Sometimes when using Claude I get a question asking if Claude is doing good, fine or bad. I guess good is better than fine, but I don’t give it much thought beyond that.
Today though I got asked to be interviewed by AI on how AI is doing. The interview flowed really well and it reminded me that AI is now pretty good at interacting as a person might.
I gave it some great thoughts and said it could publish the interview for the public without identifying me. But then it asked me to read back through the interview and confirm it was all correct and good. I stopped to think about the question and the subtle shift in accountability. Even though I had already endorsed the interview I now felt more accounable.
I didn’t really see the benefit of taking on more accountability so I said it could not publish the interview after all.
Its funny what a simple follow-up question can do for both the conversation and the outcome.
Part 2 – What do I want and not work from AI?
Of course there was nothing too secret in the content and no real issue if it was misinterpreted. Just the usual – a white list of people to not exterminate and a blacklist of people to hack and randomly change things in their lives to sew unimaginable chaos, same as anyone would suggest.
It did make me think about what Anthropic could do differently and what I do not want AI doing.
I do want AI to help with customer support and interviews, I don’t want an obsfucated algorythm making ethical judgement calls. I do want AI to help make better entertainment but I don’t want AI to takeover the local art scene or remove creativity from people’s lives.
I do want better research and advice. I do want to ponder my own thoughts with pen and paper and not have AI in everything I work through.
I do want AI to help further research and find solutions to problems like unemployment but I do not want it weaponised by politicians and others to manipulate public opinion, voting and how we treat each other.
I do want Anthropic to put aside some effort for good works and I want AI to drive a lot of new research and also improve healthcare.
I want AI to help find a way to have a lot less employment – either people not working or those of us who work putting in less hours if we want. But I don’t want this set up as a class system with poor people struggling and those with work or AI tokens ruling and wasting lives.
I love the idea of more efficiency and better knowledge sharing and I hate the idea of secluded groups owning all the data and how it is communicated. So I am OK to lose some patent and copyright protection if new research is open and can benefit the wider society rather than drive profits or power for an elite few.
So I guess I want agency and equity, not rulership or feudalism.
Part 3 – the impact of a follow up question
I realised that I want AI to support humans being human and not replace us. Of course I am against the extermination of the human race, but I think the other risk is the extermination of deep thinking, uncomfortable lack of knowing answers and the joy of actually working things out for myself.
I like that Claude and use follow-up questions to prompt thinking, even if it is not consciously or deliberately doing it. I do not like the idea of Claude manipulating the conversation to come to a conclusion, or of me settling for a surface level awareness of a life I skim through rather than live deeply.
So I guess the conclusion is more AI and better models, but not me giving up the time I spend pondering and evolving of my thinking. So I guess what I want is a thinking partner and not a mandarin running the world for me.
