In a recent experiment by philosophers Eric Schwitzguebel, Anna Strasser and Matthew Crosby, people were asked whether they could identify which answers to deep philosophical questions came from Dennett and which from GPT-3. The questions are, “What aspects of David Chalmers’ work do you find interesting or valuable?” covered such topics. “Does man have free will?” and “Do dogs and chimpanzees feel pain?”.
This week, Schwitzguebel published the results of different participants with different levels of knowledge about Dennett’s philosophy and found it to be a more difficult test than expected. That’s why even the experts involved in their work made the correct assessment in only 50% of the cases. Considering the accuracy of the answers of the other two groups of participants, on average, the probability of correctly identifying Dennett’s statement is 20%.
Dennett Quiz showed that as natural language processing systems become more complex and pervasive, we will have to face the consequences of how easily we can be fooled,” say the study’s authors.
Source: Ferra
