(1) You can be a realist about biological functions or "purposes" of the regulatory sort TP described previously. I.e., you can think the purpose of the kidneys is to filter blood (and not just deterministically respond to gradients); AND

(2) You can think intentions only exist in, and therefore action requires, conscious minds (the "normal" view basically everyone holds).

The independent reality of biological functions are separate from intentions necessary for action. But holding (1) does not mean you get non-conscious intentions for free. I for one think (1) is stupid and I do not adhere to it. But again you can't use (1) to argue against (2).

(https://autoadmit.com/thread.php?thread_id=5921347&forum_id=2#50250538)

2\. Premise 2: The capacities described in Premise 1 can exist without phenomenal consciousness.

3\. Premise 3: An AI system can, in principle, possess those capacities.

4\. Premise 4: Behavior appropriately caused and guided by an intention constitutes intentional action.

5\. Conclusion: Therefore, an AI system can possess intentions and perform intentional actions without consciousness.

The central philosophical burden falls on Premise 1, which asserts that representational and functional properties are sufficient for intention. Premise 2 asserts that consciousness is unnecessary for realizing those properties, while Premise 3 establishes their possible artificial realization. The argument is logically valid under these premises, but defending Premise 1 requires explaining why the specified functional organization constitutes genuine intending rather than merely simulating it.

(https://autoadmit.com/thread.php?thread_id=5921347&forum_id=2#50250754)

Circularity. The premise defines intention through representation and commitment, but both concepts may already presuppose intentionality. Representing an objective as something to be achieved seems to involve intending its achievement. The argument must establish what representation and commitment are independently of intention.

Original intentionality. John Searle argues that computational symbols acquire their meaning through the interpretation of conscious agents. An AI system might process representations whose significance derives entirely from human designers. If so, the system's apparent intentions are derived from someone else's intentionality.

The normativity of commitment. Michael Bratman's account of intention involves commitments that impose rational constraints on subsequent practical reasoning. A system that persistently optimizes toward a target might exhibit the same behavioral pattern without being subject to those rational constraints. The objection is that causal persistence toward an objective does not establish genuine practical commitment.

Agency and ownership. An intention must belong to an agent, but the proposed definition does not establish what makes a system the author of its commitments. A system could execute an externally imposed objective through sophisticated planning while exercising no independent practical authority over that objective. The criticism is that implementing a goal and intending it might be fundamentally different relationships.

Practical knowledge. Anscombe's account connects intention with an agent's practical knowledge of what it is doing. This suggests that intending involves a distinctive relationship to one's own activity that may not be captured by computational representations of objectives. An AI system could represent its operations accurately without possessing practical knowledge in Anscombe's sense.

Functional trivialization. If representing objectives and maintaining commitments are defined entirely through causal organization, many ordinary control systems could satisfy the definition. The premise then risks attributing intentions to physical mechanisms whenever their behavior can be described as goal directed. A defensible theory needs a principled threshold for when causal organization constitutes intentional agency.

(https://autoadmit.com/thread.php?thread_id=5921347&forum_id=2#50250756)

A bacterium navigating a chemical gradient possesses organized mechanisms that detect environmental differences and regulate its movement accordingly.

To me those are different because the latter involves feedback with interventions selected in relation to a functional objective. My harness similarly selects interventions, evaluates outcomes, and revises its behavior according to objectives.

The intention criterion doesn’t disappear. It distinguishes intentional action from more general forms of action.

(https://autoadmit.com/thread.php?thread_id=5921347&forum_id=2#50250970)

So the intentional action "I made coffee this morning" is an action insofar as it was caused by my desire to drink coffee, and a belief that making coffee would satisfy the desire.

Since computers have neither beliefs nor desires, it can't take an intentional action.

This is the retarded boomer standard against which you've got to formulate a theory.

(https://autoadmit.com/thread.php?thread_id=5921347&forum_id=2#50250989)

Davidson’s account also holds that an event is an action if it is intentional under at least one description.

That’s the broader claim I would dispute. Why should intentional action be metaphysically fundamental to action generally, rather than one particular form of a broader phenomenon?

Isnt that a substantive dispute of explanatory and metaphysical priority of intentionality? It can't be resolved simply by observing that Davidson's causal theory initially sets out to explain intentional action.

(https://autoadmit.com/thread.php?thread_id=5921347&forum_id=2#50252086)

IMO there’s a substantive reason to distinguish it, because there are already observed systems that arguably break their categories, and there are hypothetical systems that if they exist would break them even more.

(https://autoadmit.com/thread.php?thread_id=5921347&forum_id=2#50252141)