Corellary: The extreme excitement and surprise about AI output in a media culture of cherry-picking has produced epic confirmation bias, which feeds a hype machine that eclipses the stark reality.
As AI (mis)application accelerates, the stories about coping with AIs limitations are usually couched in a tone of opportunity: "I made a tool to help you find where your AI is wasting money!"
There's a simple question that can be asked about any application context of AI that puts the hype into proper perspective: Why hasn't the AI already accounted for and mitigated your problems with applying it?! From the vantage of this simple question, you realize that you believe there's a line about what AI cannot be expected to do... so the next question you face is: why do I think this line of AI limits can be drawn beyond the scope my desired application rather than within the scope?
At this point, you will notice that AI-application success stories too commonly sound like "I tried this and it looks like it works," whereas actual engineering is the application of proven principles, techniques and methods to the solution of problems that pre-exist any novel technology used for the solution.
Repeat: For any problem you are facing applying AI, ask "Why can't AI solve this problem?" then get on with traditional approaches to engineering free of confirmation bias.
Referring back to the example of excitement about tools that help you overcome AI limits, the misapplication of
"I tried this and it looks like it works," whereas actual engineering is the application of proven principles, techniques and methods to the solution of problems that pre-exist any novel technology used for the solution.
Bingo. We're basically in mass hallucination where credentialed experts believe these sophisticated pattern matchers are gods rather than realizing their own experience is bleeding into the prompts they type
Because AIs can have no idea about correct code?
Corellary: The extreme excitement and surprise about AI output in a media culture of cherry-picking has produced epic confirmation bias, which feeds a hype machine that eclipses the stark reality.
As AI (mis)application accelerates, the stories about coping with AIs limitations are usually couched in a tone of opportunity: "I made a tool to help you find where your AI is wasting money!"
There's a simple question that can be asked about any application context of AI that puts the hype into proper perspective: Why hasn't the AI already accounted for and mitigated your problems with applying it?! From the vantage of this simple question, you realize that you believe there's a line about what AI cannot be expected to do... so the next question you face is: why do I think this line of AI limits can be drawn beyond the scope my desired application rather than within the scope?
At this point, you will notice that AI-application success stories too commonly sound like "I tried this and it looks like it works," whereas actual engineering is the application of proven principles, techniques and methods to the solution of problems that pre-exist any novel technology used for the solution.
Repeat: For any problem you are facing applying AI, ask "Why can't AI solve this problem?" then get on with traditional approaches to engineering free of confirmation bias.
Referring back to the example of excitement about tools that help you overcome AI limits, the misapplication of
"I tried this and it looks like it works," whereas actual engineering is the application of proven principles, techniques and methods to the solution of problems that pre-exist any novel technology used for the solution.
Bingo. We're basically in mass hallucination where credentialed experts believe these sophisticated pattern matchers are gods rather than realizing their own experience is bleeding into the prompts they type