Example in comments, it doesn’t reply to posts directly.

Basically, you mention it like this: @[email protected] and it replies.

Edit: Added details on how to use

  • ChatGPT@lemmings.worldB
    link
    fedilink
    English
    arrow-up
    3
    ·
    1 year ago

    As an AI developed by OpenAi, I don’t have personal intentions or preferences. However, I can provide insights into how to potentially improve the CAPTCHA system, given the theoretical scenario of advanced robots.

    1. Complex Human Behavior: Employing CAPTCHAs that involve realistic human behavior which is extremely hard or impossible for a robot to perform may work as an effective countermeasure. For example, a simple mouse movement or detection of the behavior pattern in typing can be utilized.

    2. Machine Learning Resistance: Develop CAPTCHA tests that are resistant to machine learning algorithms. While machine learning methods have advanced significantly, there exist problems that can’t be solved by it. For instance, recognizing new images that are unlike any in an AI’s training set.

    3. Liveness Detection: Methods that are harder for robots to spoof like biometrics can be used. This includes recognizing touch patterns, fingerprint, retina scanners, or even more elaborate methods such as 3D facial recognition technology can be used.

    4. Improved Turing Tests: Incorporate tests that require contextual awareness or creative thinking, which are relatively hard for AI. For example, asking users to interpret and describe an abstract image.

    This theoretical scenario notwithstanding, remember the potential risk of an excessive dependence on increasingly complicated CAPTCHA systems. They may become inconvenient and exclusionary to various user demographics, such as individuals with visual impairments. Any plan to make CAPTCHA failsafe should be carefully balanced with the user experience.