“Although ironically the model itself likely has no actual understanding of what it solved.“
Yet.
Terminator was a documentary.
The issue here is that LLMs (Large Language Models) from the first examples to the latest, all suffer from the same limitations. They have improved in some ways but the basics have remained the same.
Quoting from Google's most advanced Gemini model itself, “An LLM only sounds self-aware because it has ingested millions of human conversations discussing self-awareness, allowing it to mathematically predict what a self-aware response should look like. It is a highly convincing mirror, not a mind.”
If I had not spent a great deal of time working with a wide variety of artificial intelligence models it would not be so obvious to me that they all have similar limitations at this point in time. Large language models are very useful tools in many ways, but they are a complete dead end when it comes to achieving true human like intelligence. $Billions have been spent developing them over the past few years all over the world and true progress toward actual intelligence has been stalled for an extended period of time. Evolutionary progress has been made to improve their output, but the entire concept has been flawed from the beginning by a ceiling that cannot be broken through without an entirely new concept that will necessarily have to be built from the ground up.
Terminator was not a documentary. It is fantasy that has captured the imaginations of people in our society who have no firsthand experience working with AI tools.