Small Language Models, Data exhaustion, Prompt engineering, LLMs hallucinations, how they can help make new scientific discoveries and whether their emergent capabilities are just a mirage.