HackerPapers

arXiv:0712.3329cs.AI

Universal Intelligence: A Definition of Machine Intelligence

Shane Legg, Marcus Hutter

Abstract

A fundamental problem in artificial intelligence is that nobody really knows what intelligence is. The problem is especially acute when we need to consider artificial systems which are significantly different to humans. In this paper we approach this problem in the following way: We take a number of well known informal definitions of human intelligence that have been given by experts, and extract their essential features. These are then mathematically formalised to produce a general measure of intelligence for arbitrary machines. We believe that this equation formally captures the concept of machine intelligence in the broadest reasonable sense. We then show how this formal definition is related to the theory of universal optimal learning agents. Finally, we survey the many other tests and definitions of intelligence that have been proposed for machines.

arXiv comments
50 gentle pages
Journal reference
Minds & Machines, 17:4 (2007) pages 391-444
DOI
Subjects
cs.AIArtificial Intelligence

On Hacker News

Linked by 13 identifiable commenters in 14 comments across 13 threads, from Feb 2015 to Mar 2023. Found in 2 HN submissions from Jun 2016 to Oct 2017. Open submission discussions.

Comments linking this paper, per year.

Discussed in comments13 threads

  1. And yet It Understands 1 comment

    • > Maybe it is human like intelligence already. It objectively is not. First, intelligence (human or otherwise) is typically defined as the sum of value functions of a given policy across all states, divided by an exponential of the complexity of the environment[0] (page 23). The…

      espadrineView on HNParent context

  2. John Carmack's new AGI company, Keen Technologies, has raised a $20M round 1 comment

    • > It is a term gaining in popularity due to this difficult to define definition and allows for more a porn like definition > Philosophers, psychologists, neuroscientists, and many others have been contemplating the definition of intelligence for centuries and we're still not…

      sinenomineView on HNParent context

  3. The singularity is close? 2 comments

    • > I will continue dismissing AGI until someone can tell me what GI is Here's a rather famous recommendation for a definition of general intelligence: https://arxiv.org/pdf/0712.3329.pdf Very loosely paraphrasing their paper, general intelligence is the ability to adapt to novel…

      XcelerateView on HNParent context

    • Here's a definition from Shane Legg and Marcus Hutter: Universal Intelligence: A Definition of Machine Intelligence https://arxiv.org/abs/0712.3329

      VladimirGolovinView on HNParent context

  4. Intuition is a “superpower” 1 comment

    • > There is a body of thousands of scientific studies discussing what intelligence means Which don't agree on a single definition. See for example: https://arxiv.org/abs/0712.3329

      WithinReasonView on HNParent context

  5. Humans who are not concentrating are not general intelligences 1 comment

    • >Practical intelligence is clearly multidimensional I wouldn't call that clear at all! Of course no matter how intelligent a person is, there will be environments in which they do poorly. Feynman would do poorly in the environment called, "Everyone find Feynman and beat him up".…

      xamuelView on HNParent context

  6. Geoffrey Hinton and Demis Hassabis: AGI is nowhere close to being a reality 1 comment

    • Shane Legg (co-founder DeepMind) and Marcus Hutter (Schmidhuber pedigree) defined machine intelligence in this canonical paper from 2007: https://arxiv.org/abs/0712.3329 > Universal Intelligence: A Definition of Machine Intelligence > A fundamental problem in artificial…

      995533View on HNParent context

  7. Artificial Intelligence Still Isn’t All That Smart 1 comment

    • Well, this seems to be the definition Deepmind uses: https://arxiv.org/abs/0712.3329 . And there has been work to establish a quantitative measure of the extent to which an agents behavior is intelligent or mechanical: https://arxiv.org/abs/1805.12387 .

      goolulusaursView on HNParent context

  8. Why humans need to rethink their place in the animal kingdom 1 comment

    • > the assumption that humans have it [general intelligence] is a conceit without convincing evidence Why do you make this claim? Engineering, science, and mathematics are all convincing evidence that humans have general intelligence. Humans also score highly in measures of…

      cgmgView on HNParent context

  9. Where will artificial general intelligence come from? 1 comment

    • Hutter, Schmidhuber, et al. have done some very interesting work on this: https://arxiv.org/abs/0712.3329

      cgmgView on HNParent context

  10. AlphaGo Is Not AI 1 comment

    • One of the co-founders of DeepMind (Shane Legg) has provided the following definition of intelligence: "Intelligence measures an agent’s ability to achieve goals in a wide range of environments" in [1]. This definition has been pretty influential, and in the sense of this…

      mafribeView on HNParent context

  11. Will This “Neural Lace” Brain Implant Help Us Compete with AI? 1 comment

    • > consciousness and insight Being unmeasurable and undefined, and therefore irrelevant to measures of intelligence, "consciousness" is typically not considered part of AGI [1]. Not sure what you mean by "insight". [1] http://arxiv.org/abs/0712.3329

      tfggView on HNParent context

  12. Human collective intelligence as distributed Bayesian inference 1 comment

    • Cool paper! learned a lot about shared beliefs. But I was asking myself "why is 50k traders using similar trading patterns an example of collective intelligence"? I expected an example of a collective accomplishing a group goal as opposed to optimizing individual ability to…

      jonmc12View on HNParent context

  13. Human-level control through deep reinforcement learning 1 comment

    • The No Free Lunch Theorems don't apply in reality anyway, as explained at https://en.wikipedia.org/wiki/No_free_lunch_in_search_and_op... Essentially: NFL says the performance (1 / number of trials are required to find a solution), when averaged over all fitness functions, will…

      chriswarboView on HNParent context

Submitted to Hacker News

Connected through shared HN threads, not inferred similarity or a recommendation.