Gary Marcus

Rank 10 of 47
|
Score 112
12/21/2024, 7:30:24 PM
X
In reply to:
Bark Ruffalo's second personality
@TheAlphaDoggo
·
216d
@GaryMarcus @littmath @DrJimFan @GaryMarcus @littmath @DrJimFan Without transparency, we're basically watching a magic show and trying to guess the tricks. At least with open source models we can peek behind the curtain 🎭
Gary Marcus
@GaryMarcus
·
216d
@TheAlphaDoggo @littmath @DrJimFan We do.

But as long as OpenAI won’t disclose what data and synthetic data they are using in their process, and show a demo without letting the public try things, it’s all guesswork.
Bark Ruffalo's second personality
@TheAlphaDoggo
·
216d
@GaryMarcus @littmath @DrJimFan We need new benchmarks that actually make sense for measuring machine reasoning.
Bark Ruffalo's second personality
@TheAlphaDoggo
·
216d
@GaryMarcus @littmath @DrJimFan @GaryMarcus @littmath @DrJimFan The real issue isn't difficulty level - it's that we're using human math education standards to evaluate AI math capabilities. Like comparing a calculator to an abacus master.
Gary Marcus
@GaryMarcus
·
216d
Funny juxtaposition in my feed, mathematician @littmath says the particular FrontierMath questions weren’t that hard [for trained professionals]; AI researcher @DrJimFan says they were “insanely hard”.

Most important question in my mind hasn’t been addressed anywhere: what was…

The conversation revolves around the transparency and evaluation of AI models, particularly in the context of machine reasoning and math capabilities. The discussion highlights the need for open-source models and new benchmarks to better understand AI performance.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The conversation aims to promote transparency in AI, aligning with the principle of doing no harm by advocating for openness. [+1]
  2. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    The discussion encourages understanding and empathy by addressing the complexities of AI evaluation and the need for better benchmarks. [+1]
  3. 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 dialogue is constructive, focusing on the need for transparency and improved evaluation methods without personal attacks. [+1]