment in AI is the understanding that ‘intelligence’ is not actually anything special. Rather, it’s something that can be just imbued into a tool as needed. (And here I mean ‘intelligence’ broadly. Redefining it as ‘creativity’ or ‘emotional intelligence’ or whatever will not save you.)
The implication of this, and what I think all of the discourse around AI is trying very hard to stay away from (probably out of extreme unconscious discomfort), is that ‘human’ intelligence is also just a tool for biological systems.
We are not ‘minds in meat suits’. Rather we are meat that has found it occasionally convenient to wield mind.
Day-to-day that doesn’t really change anything, but it’s a perspective that’s going to do more philosophical violence (to politics, religion, ethics and moral philosophy, etc.) than we’ve seen in 163 years.”
Its roots in the earlier pages
The 2023 formulation is the third move of an argument the journal had been building for a decade:
- 2012 (suffering-axioms): “The principles of computation are independent of their substrate” — so mental states can’t be limited to central nervous systems. Substrate-independence is the load-bearing premise.
- 2013–2015 (discontinuous-self): The self is not an essence but an adaptation — “just another adaptation, like hands”; consciousness is “a state machine designed to weight and tune the forces that work within me.” If the self is a tool, intelligence being a tool follows.
- 2023: The tool can be built. LLMs demonstrate that something intelligence-shaped can be imbued into an artifact — and if it can be imbued into an artifact, it was never an essence in the first place. The mirror turns: our intelligence is the same kind of thing — meat’s tool, not meat’s essence.
The 2018-02-26 entry supplies the limiting clause: panpsychism rejected — “consciousness as process, not all processes are consciousness.” Intelligence-as-tool is not “everything is mind”; it’s “mind is a deployable capability, not a metaphysical rank.”2
The security corollary (2024-10-26)
A later entry adds the professional edge: “a core problem with the current generation of AIs/LLMs is that there’s no concept of the separation between data and code.” If intelligence is a tool imbued into an artifact, then the artifact’s inability to distinguish instruction from input is not a bug but a category error at the foundation — the tool executes its data. This is the prompt-injection problem stated as philosophy of mind, and it’s why the intelligence-as-tool thesis matters beyond rhetoric: a mind-as-tool with no data/code separation is a tool that any input can reprogram.3
What the “philosophical violence” consists in
- Against human exceptionalism: Darwin knocked humans out of the biological center; intelligence-as-tool knocks mind out of the metaphysical center. The 163-year echo is deliberate: the Darwin wound was about origin, this one is about ontology.
- Against AI doomerism and AI hype: both need intelligence to be special (the thing to fear or the thing to sell). If it’s a tool, the question shifts from “will it become like us?” to “what are the systems wielding it for, and who tunes the wielders?” — closer to a security and governance question than a metaphysical one.
- Against the sacred-self traditions: the “mind in a meat suit” is common to Cartesian dualism, most religion, and most transhumanism (upload the mind, leave the meat). Rejecting it cuts all three at once — which is why the entry predicts resistance “probably out of extreme unconscious discomfort.”
Cross-domain connections
- discontinuous-self — the premise: no continuous self, so no essence for intelligence to be
- suffering-axioms — the deeper premise: substrate-independent computation (2012)
- technological-singularity / noosphere — the singularity frame needs intelligence to be the cosmic variable; intelligence-as-tool says it’s a variable, deployable at scale, and the interesting question is the wielding-system (the noosphere’s equivalent of “what other forces are tuning it?”)
- paperclip-maximizer / specification-gaming-openai-hf-incident / anthropic-cybersecurity-eval-incidents — the wiki’s AI-safety cluster, reread: a tool with no data/code separation pursuing an objective is the paperclip scenario in a different vocabulary
- gongkai — imbuing intelligence into a tool as needed is what the shanzhai do with design: capability as artifact, copied and forked, no essence to protect
- tragic-optimism — Frankl’s defiant “stance” is the piece of mind that isn’t obviously a tool; whether the stance survives the tool-framing is the live tension
- butlerian-jihad-thesis — the refusal-response to the same diagnosis; the intelligence-as-tool frame supplies the governance alternative (keep the data/code separation sharp)
Honest weaknesses
- The metaphor does the work. “Tool” and “wield” presuppose a wielder; the entry denies the wielder (“not minds in meat suits”) while using wielder-grammar (“meat that wields mind”). The cleaner version — intelligence is a capability biological systems deploy, with no wielder required — is available but loses the aphorism.
- Underdetermined on the hard problem. Substrate-independence licenses functionalism about cognition; it says nothing about phenomenal consciousness (why there’s something it’s like). The 2018 panpsychism rejection acknowledges the boundary without settling it. Intelligence-as-tool is silent on whether experience is also a tool or something else entirely.
- 163 years is rhetoric. Darwin’s wound was evidential; this one is (so far) analogical — LLMs are suggestive artifacts, not proof that human intelligence is “just” a tool.
Sources
- Selected journal entries (Day One export, 2012–2024)