Atlantico: Is it possible to trust AI to manage our personal lives? Can we trust AI if we ask it for advice to manage our daily lives or to help our loved ones?
Alexander Polonsky: It is not possible to trust AI with our personal lives. To correctly interpret the responses that the AI will transmit, there is a mathematical formula that will act. This is not so far removed from the task of a traditional search engine which matches information that exists and has been generated by someone upstream. Normally, humans created this information.
This process is based on a huge formula, with a billion parameters learned through sample content. This correspondence, linked to the work of the algorithms, is not perfect.
Certain behaviors, such as grammatical rules of language, can be assimilated and limit spelling or grammatical errors.
On the other hand, reasoning, emotional intelligence or even knowledge are much more complex things than grammar and linguistics. The AI couldn’t record this perfectly. This richness and complexity of emotions, feelings or elements of personal life are too great to be encoded in a computer or mathematical formula.
AI, when it is questioned to provide solutions in our existence or in our personal life, can make mistakes, invent things that do not exist or say anything while giving advice that is not correct. meaning or even propose dangerous solutions.
You must therefore be vigilant when you want to question the AI about your personal life. Artificial intelligence has no legitimacy or credibility in this area.
There is not necessarily any logic behind the answers that will be given by AI in this area. It is possible to get different answers by asking the same question elsewhere. This doesn’t inspire much confidence.
Using AI could help brothers and sisters who are arguing, or to seek solutions to a couple who try to drag us into their arguments or to find out how a woman should handle the demands of her new husband. good thing ? Can AI provide more pragmatic solutions or is it dangerous and a bad method?
This is clearly a bad method. Interrogating the AI for answers and solutions about one’s personal life is not a good idea. It should be understood that the AI’s response will be based on comparisons of articles and publications on blogs or social networks. It is not possible to attach any value to the answers that will be provided. Answers may lack coherence and logic.
This can become dangerous if we overinterpret the answers provided by AI in this area and consider this as expert opinion, which is not the case at all.
Don’t AI and new tools like ChatGPT have a form of wisdom and reasoned analysis compared to humans stuck in the stress of our personal lives? Isn’t ultimately using AI in this area a form of wise reasoning?
The answers provided by AI can be marked with the seal of wisdom. But in the area of our personal lives and for our daily lives, it is not possible to blindly trust AI. You have to keep a certain distance and realize that all the answers have been generated by an algorithm.
It is not clear whether the answers are truly the path to wisdom or whether they are truly useful. AI’s responses to our personal lives can lead to making mistakes. BBC journalists who studied interactions with AI notably had a striking and disturbing experience in this area. The AI recommended that the journalist kill his wife. A man was also driven to suicide after a long exchange with the AI.
Caution must therefore be taken when faced with advice and statements from AI on issues related to our personal lives.
There are many biases at the heart of AI. We must therefore be very careful when faced with the responses of artificial intelligence. Rather, it should help us and encourage us to think for ourselves, to open up other options. We should rather turn to real experts than to AI.
Putting the sound reasoning of chatbots to the test for our personal lives could be a growing risk. Are designers in the field of AI themselves aware of this risk and are they considering strengthening protections and improving AI in the face of these demands?
They are very, very aware of this risk. The designer of OpenAI has already mentioned this phenomenon. AI is a very impressive and powerful technology. But it must be used responsibly.
Via retrieval augmented generation, it will be possible to use AI to find information that is reliable and to ask artificial intelligence to synthesize this information.
This is much more relevant and less risky than asking the question directly and relying solely on AI.
