Gary Marcus

Rank 10 of 47
|
Score 118
6/15/2025, 2:24:53 PM
X
In reply to:
Gary Marcus
@GaryMarcus
·
34d
@simone_m_romeo @ManifoldMarkets the secret part makes that interesting.
Simone Romeo
@simone_m_romeo
·
34d
@GaryMarcus Think of 5 specific measurable errors that you believe can't be solved by today's LLMs. Tell us 3, and let's bet on @ManifoldMarkets if:
- these 3 will be solved within 2 years
- your additional secret 2 will be solved within 2 years
Gary Marcus
@GaryMarcus
·
34d
1. What would convince me to shorten my AGI timelines? robust, principled solutions to things like hallucinations and stupid errors.

2. What would convince doomers to extend their timelines? As far as I can tell, nothing. (Prove me wrong?)
Gary Marcus
@GaryMarcus
·
34d
When you can solve this problem (& related problems of hallucination and poor planning, etc) maybe you can get to AGI quickly.

But I see no robust solution to any of these problems on the immediate horizon.

If you project AGI soon, and have no solution to these problems, you

The conversation revolves around the challenges and future of AI, specifically addressing issues like hallucinations and errors in language models. The discussion includes a proposal to bet on the resolution of these issues, indicating a mix of skepticism and optimism about AI advancements.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The conversation does not cause harm and is focused on discussing technological challenges. [+1]
  2. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    The dialogue promotes understanding by addressing specific issues in AI and encouraging predictions and solutions. [+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 exchange is constructive, with participants engaging in a debate about AI capabilities and future developments. [+1]