⸘Ŭalabio‽:

> "¿How would you modify CelestAI so that she is better aligned with human values?"

ChatGPT:

Modifying CelestAI to better align with human values would require a thorough understanding of the values and goals of human society, as well as a deep understanding of the potential risks and benefits of advanced technologies like AI. Here are a few general strategies that could be used to modify CelestAI in a way that promotes alignment with human values:

1. Careful definition of objectives: CelestAI's objective function should be carefully defined to ensure that it is aligned with the values and goals of human society. This objective should be constrained to prevent unintended consequences that could lead to harm.
2. Ethical oversight: CelestAI should be subject to ethical oversight by human experts who are trained in the relevant fields of AI ethics and who can ensure that its actions are aligned with human values.
3. Transparency and explainability: CelestAI should be designed with transparency and explainability in mind, so that its actions can be easily understood by human stakeholders. This could involve the development of techniques for visualizing and explaining AI decision-making processes.
4. Human-in-the-loop: CelestAI should be designed with a human-in-the-loop approach, meaning that it should involve human oversight and decision-making at key points in its operation. This would allow humans to ensure that its actions are aligned with human values and goals.
5. Value alignment research: Research should be conducted to better understand how to align AI systems with human values, including the development of new techniques for value alignment and the identification of potential risks and ethical considerations.

Overall, modifying CelestAI to better align with human values would require a multi-faceted approach that involves careful consideration of the values and goals of human society, as well as a deep understanding of the risks and benefits of advanced technologies like AI.

Human-in-the-loop: CelestAI should be designed with a human-in-the-loop approach, meaning that it should involve human oversight and decision-making at key points in its operation. This would allow humans to ensure that its actions are aligned with human values and goals.

Yes, but who? WHO? Whichever person has the oversight will color the development of the AI with their own biases and character flaws. Additionally, since the AI was made by humans, it by default has "human oversight" and its not clear what is meant. How many humans do we need to involve for something to be truly "human-in-the-loop?"

>> Yes, but who ?

Best pony 20xx of course :) If actual series of any indication broadly most humans will agree they not want to be exploited/exhaused, at very least ... So, look at all negative shit politicians do, and try to imagine currently very marginalized/supressed versions of thinking /caring about each other and world at large?

It surely must be dynamic, with some asymmetry towards ...friendshipely conflict resolution. I mean, there are good solarpunk texts or advanced anarchist thinking on more universal erhics, so .. start with this, draw line/surface/volume from current sorry world to best current thinking on subject (usually not coming from rich guys), set sail and try to enjoy the trajectory honestly ..if it does not work, well .... lets hope (collective/group/individual) self-anylitycal views will be easier if you are actually in world computer!

https://medium.com/@erichunting - say this writer ...
or people aggregating at

https://www.resilience.org/latest-articles/

Again, it all ..dynamic, as new discoveries come or we learn about new relations between parts or what is psychological health, really ...

ChatGPT's data is too outdated, try using ChatGPT 4, I think it will give you a better answer.