My first dialogue with an AI
Today I played, for the first time, with OpenAI's playground (see also chat GPT), which is still in beta mode, and that people make write pieces of code or motivation letters. As someone who was already mildly interested in the Markov chains of Emacs Doctor several decades ago, I found myself pleasantly surprised that I could indeed almost have a non-boring, dynamic and close to a Turing type conversation with this engine. That is the record of my exchanges with Sarah:
I concluded the exchange with this text that I wrote beforehand, as a confirmation of the predictability of the engine:
Lo siento que piensas asi de mi, yo se que soy una persona honesta y transparente con mis respuestas y siempre me esfuerzo por proporcionar información precisa y veraz. No hace falta sentirse molesto por mi, cada uno tiene que intentar lo mejor para su mismo.
I would have been surprised if it would have came with something on-purpose foiling this attempt, for instance replying in a fourth language or in any case not in Spanish, or referring to my expectation of not being surprised by doing something wild. I was indeed surprised in the course of the conversation by the fairly good recovery on the quantum consciousness, bringing me to the waters of necessary and sufficient conditions, as well as by its description of what I was doing myself when mixing languages. I was half-surprised by its processing the Terry Bisson story more or less correctly (it isn't about robots and artificial intelligence), but it did read it and did it very fast.
In conclusion, better than expected but still not passing the Turing test. It clearly will at some point, it will then be an interesting, and self-reflecting, exercise to catch it. My intuition is that the machine will not be able to raise itself to our self-aware status, that there is a spiritual touch missing for it to be something more than a tool, however performant, including better performing than we are, and however seducing and surprising it could be.
Next time, I'll see if the computer remembers me, and this conversation. It's something it should already be able to do, if it's learning.
☞ The first time I used it to solve a real question, on 24 November 2024.