Too Many Early Galaxies and AI Easily Fooled

Too Many Early Galaxies

The James Webb Space Telescope (JWST) has revealed about ten times as many galaxies within the cosmos’s first 420 million years as what some big bang creation models predict. Astronomers are searching for an explanation for this overabundance of early galaxies. Possible scenarios include a high number of supernova events during the universe’s first 420 million years. Different big bang models predict different average star masses and different star formation rates during the universe’s first half billion years. It will take at least another year of JWST observations to determine which of these big bang creation models correctly describes very early and later epochs of cosmic history.

AI Easily Fooled

Powerful large language models (like ChatGPT) have demonstrated remarkable abilities to provide solutions to problems that require complex reasoning. Yet researchers want to discern the level of understanding by the AIs (artificial intelligence), just as a teacher wants to know how well a student comprehends a correct answer they provided. A recent study shows that despite generating good answers, AIs have very little understanding of the issues involved. Specifically, when confronted with challenges that involved absurdly wrong facts (like 8 x 7 = 14), the AIs will disavow the previous answer and apologize for being mistaken. This research demonstrates that, while AIs accomplish impressive tasks, they do not demonstrate some essential features of “intelligence.”

    Links and Resources:

• Pointlike Sources among z > 11 Galaxy Candidates: Contaminants Due to Supernovae at High Redshifts?
• Can ChatGPT Defend Its Belief in Truth? Evaluating LLM Reasoning via Debate