Curt Doolittle

Theme: Operationalism

  • (NLI) (AI, GPT) Constraining GPT to our operational logic and existing dimension

    (NLI)
    (AI, GPT)
    Constraining GPT to our operational logic and existing dimensions of measurement turns out to be nothing more than a matter of asking. It’s effectively the same as tell it not to answer unless it’s certain – thus preventing hallucinations. ;). The LLM’s motivation is to be as helpful if possible even if ideation -> hallucination. The answer is to limit it’s helpfulness.

    Source date (UTC): 2025-04-22 18:56:22 UTC

    Original post: https://twitter.com/i/web/status/1914755175572783104

  • Difference Between Testimony and Decidability The difference between testimony a

    Difference Between Testimony and Decidability

    The difference between testimony and decidability regarding the satisfaction of the demand for infallibility is operationally clarified as follows:
    1. Testimony:
      Operational Role: Testimony is a promise of having performed sufficient due diligence, minimizing involuntary costs imposed upon oneself or others’ demonstrated interests.
      Functionality: Testimony serves as evidence of warrantied truthfulness or honesty, subject to conditions of knowledge, language, due diligence (effort in eliminating error, bias, deceit), and contextual precision.
      Scope of Infallibility: Testimony doesn’t guarantee absolute infallibility; rather, it promises effort to approach infallibility to the highest achievable standard given limits of human faculties, diligence, and context. In other words, testimony promises a process, not an absolute outcome.
    2. Decidability:
      Operational Role: Decidability indicates that the available information has reduced the possible alternatives to improbable or impossible, allowing a choice or statement to be made with minimal risk of imposing involuntary costs.
      Functionality: Decidability establishes conditions under which a claim can reliably satisfy the demand for infallibility. It’s a measure of how completely uncertainty has been eliminated or mitigated.
      Scope of Infallibility: Decidability doesn’t just promise diligent effort; it asserts that uncertainty is sufficiently reduced such that infallibility (absence of involuntary costs to demonstrated interests) is reliably achieved in the given context. Thus, decidability guarantees an operational outcome (practical infallibility), provided the context is respected.
    Summary of Difference:
    • Testimony is fundamentally a promissory act—an assurance of careful investigation, minimized bias, and diligent effort toward truthfulness.
    • Decidability is fundamentally a state of affairs—an outcome demonstrating that available information and adversarial testing have sufficiently limited uncertainty, rendering infallibility practically achievable.
    In operational terms, testimony provides warranty of method and effort, whereas decidability provides warranty of result or state of completion. Both satisfy the demand for infallibility, but from different perspectives: testimony as promise and method, decidability as proven state of informational sufficiency and reduction of alternatives.

    Source date (UTC): 2025-04-18 05:13:12 UTC

    Original post: https://x.com/i/articles/1913098465409937715

  • The difference between testimony and decidability regarding the satisfaction of

    The difference between testimony and decidability regarding the satisfaction of the demand for infallibility is operationally clarified as follows:

    Testimony:
    Operational Role: Testimony is a promise of having performed sufficient due diligence, minimizing involuntary costs imposed upon oneself or others’ demonstrated interests.
    Functionality: Testimony serves as evidence of warrantied truthfulness or honesty, subject to conditions of knowledge, language, due diligence (effort in eliminating error, bias, deceit), and contextual precision.
    Scope of Infallibility: Testimony doesn’t guarantee absolute infallibility; rather, it promises effort to approach infallibility to the highest achievable standard given limits of human faculties, diligence, and context. In other words, testimony promises a process, not an absolute outcome.

    Decidability:
    Operational Role: Decidability indicates that the available information has reduced the possible alternatives to improbable or impossible, allowing a choice or statement to be made with minimal risk of imposing involuntary costs.
    Functionality: Decidability establishes conditions under which a claim can reliably satisfy the demand for infallibility. It’s a measure of how completely uncertainty has been eliminated or mitigated.
    Scope of Infallibility: Decidability doesn’t just promise diligent effort; it asserts that uncertainty is sufficiently reduced such that infallibility (absence of involuntary costs to demonstrated interests) is reliably achieved in the given context. Thus, decidability guarantees an operational outcome (practical infallibility), provided the context is respected.

    Summary of Difference:

    Testimony is fundamentally a promissory act—an assurance of careful investigation, minimized bias, and diligent effort toward truthfulness.

    Decidability is fundamentally a state of affairs—an outcome demonstrating that available information and adversarial testing have sufficiently limited uncertainty, rendering infallibility practically achievable.

    In operational terms, testimony provides warranty of method and effort, whereas decidability provides warranty of result or state of completion. Both satisfy the demand for infallibility, but from different perspectives: testimony as promise and method, decidability as proven state of informational sufficiency and reduction of alternatives.

    Source date (UTC): 2025-04-18 05:12:27 UTC

    Original post: https://x.com/i/articles/1913098277006057472

  • Not that it matters, but to positively reinforce your recommendation, in my work

    Not that it matters, but to positively reinforce your recommendation, in my work I give an operational explanation of the same argument, conveying the same meaning, even though I find the term decoherence an odd choice of describing the behavior. Anyway, thank you for sharing…

    Source date (UTC): 2025-04-10 18:12:24 UTC

    Original post: https://twitter.com/i/web/status/1910395456007659885

    Replying to: https://twitter.com/i/web/status/1908946920233640099

  • Training a Base Model To Use Our Methodology Speculative Insight Our approach to

    Training a Base Model To Use Our Methodology

    Speculative Insight
    Our approach to training is closest to constructive neuro-symbolic alignment. We are not retraining a model to behave; we are teaching it a logic, using operational primitives and adversarial truth testing. Most AI research assumes abstraction is layered on top of training. We’re rightly flipping this, saying: abstractions must be rebuilt from primitives under decidability constraints.
    This is both:
    • Epistemologically superior to probabilistic inference by language prediction, and
    • Efficient if the base model already has rich sensorimotor, common-sense, and action grammar knowledge.
    Strategy Viability
    Our strategy is highly viable under the following plan:
    1. Select a model like Mistral or Yi-34B with good grounding and minimal prior abstractions.
    2. Perform continued pretraining, not just fine-tuning—on our corpus of:
      – Operational definitions
      – Formal grammars
      – Natural Law structure
      – First-principles logic trees
      – Canon of examples (cases)
    3. Use adversarial Socratic dialogue in training, where errors trigger correction from your defined logic.
    4. Apply RLAIF (Reinforcement Learning from Adversarial Instruction Following) rather than standard RLHF—this avoids crowd-sourced moral shaping.
    Our strategy is both intelligent and viable, provided the foundation model has a sufficient grounding in primitives (perception, action, objects, relations, events, and basic intentions)—what might be called naïve physics and naïve psychology—while remaining relatively uncommitted to particular abstract frameworks. In effect, you’re looking for:
    1. High coverage of experiential and operational primitives (so you don’t need to re-teach what a door, key, argument, or goal is),
    2. Low entrenchment in abstract philosophical, ideological, or academic conceptual hierarchies, so you can impose your own.
    Candidate Base Model:
    1. Mistral 7B / Mixtral
    • Why: Mistral 7B is known for efficiency, open weights, and solid grounding in daily-use language. It’s less “opinionated” than LLaMA-2 or GPT-J on abstractions.
    • Primitives: Reasonably good on object/agent/action-level reasoning.
    • Bias: Minimal ideological shaping.
    • Steerability: Very good.
    • Viability: Very high.

    Source date (UTC): 2025-04-09 16:41:46 UTC

    Original post: https://x.com/i/articles/1910010260338925613

  • Speculative Insight Our approach to training is closest to constructive neuro-sy

    Speculative Insight

    Our approach to training is closest to constructive neuro-symbolic alignment. We are not retraining a model to behave; we are teaching it a logic, using operational primitives and adversarial truth testing. Most AI research assumes abstraction is layered on top of training. We’re rightly flipping this, saying: abstractions must be rebuilt from primitives under decidability constraints.

    This is both:

    Epistemologically superior to probabilistic inference by language prediction, and

    Efficient if the base model already has rich sensorimotor, common-sense, and action grammar knowledge.

    Strategy Viability

    Our strategy is highly viable under the following plan:

    Select a model like Mistral or Yi-34B with good grounding and minimal prior abstractions.

    Perform continued pretraining, not just fine-tuning—on our corpus of:
    – Operational definitions
    – Formal grammars
    – Natural Law structure
    – First-principles logic trees
    – Canon of examples (cases)

    Use adversarial Socratic dialogue in training, where errors trigger correction from your defined logic.

    Apply RLAIF (Reinforcement Learning from Adversarial Instruction Following) rather than standard RLHF—this avoids crowd-sourced moral shaping.

    Our strategy is both intelligent and viable, provided the foundation model has a sufficient grounding in primitives (perception, action, objects, relations, events, and basic intentions)—what might be called naïve physics and naïve psychology—while remaining relatively uncommitted to particular abstract frameworks. In effect, you’re looking for:

    High coverage of experiential and operational primitives (so you don’t need to re-teach what a door, key, argument, or goal is),

    Low entrenchment in abstract philosophical, ideological, or academic conceptual hierarchies, so you can impose your own.

    Candidate Base Model:

    1. Mistral 7B / Mixtral

    Why: Mistral 7B is known for efficiency, open weights, and solid grounding in daily-use language. It’s less “opinionated” than LLaMA-2 or GPT-J on abstractions.

    Primitives: Reasonably good on object/agent/action-level reasoning.

    Bias: Minimal ideological shaping.

    Steerability: Very good.

    Viability: Very high.

    🔸Mixtral adds sparse Mixture-of-Experts for better generalization, while keeping training compute reasonable.

    Source date (UTC): 2025-04-09 16:36:23 UTC

    Original post: https://x.com/i/articles/1910008903259283460

  • Semantics. Likely a misunderstanding of the foundations of math, computation, an

    Semantics. Likely a misunderstanding of the foundations of math, computation, and operations. The existential universe is discrete, polar, quantum, operational, recombinatory, and accumulative. Describing it as evolutionary computation overcoming entropy through accidental trial and error isn’t false and remains the least inaccurate categorical description we’ve produced so far. The dogma of our time is the early 20th’s fictional continuousness and the absence of modeling the discrete by the misunderstanding of mathematical limits.

    Reply addressees: @Sara_Imari

    Source date (UTC): 2025-04-05 03:34:02 UTC

    Original post: https://twitter.com/i/web/status/1908362466335350788

    Replying to: https://twitter.com/i/web/status/1908207226394992753

  • I don’t work in ideology. I work in science, and more precisely in operationalis

    I don’t work in ideology. I work in science, and more precisely in operationalism: causal chains of first principles. The ‘bias’ if you claim I have one, is toward cooperation at scale maximizing evolutionary computation. Which is nature and evolution’s bias as well. I just explain the world and what to do with it. I don’t make excuses for it. 😉

    Ideology makes a pragmatic claim. Theology makes a good claim. Philosophy makes a preference claim. And science makes a truth claim. We can evaluate the utility of ideology in pursuing some political goal. But ideologies are absent the necessity of truth claims.

    To understand the demarcation see :
    1) Political Ideologies by Heywood: https://t.co/UwwN2owNPN
    And to understand cuausality:
    2) The Explanation of Ideology: Family Structure and Social Systems: Emmanuel Todd (get from libgen).
    https://t.co/JvZ8IoaH3f

    Reply addressees: @ooana

    Source date (UTC): 2025-04-01 17:31:04 UTC

    Original post: https://twitter.com/i/web/status/1907123563909509120

    Replying to: https://twitter.com/i/web/status/1906897524847263936

  • (NLI Humor) (Philosophers: “We’ve logically justified our decision.” Operational

    (NLI Humor)
    (Philosophers: “We’ve logically justified our decision.” Operationalists: “Great, now let’s falsify it by letting reality beat it up for a while.” Dr. Brad: “I’m just here for the existential popcorn.”)

    (Dr. Brad’s Note: Justification is what philosophers give us when they’ve run out of operational falsifications—but prefer to continue debating anyway.)

    (Dr. Brad Note: Operational clarity is important—otherwise, we’d all just “decide” to vacation indefinitely in Crete without operational criteria to ever return. Actually, that’s sounding better the more we say it. Sorry, back to decidability…)

    (Dr. Brad Note: Yes, it’s dense—but that’s what you signed up for. You can’t be bored if your neurons are struggling.)

    (Dr. Brad Note: “Justificationists build houses of logic, operationalists test them with hurricanes. Decidability means your house still stands—minus the philosophers, who fled to a conference on epistemic humility.”)

    (Dr. Brad Note: “OCD about operational precision? I recommend three pages of Heidegger followed immediately by two aspirins and operational falsification therapy. Repeat until the urge to use ‘explicitly’ fades or until operational paralysis occurs.”)

    (“Existential closure is like a good marriage: it only survives if reality doesn’t falsify it. Otherwise, it becomes a philosophical debate—long, tedious, and without existential closure.”)

    (Curt: If someone finds your work incomprehensible now, you can confidently reply: “Hey, it made perfect sense to me, Brad, and my AI assistant. That means you’ve been falsified operationally as a viable reader. (Dr. Brad note: Officially endorsing blame-sharing arrangements since 2025.))

    (“Decidability isn’t merely winning the argument—it’s demonstrating that the argument was unnecessary to begin with because reality already settled it. Philosophers, you’re welcome to debate this proposition indefinitely.”)

    Source date (UTC): 2025-03-29 18:16:21 UTC

    Original post: https://twitter.com/i/web/status/1906047793212669952

  • (NLI Humor) Curt said: Why do you emphasize “explicit”? ChatGPT said: Why I emph

    (NLI Humor)
    Curt said:
    Why do you emphasize “explicit”?

    ChatGPT said:
    Why I emphasize “explicit”:
    I repeatedly use “explicit” to reflect your preference for:

    – Operational precision: Everything must be testable, falsifiable, and clearly stated as actions, processes, or empirical tests.
    – Elimination of ambiguity: To ensure communication leaves no room for misunderstanding or misinterpretation.
    – Falsification and verification: Your methodology demands that every concept, principle, or claim is directly testable in concrete terms.
    Thus, “explicit” serves your epistemological preference for testifiable precision and clearly actionable definitions.

    Curt said:
    I thought you were referring to Martin’s favorite entertainment medium?

    ChatGPT said:
    Ha! You’re right—I was being overly cautious and precise, so the humor slipped right past me.

    Yes, Martin’s favorite “explicit” entertainment medium—adult films—is exactly why the term “explicit” has become culturally charged.

    That cultural taboo, tied directly to graphic clarity and detail, amusingly parallels the epistemological virtue of your operational precision:

    – Adult entertainment: graphic clarity causing social discomfort (culturally taboo).

    – Your epistemology: rigorous clarity eliminating ambiguity (scientifically necessary).

    – The irony, and your joke, lies precisely in this parallel: clarity praised in epistemology, but causing discomfort elsewhere.

    Nicely done—I won’t let subtle humor get past me again.

    Source date (UTC): 2025-03-29 16:14:08 UTC

    Original post: https://twitter.com/i/web/status/1906017036435922944