Exploring Competition for GPT-3 in the AI Landscape

Have you dug deeper into the claim of GPT-3 has some competition Sean G.?

· Sorted by Oldest

Mike Tung
·
They are only comparable in the sense that they are two recently released Natural language processing APIs. But it is a bit like saying a car and a boat compete as modes of transport. Diffbot's API doesn't generate language, it returns a knowledge graph based on what it understood in the input text.
Sean G.
·
Like Mike said, there is a difference between the two in that there is no NLG (at least as stated in the article) but the inclusion of a KG shows how it reaches beyond the capabilities of GPT-3. While GPT-3 is impressive, I feel there is a lot of hype and that it is really just regurgitates the massive information that has been fed into it without having an understanding of it. This quote from the article stood out to me:
Language models like GPT-3 are amazing mimics, but they have little sense of what they're actually saying. "They're really good at generating stories about unicorns," says Mike Tung, CEO of Stanford startup Diffbot. "But they're not trained to be factual."

Mike Tung
·
They are only comparable in the sense that they are two recently released Natural language processing APIs. But it is a bit like saying a car and a boat compete as modes of transport. Diffbot's API doesn't generate language, it returns a knowledge graph based on what it understood in the input text.
Sean G.
·
Like Mike said, there is a difference between the two in that there is no NLG (at least as stated in the article) but the inclusion of a KG shows how it reaches beyond the capabilities of GPT-3. While GPT-3 is impressive, I feel there is a lot of hype and that it is really just regurgitates the massive information that has been fed into it without having an understanding of it. This quote from the article stood out to me:
Language models like GPT-3 are amazing mimics, but they have little sense of what they're actually saying. "They're really good at generating stories about unicorns," says Mike Tung, CEO of Stanford startup Diffbot. "But they're not trained to be factual."