53
orhun
4y

Unpopular opinion: AI is overrated.

Comments
  • 19
    Correct opinion.
  • 11
    AI does not exists, so - it is an incorrectly used term.
  • 3
    @magicMirror You know what I meant. I'm talking about the half-experimental half-spaghetti things that 'data scientists/developers' do and we call it AI.

    Sure, the real one doesn't exist yet.
  • 0
    @PublicByte that's a pretty good story actually. Not that I'm against "AI" when needed or something but it's... overrated.
  • 3
    @orhun AI using blockchain on the cloud?
    sure. I will be right there.
  • 6
    The term has unfortunately been wrongfully used by the marketing guys since forever. The same can be said about pretty much anything they're saying (cf. "Smart" objects)
  • 0
  • 0
    @PublicByte more like artificial common-sense
  • 1
    Over-hyped for sure. It's fun though
  • 0
    Define AI.

    Is it a machine that performs a compute task?

    Is it a machine the mimics human behavior?

    Is it a machine at all?

    Is it a conscious machine?

    Is it a conscious machine with self awareness?

    Is it a machine that can make a decision?

    Is it a machine that can make a decision about life and death?

    It is a machine that maps data and acts like a filter?
  • 1
    @magicMirror AI exists. Take the deep fakes algorithms for example. You have facial recognition software trained on real faces and this is the system you need to trick to pass the test. You have a face generator that is learning to trick the facial recognizer. There is no human intervention or adjustment of the algorithms needed. You can accomplish a task that is labour impossible for humans. Is it “intelligent”? No. Is it a self training algorithm that can assemble itself in a way that would be time prohibitive for a human to do? Yes.

    Building its own algorithm is self-automation. That learning process is artificially intelligent.
  • 0
    @Demolishun Nowadays theres no possible misunderstanding, he's obviously talking about neural networks and whatever floats around them
  • 1
    @orhun Amen bro! I work with AI on daily basis, what you say is so fucking right
  • 1
    @irene
    in your own words: "is it inteligent?"
    Inteligence is defined using a simple measure: true self awareness. This is the only way to pass The Turing Test.

    You are talking about ML. ML is a way to take a large amount of known observed results, along with what the human assumes is having an actual effect on the results, and use an algorithem to automatically constracting a new non trivial algorithem to predict future results, based on new obeservations.

    Take the result algorithem and try to use it on anything else, will result in total failure. An ML classifier that can identify a cat in a picture cannot identify a dog. Or a chair. It is able to do cat/not a cat.
    Do not mistake a classifier that can identify both - it is a meta classfier that uses a dog classifier, and a cat classifier, and decides which is more correct.
  • 3
    @magicMirror Exactly a lot of "AI" is just a fancy filter. It is really really cool, but it still just maps things.
  • 0
    πŸ“πŸ“ŒπŸ§·
  • 0
    Machine learning is super overrated. Algorithms like monto Carlo https://en.m.wikipedia.org/wiki/... are super awesome and still artificial intelligence
  • 1
    @magicMirror Seems philosophical. If artificial intelligence is about perceived intellect does it matter how it arrives at the seemingly intelligent solution? I’m not sure there is a strict definition of AI. Maybe there is?
  • 0
    A definition of AI: Javascript developers putting Java on their resume.
  • 0
    AI thinks you are overrated
  • 1
    @irene It is called the Turing test.
  • 0
  • 0
    VR/AR is still plausible. Only companies use VR/AR but not customers/consumers
  • 0
    @irene but is it really though, isn't it more that humans are defining a sandbox of parameter spaces for a human created algorithm to play around in it. what you are suggesting is that the system is capable of adjusting its internal structure / logic, by adjusting to its environment.

    For the aspect of time intensive for human labour in context of face recognition. I suggest you look into Eigenfaces which is nothing more than what the network is doing "automatically".

    It is great that "AI" has regained in popularity but please don't try to solve everything just by throwing CPU/GPU power at the problem. The electricity invested in those endeavors is most of the time not worth the outcome. Use your personal BPU first😁
  • 0
    @Yggdrasil What we need is an AI that cleans code and detects inefficiencies trained on good coding patterns. Then it can clean all of the AI code and the problem goes away. 🀣
  • 3
    AI is Machine Learning. Machine Learning is Statistics. Statistics is not new.

    AI is not new
  • 0
    A delusion
Add Comment