All LLM models (ChatGPT, Gemini, Grok...) work to understand and generate text that resembles how humans use language. We ask them something, they understand what we say, they think about what they are going to say, and they answer the question we ask them.

 

LLMs are dominating the market and society, we use them to think of answers for us, create content for our work or study, find answers faster than we would otherwise, create virtual assistants and optimize our results. These LLMs (like ChatGPT) train their data with books, articles, websites, forums, conferences and other information they find on the internet. They are pre-trained to predict the next word and thus learn linguistic and semantic patterns; and then they adjust to the tasks required of them (like answering the questions we used to ask Google).

 

The problem is that these models fail to understand how people manage stress, emotions and make decisions. They fail to understand what it means to be human and how emotions govern our rationality. These models try to inferring personality traits, but they are flawed because the interactions they have led to biased results that do not accurately reflect a person's true personality traits. They are therefore unreliable for personal assessments in sensitive areas such as hiring, education and mental health.