Key Points:
- Gary Marcus critiques GPT 4.5, describing it as mostly insignificant or a “nothing burger.”
- He suggests that the advancements in GPT 4.5 do not meaningfully progress beyond previous iterations.
- The anticipation for GPT 5 is characterized as unrealistic or “still a fantasy.”
- Marcus asserts that scaling data and compute does not equate to guaranteed improvements, challenging the underlying assumptions about AI development.
References:
- The commentary can be found in Marcus’s article on Substack.
Executive Summary:
In his article, Gary Marcus expresses skepticism about the advancements in GPT 4.5, labeling it as largely inconsequential and a precursor to a still-distant GPT 5. He argues that merely increasing data and computational power does not inherently lead to breakthroughs in AI, calling into question commonly held beliefs in the field.
12ft.io Link: https://12ft.io/https://open.substack.com/pub/garymarcus/p/hot-take-gpt-45-is-a-nothing-burger?r=2fj9s&utm_medium=ios
Archive.org Link: https://web.archive.org/web/https://open.substack.com/pub/garymarcus/p/hot-take-gpt-45-is-a-nothing-burger?r=2fj9s&utm_medium=ios
Original Link: https://open.substack.com/pub/garymarcus/p/hot-take-gpt-45-is-a-nothing-burger?r=2fj9s&utm_medium=ios
User Message: Hot take: GPT 4.5 is a nothing burger - by Gary Marcus
for more on see the post on bypassing methods