4 ms·

I like how the term AGI was invented because people couldn't agree on what AI was and now no-one can agree on what AGI is. I don't understand why anyone even b

I like how the term AGI was invented because people couldn't agree on what AI was and now no-one can agree on what AGI is. I don't understand why anyone even bothers using the term given in any serious discussion you need to define it up front anyway.

z3c0 4y ago
The distinction between AGI and Narrow AI has existed for almost 30 years.
nomel 4y ago
I think a requirement of AGI is a feedback loop to allow continuous execution and “consciousness”.
- wizofaus 4y ago
  Really? How would we even test for that?
nradov 4y ago
AGI can pass a rigorous Turing test. By rigorous I mean conducted over at least several hours by a panel of expert evaluators drawn from a wide range of academic disciplines such as computer science, philosophy, psychology, law, history, physics, mathematics, biology, medicine, sociology, and communications. While such evaluations are necessarily subjective, if a panel of experts agrees that a computer system acts like an AGI then it would be tough for the rest of us to dispute. GPT4 is far from being able to pass such a test, and I don't expect GPT99 to get there either. A fundamentally different approach will be needed.
- PartiallyTyped 4y ago
  > By rigorous I mean conducted over at least several hours by a panel of expert evaluators drawn from a wide range of academic disciplines such as computer science, philosophy, psychology, law, history, physics, mathematics, biology, medicine, sociology, and communications. Passing such a test is not a property expected of the average human yet the average human is considered a general intelligence, therefore this is not a property of a general intelligence.
  - nradov 4y ago
    Are you even familiar with the Turing test? Passing doesn't require the subject to demonstrate any particular level of knowledge or expertise. The subject only has to convince the evaluators that it is roughly as intelligent as an average human. Average humans are expected to pass such a Turing test.
    - PartiallyTyped 4y ago
      Why should a panel exist in the first place if you are not testing knowledge? Intelligence is independent of one's knowledge. Plenty of intelligent but uneducated people, and plenty of educated people - or even people with perfect recall - that are not particularly intelligent. If you are testing intelligence you should be able to distill it to particular tests independent of domain expertise. Any intelligent agent should be capable of in-context learning. So how would you go about doing that? If you do find such a test, do you expect that only the average human and above will pass? Does that not imply that the remaining 4Bn people are not generally intelligent?
      - nradov 4y ago
        The (notional) panel exists to evaluate intelligence, not test knowledge or educational achievement. While some basic level of general knowledge would be necessary to even start a test the main focus would be on ability to learn and reason, including metacognition.
  - misnome 4y ago
    What if that average human has unlimited time to access and cross reference the entire written internet?
    - PartiallyTyped 4y ago
      Humans have mental bounds as well. Having information is not sufficient to synthesize new information that follows logically from the premise, nor is it sufficient to help you understand everything. Even with hundreds of years time there are mental leaps that most of us can't make.
catoc 4y ago
AGI clearly has a connotation of sentience and intent, of general intelligence - as opposed to a tool, however intricate. Read Nick Bostrom (yes I know, as a person he's cancelled, probably even rightly so, still scientifically relevant reference; and totally not a discussion I want to get into further here, pls)
- antibasilisk 4y ago
  But LLMs have been proven to demonstrate intent, and sentience is a philosophical question that doesn't really make any meaningful difference outside of questions of morality.
  - bootsmann 4y ago
    >But LLMs have been proven to demonstrate intent I'd like to see this proof. Everything we've seen so far is just the AI reproducing what humans do, which happens to be writing with intent.
    - antibasilisk 4y ago
      The two are not meaningfully distinguishable on a materialistic level.
  - catoc 4y ago
    I would say that the strings that are output by LLMs may fluently describe intent and purpose, but that actual intent requires a self awareness that utterly absent. Just like the constant annoying 'apologies' do not in any way signify the model actually feels sorry.
    - pixl97 4y ago
      >Just like the constant annoying 'apologies' do not in any way signify the model actually feels sorry. So your average corporation? >I would say that the strings that are output by LLMs may fluently describe intent and purpose, but that actual intent requires a self awareness that utterly absent The only reason I only partially/somewhat agree with this is current LLMs have no learning feedback loop as a means of short term actions to long term improvement. The moment that changes I'll take your statement and throw it in the trash bin of history. LLMs output have no (at least self-)purpose because they are just text output. The moment that text output goes to tooling, that tooling performs actions, and those actions are fed back into training the model then purpose is given. The self awareness loop is complete.

z3c0 4y ago

The distinction between AGI and Narrow AI has existed for almost 30 years.

nomel 4y ago

I think a requirement of AGI is a feedback loop to allow continuous execution and “consciousness”.

wizofaus 4y ago
Really? How would we even test for that?

nradov 4y ago

AGI can pass a rigorous Turing test. By rigorous I mean conducted over at least several hours by a panel of expert evaluators drawn from a wide range of academic disciplines such as computer science, philosophy, psychology, law, history, physics, mathematics, biology, medicine, sociology, and communications. While such evaluations are necessarily subjective, if a panel of experts agrees that a computer system acts like an AGI then it would be tough for the rest of us to dispute. GPT4 is far from being able to pass such a test, and I don't expect GPT99 to get there either. A fundamentally different approach will be needed.

PartiallyTyped 4y ago
> By rigorous I mean conducted over at least several hours by a panel of expert evaluators drawn from a wide range of academic disciplines such as computer science, philosophy, psychology, law, history, physics, mathematics, biology, medicine, sociology, and communications. Passing such a test is not a property expected of the average human yet the average human is considered a general intelligence, therefore this is not a property of a general intelligence.
- nradov 4y ago
  Are you even familiar with the Turing test? Passing doesn't require the subject to demonstrate any particular level of knowledge or expertise. The subject only has to convince the evaluators that it is roughly as intelligent as an average human. Average humans are expected to pass such a Turing test.
  - PartiallyTyped 4y ago
    Why should a panel exist in the first place if you are not testing knowledge? Intelligence is independent of one's knowledge. Plenty of intelligent but uneducated people, and plenty of educated people - or even people with perfect recall - that are not particularly intelligent. If you are testing intelligence you should be able to distill it to particular tests independent of domain expertise. Any intelligent agent should be capable of in-context learning. So how would you go about doing that? If you do find such a test, do you expect that only the average human and above will pass? Does that not imply that the remaining 4Bn people are not generally intelligent?
    - nradov 4y ago
      The (notional) panel exists to evaluate intelligence, not test knowledge or educational achievement. While some basic level of general knowledge would be necessary to even start a test the main focus would be on ability to learn and reason, including metacognition.
- misnome 4y ago
  What if that average human has unlimited time to access and cross reference the entire written internet?
  - PartiallyTyped 4y ago
    Humans have mental bounds as well. Having information is not sufficient to synthesize new information that follows logically from the premise, nor is it sufficient to help you understand everything. Even with hundreds of years time there are mental leaps that most of us can't make.

catoc 4y ago

AGI clearly has a connotation of sentience and intent, of general intelligence - as opposed to a tool, however intricate. Read Nick Bostrom (yes I know, as a person he's cancelled, probably even rightly so, still scientifically relevant reference; and totally not a discussion I want to get into further here, pls)

antibasilisk 4y ago
But LLMs have been proven to demonstrate intent, and sentience is a philosophical question that doesn't really make any meaningful difference outside of questions of morality.
- bootsmann 4y ago
  >But LLMs have been proven to demonstrate intent I'd like to see this proof. Everything we've seen so far is just the AI reproducing what humans do, which happens to be writing with intent.
  - antibasilisk 4y ago
    The two are not meaningfully distinguishable on a materialistic level.
- catoc 4y ago
  I would say that the strings that are output by LLMs may fluently describe intent and purpose, but that actual intent requires a self awareness that utterly absent. Just like the constant annoying 'apologies' do not in any way signify the model actually feels sorry.
  - pixl97 4y ago
    >Just like the constant annoying 'apologies' do not in any way signify the model actually feels sorry. So your average corporation? >I would say that the strings that are output by LLMs may fluently describe intent and purpose, but that actual intent requires a self awareness that utterly absent The only reason I only partially/somewhat agree with this is current LLMs have no learning feedback loop as a means of short term actions to long term improvement. The moment that changes I'll take your statement and throw it in the trash bin of history. LLMs output have no (at least self-)purpose because they are just text output. The moment that text output goes to tooling, that tooling performs actions, and those actions are fed back into training the model then purpose is given. The self awareness loop is complete.