GPT-4 Explains — “From Unitary Models to Game Theory: The Evolution of Economic Theories of Marriage and Their Impact on Understanding Domestic Violence”
GPT-4 Explains — “From Unitary Models to Game Theory: The Evolution of Economic Theories of Marriage and Their Impact on Understanding Domestic Violence”
causalinf.substack.com
Tracing the Theoretical Lineage of the Outside Option in Modern Empirical Work on Family Violence by Cunningham (r) and GPT-4 (r)
GPT seems quite happy to regurgitate the views of others and, I am afraid, would uncritically do the same with ‘fake news’. I mean, we know that. GPT doesn’t claim to critical thinking abilities nor to develop original ideas.
This essay reads as plausible to the non-expert as any other, despite even after what appear to be several iterations by Scott to improve its work still containing numerous errors that Scott pointed out in his footnotes.
It takes the competence of a learned expert to see through GPT’ display of overconfidence. As much as it takes the leaned expert to prompt and direct GPT in a meaningful way.
Problem is that one becomes an expert by researching and writing essays like this. If nobody does that anymore (as GPT takes over), nobody will have the skill any longer to critique GPT’s output.
"The 1980s saw the rise of Ariel Rubinstein, who significantly extended the reach of bargaining theory. He expanded Nash's cooperative bargaining model to non-cooperative scenarios, providing a deep insight into how bargaining outcomes could be influenced by factors like the order of moves, available outside options, and patience or impatience of the negotiating parties.2"
GPT seems quite happy to regurgitate the views of others and, I am afraid, would uncritically do the same with ‘fake news’. I mean, we know that. GPT doesn’t claim to critical thinking abilities nor to develop original ideas.
This essay reads as plausible to the non-expert as any other, despite even after what appear to be several iterations by Scott to improve its work still containing numerous errors that Scott pointed out in his footnotes.
It takes the competence of a learned expert to see through GPT’ display of overconfidence. As much as it takes the leaned expert to prompt and direct GPT in a meaningful way.
Problem is that one becomes an expert by researching and writing essays like this. If nobody does that anymore (as GPT takes over), nobody will have the skill any longer to critique GPT’s output.
"The 1980s saw the rise of Ariel Rubinstein, who significantly extended the reach of bargaining theory. He expanded Nash's cooperative bargaining model to non-cooperative scenarios, providing a deep insight into how bargaining outcomes could be influenced by factors like the order of moves, available outside options, and patience or impatience of the negotiating parties.2"
The Nash program goes actually back to Nash himself: https://link.springer.com/article/10.1007/s13209-020-00221-5
Another thought: How long before a large portion of what GPT is trained on is actually its own output?