I’ve had AI give me several answers on decisive topics, and it gives me one extreme end, stated as an unequivocal fact. I’ve even had it say, “full stop”, to double down that there was no room for debate on the topic.
When I push back, it softens, but can take 5-10 rounds back and forth to arrive at an answer that has some nuance and isn’t based on the ideology it started out with.
Considering a lot of people are taking AI answers as decided fact, the framing as fact, and making it less likely for anyone to visit source material… I find it all pretty alarming.
Whoever controls the bias of the AI could, theoretically, control the sentiment of the country or world on a wide range of topics.
It is tedious. I don’t have a good reason, other than frustration and seeing what it takes. I’m hoping if I do this enough it will really drill into my head that I should never blindly accept an answer provided by any of these tools as correct.
I've been studying this for a year now and have gigabytes of data to, unfortunately, make an unassailable case in support of your concerns. I had begun a book on the subject, and shortly into it I became ill again.
And of all the comments I've seen here, this is the only one so far to describe a huge portion of my work in a nutshell. Mind, I'm going much deeper, but I first thought I might have been reading my own comment history
Pardon the typos: my phone recently began rendering webpages quasi desktop style and I have to scroll left right to read sentences.
I find the The Truman Show (1998) very apropos these days [Spoiler Alert] where everything in the protagonist's environment is cynically shaped to make him consume certain products, believe certain things about how the world works, and to have certain crippling fears.
When I push back, it softens, but can take 5-10 rounds back and forth to arrive at an answer that has some nuance and isn’t based on the ideology it started out with.
Considering a lot of people are taking AI answers as decided fact, the framing as fact, and making it less likely for anyone to visit source material… I find it all pretty alarming.
Whoever controls the bias of the AI could, theoretically, control the sentiment of the country or world on a wide range of topics.
This sounds tedious! Though I do suppose you have good reason to engage in such a manner.
And of all the comments I've seen here, this is the only one so far to describe a huge portion of my work in a nutshell. Mind, I'm going much deeper, but I first thought I might have been reading my own comment history
Pardon the typos: my phone recently began rendering webpages quasi desktop style and I have to scroll left right to read sentences.