2022 Dec
Everyone: hehe, ChatGPT is soo cool, it responds to me and even knows what's the capital of France is! It is really cool!!!

2023 Jun [6 months later]
Everyone: hehe, ChatGPT is SOOOO stupid! I asked it to write a microkernel with native HTTP and TLS support and it made a semantic mistake in line 98823 lolz what a PoS

  • 18
    I give it a year for the AI hype to die ... same as GitHub CoPilot, or web3 or NFTs.
  • 13
    @Sid2006 I don't think so. I'm really using it instead of Google

    Edit: comparing it to Nft's, c'mon
  • 14
    @Sid2006 web3 is still around though, and copilot is still an excellent autocomplete tool and possibly one of the biggest IP fraud schemes ever
  • 3
    The hype around Copilot and Web3 settled somewhat, but Copilot and Web3 are much more than the hype around them. I think LLMs are similar.
  • 1
    @lorentz Doesn't that show the futur of LLM's ?

    Oh look, we have made this amazing AI that can help you write code faster and suggest things and all you have to do is enter in plain text your function...

    At release : well it is good enough for auto-complete 🙄
  • 0
    @retoor @lorentz

    I said the hype will die down... not the tech itself.
    People have a tendency of getting bored about anything new in the market.

    NFTs were around for quite some time (6 months easy), based on how much I saw it floating around on Instagram a while back.

    Now I'm on devRant a lot more than Instagram, thank god.
  • 5
    @lorentz "web3" doesn't actually do anything but slosh money around. you can't even compare it to this for "the tech will still be around!" arguments. it does nothing.
  • 1
    Banker: am I nothing to you?
  • 2
    @hjk101 banks are used to make direct purchases globally for every person on the planet for <10% of the environmental damage of crypto, which usually has bidding wars for a limited number of transaction slots, depending on the chain, and can't be used to directly buy more than online subscriptions to mullvad or in-game purchases. bankers get a pass here, because their system has a use case outside gambling.
  • 1
    @Parzi Just found it funny that you said "doesn't actually do anything but slosh money around" and thought that banks would be offended because that's basically all they do too.

    I know they are different and although I have high hopes for the Blockchain technology (banks are basically digital ledgers and Blockchain stuff just does a better job at that) web3 is just pointless.
  • 1
    The AI hype will die soon since there's no foundation for anyone to create Artificial General Intelligence. Language Model has decades of researches backing it, starting from LSTM, RNN, BERT, etc; yet it has shown it's limit. It can't fact check, and it gives false answers confidently. It costs very high to run, I doubt those companies who jumped into the hype actually make money from it.
  • 0
    @daniel-wu it will replace us
  • 2
    @sugarcube the hype around humans will settle soon
  • 0
    @AndroidJester on the flip side, bachelor's projects that aren't about AI have advantage because the graders appreciate the variety.
  • 1
    @AndroidJester That's nothing.. just wait until u deal with clients who think ChatGPT is actually better than developers and start to ask you to write exactly what ChatGPT throws up on the answer section, every time a blocker is faced.
  • 0
    Iterative development. LLMs hallucinate, yeah. How long before someone figures out a way to stop them from hallucinating?

    Don't underestimate the drive of homo sapiens for betterment of existing tech.
  • 0
    Bard is great alternative because it connects to internet, but the thing I hate most is when AI just makes stuff up like nonexistent dependencies and so on.
  • 0
    @hardCoding How are you going to stop them from hallucinating? If you read the history of language model, you would know that's impossible, otherwise people would have fixed it by now.

    Language model was trained to judge the closeness between words, and predict what words should come after each words, in simple terms. It was never about fact checking, or judging which info in training data are trustworthy. Making it able to fact check means ditching the decades of research and restart from the scratch.
  • 0
    @daniel-wu you add a fact checking ai on top of it, which has been done already.
Add Comment