• Subscribe
  • Do we need to chat with everything?

    Mohammed Kheezar Hayat
    8 replies
    Loads of AI tools are gravitating towards chat as the main interface (chat with your files, chat with your knowledge base, chat with your emails etc etc). Is it just because chat is easier to build? Or maybe because this is where expectations are settled for the moment (possibly because of the awareness that ChatGPT generated)? Thoughts?

    Replies

    Elaine Lu
    While chatting with everything might seem overwhelming, it can enhance user experience by providing instant, personalized interactions. However, it's crucial to balance automation with human touch to maintain authenticity and trust.
    Share
    André J
    LLM is non deterministic. The output is not precise or stable. By chatting the user can narrow down that deterministic output they seek. If you ask an LLM to give you some code on some problem you have. It will always give slightly different answers. That's my observation. When AI becomes precise, then things get interesting.
    Share
    Sree
    @sentry_co Technically speaking, this is the reason why most AI-based stuff is chat interfaced but that doesn't mean a chat interface is ideal for every tool. Everyone is baking a chat-based AI feature into their app when in reality, it's a really poor decision that just degrades the user experience.
    André J
    @sreenington I think that's the easy thing todo right now. Other ways of doing isn't cost effective in the short term, or it would be done. Like everyone is trying to do agents workflows, no one has got it actually working even tho their marketing says it works 😅. It can be done but it's too costly, so everyone is banking on just saying they have it, and then get 5.0 comes out and they can actually have it. Maybe its the same with chat interface, that they just want a wedge in the market, and then when APIs that support other things becomes available, people migrate towards that.
    Chat interfaces are indeed becoming more common in AI tools, but it's not just because they're easier to build. They provide a familiar and user-friendly way to interact with technology. At Contentify, we understand the importance of user experience and offer a comprehensive AI-powered marketing automation solution to enhance your online presence. Problem: Generating engaging social media content consistently Solution: Contentify's AI agents can create visuals, captions, and even optimize SEO for your social media posts, saving you time and ensuring a consistent brand presence. Try Contentify AI for Free here just look up Contentify AI
    Share
    Matt Bonacini
    imo no, if prompts are tuned correctly, my experience is that the outcome will be incredibly similar and "reliable" - with this I don't mean it will be 100% correct, I mean that if you give the same (or similar) inputs, different outputs will be often incredibly similar between each other. Long story short I had to run different versions of the same prompt literally more than 500 times during the past month. The goal with what I'm building is that AI will work with the user uploading some content, and they just need to press a button. There's no chat. tbh at some point there will not even be a "start" button, just an "upload" field and the app will trigger AI at specific time frames based on the uploaded content Here's I'm talking about an app that generates reports and a lot of text. The thing is also to understand when the typical code solution ends and where AI starts. You can see tons of AI wrappers in the market nowadays. But it's necessary to be a little bit more creative and really use AI where you need it instead of slapping a chat everywhere :D EDIT: TL;DR - at some point implementing a chat when it's not necessary is just an anti pattern
    Gurkaran Singh
    Do we really need to chat with everything? Maybe AI just wants us to practice our digital small talk skills—it's like networking for our data!
    Mileva
    I think the conversation is very much important factor of the life but sometime it also depends upon content and purpose