In the 80s there were AI researchers who believed that once all of human knowledge was stored on a computer system, it would spontaneously become intelligent. This, of course, was magical thinking, and no such thing happened.
It still won’t happen.
Listen to what the actual experts are saying
I do, and the actual experts all understand that the current generation of machine learning models are optimization algorithms that have been refined to locate specific areas of large datasets based on input prompts. They’re fancy search query tools, nothing more. There’s no “intelligence” at work.
When you have a “conversation” with an LLM, and it gives you a seemingly-intelligent response, what you’re getting is a dump of some copied text that is related to the input you’ve been feeding the text entry field. The “intelligence” that you perceive in the response is actually that of the human who made the original writing that was scraped for training data.
Keep your credulity in check. The algorithm has no understanding of what it regurgitates, no comprehension of the information it handles. It does keyword association really really well, and it can sift through lots and lots of data impressively fast. It’s a great, big, environmentally destructive Chinese room.
What agents?
In the 80s there were AI researchers who believed that once all of human knowledge was stored on a computer system, it would spontaneously become intelligent. This, of course, was magical thinking, and no such thing happened.
It still won’t happen.
I do, and the actual experts all understand that the current generation of machine learning models are optimization algorithms that have been refined to locate specific areas of large datasets based on input prompts. They’re fancy search query tools, nothing more. There’s no “intelligence” at work.
When you have a “conversation” with an LLM, and it gives you a seemingly-intelligent response, what you’re getting is a dump of some copied text that is related to the input you’ve been feeding the text entry field. The “intelligence” that you perceive in the response is actually that of the human who made the original writing that was scraped for training data.
Keep your credulity in check. The algorithm has no understanding of what it regurgitates, no comprehension of the information it handles. It does keyword association really really well, and it can sift through lots and lots of data impressively fast. It’s a great, big, environmentally destructive Chinese room.