Are LTMs the next LLMs? This new type of Al can do what large-language models can't