ChatGPT and A.I. ‘hallucinations’ will ever go away: ‘This isn’t fixable’
fortune.com
fortune.com
“ Yes, in the context of AI, “hallucination” refers to a phenomenon where a machine learning model generates or perceives something that isn’t present in the input data. This can occur in various tasks like image recognition, natural language processing, or generative tasks.
Hallucinations in AI models may result from biases or artifacts in the training data, or from the model overfitting to noise or unimportant details. They can lead to incorrect or misleading outputs and are a subject of ongoing research and concern in the field of AI. Ensuring that models generalize well from the training data to unseen examples is a key challenge in preventing hallucinations.”
HN: Discussion - (35-comments - 2023-08-02) : https://news.ycombinator.com/item?id=36977096
Is "intelligence" merely a matter of data collection and statistics?
I think not.
But a poor simulation of it might be.
And this is where we are today. It is where AI has always been.
My basic contention is this: It's not possible to engineer/build something that has yet to be defined.
And "intelligence" has yet to be defined. But I'm pretty sure that if/when it is, the definition will involve something more than a statistical database or a word prediction engine.