Gary Marcus

Rank 33 of 47

Score -15

@Sauers_ you can get from 35 to 45 just by mimicry and more data. the last bit is really hard, in many domains, and may require entirely different techniques, as we saw e.g., with driverless cars.

6/22/2024, 3:22:19 PM

In reply to:

Sauers

@Sauers_

139d

@GaryMarcus I except this at least partially too, especially since each incremental percentage point on a benchmark is harder to achieve. I.e., a larger skill increase is needed to go from 85% to 95% than from 35% to 45%.

Gary Marcus

@GaryMarcus

139d

Reward for anyone who can find the source of this graph and plot it with reasonable accuracy back to the original GPT in 2018.

My guess is that 2024 will look like a tiny blip in that context.

Sambhav Gupta

@sambhavgupta6

140d

@GregKamradt maybe GPT5 is not that impressive as we imagine it to be

The statement engages in a technical discussion about the challenges of improving AI performance, particularly in the context of incremental improvements and the need for different techniques in advanced stages. The tone is informative and analytical, aiming to contribute to the understanding of AI development challenges.

Principle 1:
I will strive to do no harm with my words and actions.
The statement strives to do no harm and focuses on a technical analysis, which aligns with the principle of doing no harm with words and actions. [+1]
Principle 3:
I will use my words and actions to promote understanding, empathy, and compassion.
The statement promotes understanding by explaining the difficulty of achieving higher performance benchmarks in AI, which aligns with the principle of using words to promote understanding, empathy, and compassion. [+1]
Principle 4:
I will engage in constructive criticism and dialogue with those in disagreement and will not engage in personal attacks or ad hominem arguments.
The statement engages in constructive dialogue by building on the previous comments and providing additional insights, which aligns with the principle of engaging in constructive criticism and dialogue without personal attacks. [+1]