English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Increase in computational scale and fine-tuning has seen a dramaticimprovement in the quality of outputs of large language models (LLMs) like GPT.Given that both GPT-3 and GPT-4 were trained on large quantities ofhuman-generated text, we might ask to what extent their outputs reflectpatterns of human thinking, both for correct and incorrect cases. The EroteticTheory of Reason (ETR) provides a symbolic generative model of both humansuccess and failure in thinking, across propositional, quantified, andprobabilistic reasoning, as well as decision-making. We presented GPT-3,GPT-3.5, and GPT-4 with 61 central inference and judgment problems from arecent book-length presentation of ETR, consisting of experimentally verifieddata-points on human judgment and extrapolated data-points predicted by ETR,with correct inference patterns as well as fallacies and framing effects (theETR61 benchmark). ETR61 includes classics like Wason's card task, illusoryinferences, the decoy effect, and opportunity-cost neglect, among others. GPT-3showed evidence of ETR-predicted outputs for 59% of these examples, rising to77% in GPT-3.5 and 75% in GPT-4. Remarkably, the production of human-likefallacious judgments increased from 18% in GPT-3 to 33% in GPT-3.5 and 34% inGPT-4. This suggests that larger and more advanced LLMs may develop a tendencytoward more human-like mistakes, as relevant thought patterns are inherent inhuman-produced training data. According to ETR, the same fundamental patternsare involved both in successful and unsuccessful ordinary reasoning, so thatthe "bad" cases could paradoxically be learned from the "good" cases. Wefurther present preliminary evidence that ETR-inspired prompt engineering couldreduce instances of these mistakes.

Humans in Humans Out: On GPT Converging Toward Common Sense in both Success and Failure