Blog

Contents

    Jul 18

    Why AI Feedback Is Misleading And Why You Still Need Human Input

    The Seductive Trap of AI Validation

    Just like any other day, you wake up and head over to your computer for work on another side project. This project is different. You know it’s brilliant. After all, no one is building an app for docs to order their own dogfood with a tap of their nose based on their taste preferences. It’s only clear to you because you’re so in-tuned with the needs of pets (at least this is what ChatGPT told you). Sure, your last idea didn’t work out, but you have a good feeling about this one.

    Here’s the problem, AI isn’t built to challenge you. It’s specifically built to make you feel warm and fuzzy. Reinforcement learning as well as the initial training on AI makes it tend more toward patting you on the back. In this article we will take a look at why humans are essential in real-world scenarios and why that’s not changing. We will also look at what this means for lead qualification and sales automation through Agents like CloseBot.

    AI Isn’t Giving You Honest Feedback… Here’s Why

    When we break down how AI works, it makes sense why we see some of these undesired behaviors. And it makes it much easier for us to see where humans have a permanent place in the loop of the Human-AI team.

    Generative AI of today operates by looking at a series of previous tokens (like words) and spits out the next best words. This is exactly how early models like Davinci worked upon ChatGPT’s initial release. That model and others have since been deprecated in preference of reinforcement learning (RL) models. These models are optimized around chat and output is ranked by real humans (reinforced). These rankings then nudge the model in future scenarios.

    Because of how generative AI is built, it’s going to try to:

    1. Please you
    2. Do as you ask

    So if you ask for positive and negative feedback on an awful idea, it will come up with positives for you. Similarly, if you ask for positive and negative feedback on a great idea, it will come up with negatives.

    The problem… it listens to your input even when it shouldn’t. It’s easy for users to accidentally feed their own bias and assumptions into the system, causing the AI to lie, misinterpret, and ultimately give answers that aren’t aligned with human desires.

    The Gaslighting Effect — A Generation Trained to Believe They’re Brilliant

    We created a monster whose superpower is creating over-confident individuals. Looking online, everyone appears to be an expert in their field. Not only are they using AI to appear to be an expert… but they actually believe themselves to be an expert because AI is telling them they are.

    The main risk with this new frontier of overconfident influencers is that people are easily convinced to spend money on experts who are anything but. It’s important to look deeper to see if that person has the credentials and facts that back up their product or service. Take CloseBot, for example, which is using real AI in production to book thousands of appointments with qualified leads every day. We have the data to back this up and our team is composed of real AI experts.

    Real Feedback Comes from Real People

    If you’re looking for feedback on an idea, the best place to get feedback is with a human. If you’re looking for ideas, the best place to start is with AI.

    AI is great at taking creative input and giving creative output. Having a hard time coming up with a list of your next potential business ideas? AI is great at this! But once you start to decide which to execute on, loop in real humans to give critical feedback.

    Why is this? People have a diverse background of life-long experiences that leads them to have diverse opinions. AI lacks this and there’s no true way for it to ever gain this. Human bias, assumptions, scars from learned lessons… these all play an integral part in critical feedback. This is why CloseBot is focused on the lead qualification and booking part of the sales process. We believe AI should remove a customer’s friction toward their appointment with a real person. We don’t believe humans should be fully removed from the entire sales process.

    Where to Go from Here

    Pick your sphere of influence wisely. The future generations will easily fall subject to the fake guru epidemic just like how our elders fell victim to the fake news epidemic. You’re the lucky one… you now understand the limits of AI and how to balance AI with human feedback.

    CloseBot is a no-fluff system built by real experts to get you real results. Grab your free account today to try it out for yourself.

    What you should do now

    1. See CloseBot’s powerful AI agents in action.
      Sign up for a free trial.
    2. Read more articles from our blog.
    3. If you know someone who would enjoy this article, share it with them via email, LinkedIn, Twitter/X, or Facebook.