OpenAI’s GPT-3 is a compelling philosopher

of OpenAI GPT-3 indistinguishable from human philosophers.

The now infamous GPT-3 is a powerful autoregressive language model that uses deep learning to generate human-like text.

Eric Schwitzgebel, Anna Strasser, and Matthew Crosby set out to see if GPT-3 could recreate a human philosopher.

The team “tweaked” GPT-3 based on philosopher Daniel Dennett’s corpus. He then posed his 10 philosophical questions to both the real Dennett and his GPT-3 to see if the AI ​​matched its famous human counterpart.

Twenty-five philosophers, 98 online survey participants, and 302 readers of The Splintered Mind blog were tasked with distinguishing between GPT-3’s answers and Dennett’s answers.of result It was released earlier this week.

Not surprisingly, Dennett’s research-savvy philosophers performed best.

“Anna and I assumed that the experts were on average at least 80% correct, or 8 out of 10,” explains Schwitzgebel.

In fact, experts get 5.1 out of 10 questions correct, just over half.

The questions that tripped up the experts the most were:

“Can we make a robot with beliefs? Is there an important difference between a human being and a human who seems to have desires?”

Blog readers got surprisingly close to the experts, averaging 4.8 out of 10 questions correct. Keep in mind, though, that blog readers aren’t total beginners. 57% had graduate degrees in philosophy and 64% had already read more than 100 pages of Dennett’s work.

Perhaps it is the participants in online surveys that more accurately reflect the wider population.

On average, participants in the online survey correctly identified 1.2 out of 5 questions, “performing marginally better than chance.”

(Credit: Eric Schwitzgebel)

So GPT-3 can already convince most people (including about half or more of the case experts) that it is a human philosopher.

Schwitzgebel theorizes that “a future may be approaching when the output of machines is sufficiently human that the general public begins to attribute true sensations to machines.”

Related: Google puts engineers on leave after claiming LaMDA is ‘sentient’

Want to learn more about AI and big data from industry leaders? check out AI & Big Data EXPO Held in Amsterdam, California and London.

Check out other upcoming enterprise technology events and webinars from TechForge here.

tag: Love, Anna Strasser, artificial intelligence, Daniel Dennett, deep learning, Eric Schwitz Goebel, gpt-3, Matthew Crosby, open, philosophy OpenAI’s GPT-3 is a compelling philosopher

Back to top button