top of page

AI safety and the four laws of the avatars

Updated: Nov 24, 2023

At Version of me, we are convinced of the potential for good of both AI and the Metaverse. While we acknowledge the potential dangers of both technologies, we believe there is a lot of hype around the existential threat of AI for humanity, and we propose that the details of implementation are key to the result: we will get it right if we do it right.


Several of the contacts we have discussed our technology with, have expressed some sort of concern over the danger that artificially intelligent avatars substitute their real models, so we have given long thought to the matter, and established a set of rules Version of me shall follow to ensure the highest standards of quality and safety for our AI.


To begin with, we are establishing the following committments in the development of our AI technology, which we call "the four laws of avatars", reminiscing to Asimov´s three laws of robotics:


  1. Identification: An avatar must identify itself as such and be unmistakable from a real person, while connected to the real person. The real person must be able to take control, and in such situation, be clearly identified as the real person

  2. Artificiality: An avatar cannot be programmed to establish emotional relationships with people on purpose. While such effects may happen, behaviour of this kind will be discouraged by the company, and frequent and unmistakably clear reminders of the artificial nature of the avatar shall be shared with users

  3. Non-initiative: Avatars cannot take the initiative in communicating with people. There will never be a fully independent avatar created by version of me. Avatars will not have motivations

  4. Non-escalation: Avatars must always be courteous and ignore, deflect and report harassment from humans. Avatars must be able to focus on their domain of conversation and manage deviations properly





And finally, we have created Version of me´s Terms and Conditions, which are meant to defend the rights of users, particularly minors, and to defend the platform from misuse.




These principles above are the foundation of safety and quality for Version of me´s AI, and we shall stick to them as guarantee for our customers.


Thank you for your attention, inputs are welcome in the comments,


Ignacio García-Carrillo, founder



5 views0 comments

Recent Posts

See All

Comments


Original on transparent_edited.png
bottom of page