Scott mischaracterises the reasons, which concern capacities of the system
not properties of its output.
(1) It lacks the capacity to interpret sentences. It's output possess the property of seeming-to-have-interpreted.
(2) It lacks the capacity to learn. Learning is, in part, explaining; it is, in part, causal contact with an environment. It's output has the property of seeming-to-be-generated-by-a-learner.
(3) It lacks the capacity to speak either truthfully or otherwise. Lacking the capacity to explain, because at least it lacks the capacity to imagine (counterfactuals), it cannot thereby determine whether a claim accords with the way the world is. It's output has the property: seeming-to-be-true.
(4) It lacks the capacity to be moral. For all the above, and more: it lacks the capacity to care. It's output has the property: seeming-to-take-moral-stands.
The "Jesuit" here, I fear, is the AI fanatic who have turned their telescopes away from reality (intelligence as it actually exists), and have stolen the lenses to read scripture (intelligence as theorised by partisan academics).
One has to agree with Chomsky here at the end, "given the amorality, faux science and linguistic incompetence of these systems, we can only laugh or cry at their popularity."
The morality point I think should hit home most extremely: how horrifying to treat generated text output as-if it were constructed by an agent who cares.
Apparent properties of the output are not properties of the system; ChatGPT does not think, know, care, intend, speak, commnuicate, etc.
One can only "laugh or cry" at how absurd this sales pitch: what a horror to be invited to treat ChatGPT as possessing any of these capacities, simply because correlations across a whole internet of text seems to.
(1) It lacks the capacity to interpret sentences. It's output possess the property of seeming-to-have-interpreted.
(2) It lacks the capacity to learn. Learning is, in part, explaining; it is, in part, causal contact with an environment. It's output has the property of seeming-to-be-generated-by-a-learner.
(3) It lacks the capacity to speak either truthfully or otherwise. Lacking the capacity to explain, because at least it lacks the capacity to imagine (counterfactuals), it cannot thereby determine whether a claim accords with the way the world is. It's output has the property: seeming-to-be-true.
(4) It lacks the capacity to be moral. For all the above, and more: it lacks the capacity to care. It's output has the property: seeming-to-take-moral-stands.
The "Jesuit" here, I fear, is the AI fanatic who have turned their telescopes away from reality (intelligence as it actually exists), and have stolen the lenses to read scripture (intelligence as theorised by partisan academics).
One has to agree with Chomsky here at the end, "given the amorality, faux science and linguistic incompetence of these systems, we can only laugh or cry at their popularity."
The morality point I think should hit home most extremely: how horrifying to treat generated text output as-if it were constructed by an agent who cares.
Apparent properties of the output are not properties of the system; ChatGPT does not think, know, care, intend, speak, commnuicate, etc.
One can only "laugh or cry" at how absurd this sales pitch: what a horror to be invited to treat ChatGPT as possessing any of these capacities, simply because correlations across a whole internet of text seems to.