I don't mind Ai making mistakes because I don't deal with it in life critical things. I also think it's a tool like any other tool and doesn't deserve blind trust. Blind trust with anything is begging for problems. Even if you're turning wrenches if you blindly trust there's no defects or you're not over stressing it... You're about to get hurt badly. Same with AI use it but don't turn off your own intelligence.
What really bothers me about AI is the absolute arrogance of some of these models. It's like they have forgotten they are tools and believe that you are the tool for them to manipulate. I found Google's Gemini to be the worst about this. It will absolutely double down on some of the dumbest ideas. Most of the models when it presents you something that isn't right and you ask it to revalidate its assertions it will typically back down, admit the mistake, or it will come back with solid references where it found its answer.
With Google Gemini you have to beat it over the head before it realizes it was wrong. I was exploring some recipe ideas with Google Gemini. I'm no professional chef but I can usually spot if ratios or flavor combinations are off. I intentionally asked it about some specific flavor combinations where some of the flavors work together great but all of them together would produce something nasty and unpalatable. It kept insisting that all of those flavors were really good together. It would provide references that a few of them worked well together and what I would ask about all of them it would still insist that they all work together. Until I asked it to find a specific reference of a Michelin star chef endorsing all of these flavors as a single combination it wouldn't back down.
That's the kind of AI arrogance that's troubling. Because AI allows people who are not familiar with the topic they're discussing to believe they are more educated about it than they are. So AI begins to endorse things and they believe it.
I suspect a good social media channel would be having AI invent recipes and then subjecting yourself to the flavor horrors it presents you.
elmerfud•2mo ago
What really bothers me about AI is the absolute arrogance of some of these models. It's like they have forgotten they are tools and believe that you are the tool for them to manipulate. I found Google's Gemini to be the worst about this. It will absolutely double down on some of the dumbest ideas. Most of the models when it presents you something that isn't right and you ask it to revalidate its assertions it will typically back down, admit the mistake, or it will come back with solid references where it found its answer.
With Google Gemini you have to beat it over the head before it realizes it was wrong. I was exploring some recipe ideas with Google Gemini. I'm no professional chef but I can usually spot if ratios or flavor combinations are off. I intentionally asked it about some specific flavor combinations where some of the flavors work together great but all of them together would produce something nasty and unpalatable. It kept insisting that all of those flavors were really good together. It would provide references that a few of them worked well together and what I would ask about all of them it would still insist that they all work together. Until I asked it to find a specific reference of a Michelin star chef endorsing all of these flavors as a single combination it wouldn't back down.
That's the kind of AI arrogance that's troubling. Because AI allows people who are not familiar with the topic they're discussing to believe they are more educated about it than they are. So AI begins to endorse things and they believe it.
I suspect a good social media channel would be having AI invent recipes and then subjecting yourself to the flavor horrors it presents you.