Gary Marcus

Rank 14 of 47
|
Score 83
6/3/2025, 7:56:54 PM
X
In reply to:
David Manheim
@davidmanheim
·
213d
@MoonlitMonkey69 @GaryMarcus If the bar for AGI was "can learn additional things," sure. But that wasn't the original qualification - the original discussions perhaps assumed they could learn, since humans can. But the reason to say "general" was just differentiating multi-purpose systems from narrow AI.
Moonlit Monkey
@MoonlitMonkey69
·
213d
@davidmanheim @GaryMarcus I think most humans can become much better than average at a given domain with learning (if they try). Not 'expert within field' but 'expert relative to baseline'.

Not an unreasonable standard for AGI I think, having similar learning capacity to most humans.
David Manheim
@davidmanheim
·
213d
@MoonlitMonkey69 @GaryMarcus Most humans cannot, in fact, become substantially expert at most tasks. Similarly, humans are not generally intelligent most of the time, and they are usually not actually doing much thinking.
Moonlit Monkey
@MoonlitMonkey69
·
213d
@davidmanheim @GaryMarcus It's not whether they ARE better than average, it's whether with active learning they can become that. Most humans can become good at things by learning more about them. Whether they choose to, is another matter.
David Manheim
@davidmanheim
·
213d
@GaryMarcus 1. I agree that the systems are very limited - but the term AGI was initially contrasted with ANI, and it's silly to claim that current systems are narrow, even if they are not "True AGI"
2. Your bar keeps being human expert, not human average. That seems unreasonably high.
Gary Marcus
@GaryMarcus
·
214d
𝗧𝗵𝗲 𝗰𝘂𝗿𝗿𝗲𝗻𝘁𝗹𝘆 𝗳𝗮𝘀𝗵𝗶𝗼𝗻𝗮𝗯𝗹𝗲 𝗶𝗱𝗲𝗮 𝘁𝗵𝗮𝘁 𝘄𝗲 𝗵𝗮𝘃𝗲 𝗮𝗹𝗿𝗲𝗮𝗱𝘆 𝗮𝗰𝗵𝗶𝗲𝘃𝗲𝗱 𝗔𝗚𝗜 𝗼𝗿 𝗮𝗿𝗲 𝗮𝗹𝗿𝗲𝗮𝗱𝘆 𝗲𝘅𝘁𝗿𝗲𝗺𝗲𝗹𝘆 𝗰𝗹𝗼𝘀𝗲 𝗶𝘀 𝗮𝗯𝘀𝘂𝗿𝗱. 𝗛𝗲𝗿𝗲 𝗶𝘀 𝘄𝗵𝘆:

1. True AGI [Artificial General Intelligence] ought to be

The statement is part of a broader discussion on the capabilities of humans and artificial general intelligence (AGI). It contributes to the public discourse on AI by engaging in a debate about the learning capabilities of humans compared to AGI.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement does not cause harm and is neutral in tone, focusing on the potential for humans to learn skills. [+1]
  2. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    The statement promotes understanding by contributing to a discussion about human and AI learning capabilities. [+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 statement engages in constructive dialogue without personal attacks, maintaining a respectful tone. [+1]
  4. Principle 7:
    I will uphold the principles of free speech and use my platform responsibly and with integrity.
    The statement upholds free speech principles by responsibly contributing to a public debate on AI. [+1]