AI Is Neither Artificial Nor Intelligent
Why “artificial intelligence” is brilliant marketing and bad description, while “large language model” is exactly the right name.
In an interview with Ian Pilon, Cornell-trained statistician William M. Briggs (PhD in statistics, MS in atmospheric physics; former Air Force cryptologist and weather/climate forecaster who later focused on philosophy of science) discusses AI, large language models (LLMs), reason, and related philosophical issues.
He publishes as “statistician to the stars” and is the author of Uncertainty (on the philosophy of probability and modeling) and a book on everyday fallacies (second edition forthcoming).
watch the full interview below:
Briggs treats probability as an extension and completion of logic: it measures relationships among propositions under uncertainty and resides in the mind. Knowledge requires true belief; belief itself is a further mental act beyond assigning probability.
He rejects George Box’s slogan “all models are wrong, some are useful,” noting that some models can be true while even false ones (e.g., Paul Ehrlich’s failed population predictions) can still prove useful to their creators.
AI is neither artificial nor intelligent. Computers (and abacuses) merely manipulate physical states—voltage levels or beads—to which humans assign meaning. There is no intellection, grasping of universals, or genuine reasoning inside the machine. Terms such as “neural nets” (nonlinear regression) and “genetic algorithms” are effective marketing, not accurate descriptions. Substitute “computer models” for “AI” and the hype diminishes.
LLMs, by contrast, are well-named: they are large statistical language models, fitted via nonlinear regression with billions of parameters. They excel at average-case behavior (search, pattern matching, image generation where all pixel-level causes are known and labeled) but fail at extremes and lack true understanding.
Output is probabilistic simulation, not reasoning. Arithmetic and letter-counting errors in early models illustrated the absence of hard-coded logic; later improvements came from engineering, not emergent intelligence.
Agents and “deterministic” workflows still depend on human-programmed constraints and routing. LLMs remain non-deterministic unless deliberately restricted; they route queries according to training patterns rather than comprehend goals. Claims of agents “going rogue” or approaching AGI are overstated.
Sentience and AGI are impossible. Briggs offers several arguments:
- An abacus is not intelligent no matter how large it grows; voltage states in a computer are no different.
- The “paper test”: the complete state of a computer at each clock cycle can be written as a stack of sheets of 1s and 0s. If the computer somehow achieves AGI, so does the paper stack—an absurdity.
- John Searle’s Chinese Room and related arguments distinguish syntax (formal symbol manipulation) from semantics (meaning), which only minds supply.
- Treating brains, ant colonies, or everything as computers leads to pantheism or an infinite regress of homunculi; meaning still requires a mind.
Humans reason by apprehending terms, grasping essences and universals (e.g., what a chair is), and performing logical and inductive operations—capacities computers can only mimic via pattern association. No one fully understands the “hard problem” of consciousness, but this does not prove reasoning is impossible; the fact that it occurs is evidence it is possible.
Where models succeed. When causes and conditions are fully known (as in images), models can generate high-quality outputs—cat pictures, movie scenes—under tight control. In domains with incomplete causal knowledge (medicine, complex social systems), performance is limited.
Briggs is especially critical of scientism, which he sees as a dominant secular religion that smuggles moral and political claims in under the banner of “the science.” Science describes what is; it cannot by itself dictate what ought to be done.
Phrases such as “follow the science” usually mean “follow me.” Complex systems (bodies, societies, quantum mechanics) leave wide room for dispute; declaring “the science” settled is often propaganda. Education research frequently illustrates the pattern: limited genuine progress is overlaid with moral fashion.
People rarely change entrenched beliefs because those beliefs are useful to them, even when they mismatch reality. Intellectual abilities are unequal, contrary to egalitarian myths; most people manage adequately with the models that already serve their needs.
Briggs can be found at his Substack (WM Briggs), YouTube channel, and on X/Twitter.
source www.youtube.com