Share.

    16 Kommentare

    1. I mean it makes sense. The APIs on these things are a house of cards – just layers and layers of natural language instructions. Context on context on context. At some point these limitations can’t be optimised anymore.

    2. M0therN4ture on

      Because language models are a dead end. Consciousness or AGI isn’t reached with LLMs but LNMs.

    3. SomeSamples on

      LLM’s are a dead end but Image or data analysis is a real thing. AI does a much better and faster job than humans do with it. But those types of AI are not LLM’s.

    4. GreenTrees797 on

      LLMs? The thing that has been integrated in almost every type of software we use since only two years ago, is a dead end? This isn’t 3D TVs. It’s the new normal. It’s the most basic expectation now. People barely even search through the results on Google anymore. They just read the AI overview. A summary of what you used to click through links to get manually just a few years ago. No one is giving that up, it’s already taken for granted. 

    5. Historical_Emu_3032 on

      What I find absolutely amazing about LLM use, is so much of their use is an absurd amount of implementations that do something a computer was already very good at doing but reimagined with 1000x the compute cost.

      including the idea that there are cost saving to replace humans, which it may be true in some cases up until the true cost of compute is passed back to the customer.

    6. LLM will be entering its entering its efficiencies phase in the next 2 years. GAI is decades away when quantum computing is pervasive

    7. Chill_Panda on

      This doesn’t mean AI is a dead end, nor that LLM’s are a dead end (though I do think they are reaching their current limit).

      This is an AI researcher confirming that LLM’s are a dead end for AI. Which, like yeah… we know. They’re a tool, a smart tool maybe but they’re like any other software, something you have to use with purpose. Not this magic fix all

    Leave A Reply