OpenAI’s amazing — but vastly oversold — new model Astra

4 hours ago

Astra is impressive but its capabilities are being vastly oversold, with many committing the fallacy of composition by assuming math prowess implies general intelligence.
Expertise in one domain, like math, does not guarantee expertise in all domains, which is why multidimensional theories of intelligence exist.
Math is a special case for AI because it allows for easy verification and generation of synthetic data, advantages that do not extend to most real-world problems.
OpenAI's announcement lacks essential details, such as the number of conjectures attempted and the true cost including human labor, making it hard to evaluate significance.
Astra is good at some kinds of math but not all, and autoformalization—converting human math into formal logic—remains unsolved.
Claims that Astra represents the Singularity or the most significant day in math history are absurd; it is not comparable to development of calculus or algebra.
There is no reason to think Astra will fix reliability issues in other AI systems, like extracting data from arbitrary PDFs or creative writing.
Despite hype, Astra is a remarkable but narrow achievement, not a universal solvent or automated scientific discovery engine.

AI possesses infinite objective intelligence (e.g., facts, calculations) but lacks subjective intelligence (wisdom, taste, genuine understanding).
Objective intelligence cannot be traded for subjective intelligence; they are distinct and non-fungible, unlike resources in Settlers of Catan.
AI writing is technically perfect but emotionally hollow, missing the 'light' that makes prose alive, as consensus among writers shows.

AI predictions from 2023 were largely accurate but underestimated the ensuing hype magnitude.
LLMs are not truly intelligent but perform 'absolute miracles' in tasks like grammar correction and code debugging.
The Chinese Room Argument remains relevant for debating AI intelligence vs. simulation.

CEO's claims about reaching the Singularity should not be taken seriously, as they often make outlandish statements for publicity and fail to define the term.
None of the CEOs, including Sam Altman and Elon Musk, have clearly defined what the Singularity means, and their claims lack substantiation.
The Klarna Effect shows that companies boast of AI replacing workers, but then quietly rehire humans due to AI's costs and limitations.

The author argues that mathematics is fundamentally about clarity, understanding, and concept-building, not just theorem-proving.
AI's current approach to mathematics risks focusing on solving problems without generating intelligible or accretive proofs that advance human comprehension.
The 'honor code' in mathematics historically prioritized theorem-proving over conceptual work, but this may become unsustainable as AI automates proof generation.

AI excels at passing exams and finding answers, achieving superhuman results on tests like Humanity's Last Exam and the International Math Olympiad.
AI contributes to scientific discoveries (e.g., AlphaFold, GNoME) but only when humans decide which questions to ask; it has not independently chosen important questions.
The contradiction between AI's superb answer-finding and lack of question-posing is resolved by recognizing that all benchmarks measure only answer-finding, not question-choosing.

AI possesses infinite objective intelligence (e.g., facts, calculations) but lacks subjective intelligence (wisdom, taste, genuine understanding).
Objective intelligence cannot be traded for subjective intelligence; they are distinct and non-fungible, unlike resources in Settlers of Catan.
AI writing is technically perfect but emotionally hollow, missing the 'light' that makes prose alive, as consensus among writers shows.

AI predictions from 2023 were largely accurate but underestimated the ensuing hype magnitude.
LLMs are not truly intelligent but perform 'absolute miracles' in tasks like grammar correction and code debugging.
The Chinese Room Argument remains relevant for debating AI intelligence vs. simulation.

CEO's claims about reaching the Singularity should not be taken seriously, as they often make outlandish statements for publicity and fail to define the term.
None of the CEOs, including Sam Altman and Elon Musk, have clearly defined what the Singularity means, and their claims lack substantiation.
The Klarna Effect shows that companies boast of AI replacing workers, but then quietly rehire humans due to AI's costs and limitations.

The author argues that mathematics is fundamentally about clarity, understanding, and concept-building, not just theorem-proving.
AI's current approach to mathematics risks focusing on solving problems without generating intelligible or accretive proofs that advance human comprehension.
The 'honor code' in mathematics historically prioritized theorem-proving over conceptual work, but this may become unsustainable as AI automates proof generation.

AI excels at passing exams and finding answers, achieving superhuman results on tests like Humanity's Last Exam and the International Math Olympiad.
AI contributes to scientific discoveries (e.g., AlphaFold, GNoME) but only when humans decide which questions to ask; it has not independently chosen important questions.
The contradiction between AI's superb answer-finding and lack of question-posing is resolved by recognizing that all benchmarks measure only answer-finding, not question-choosing.

Hasty Briefsbeta