The problem with chatgpt is that I can't evaluate the source it used to learn the information to see if I think it's credible, and chatgpt can't give me any indication of whether it's credible or not, and its answer will be just as authoritative in tone if it's completely incorrect as if it's correct.
I asked it for a schematic to connect an i2c device to a microcontroller, it told me to directly connect the device to the microcontroller. I asked it if pull-up resistors were required, it said yes, and then drew what it called "pull-up resistors" (describing them correctly) incorrectly in series between the bus and the device.
I know that's wrong, but someone who doesn't really know what a pull-up resistor does or why they are there would be at least somewhat confused by that interchange, and would be misled into doing it wrong. Meanwhile, I can type in "i2c schematic" to google and get thousands of correct schematics back to look at.
I asked it for a schematic to connect an i2c device to a microcontroller, it told me to directly connect the device to the microcontroller. I asked it if pull-up resistors were required, it said yes, and then drew what it called "pull-up resistors" (describing them correctly) incorrectly in series between the bus and the device.
I know that's wrong, but someone who doesn't really know what a pull-up resistor does or why they are there would be at least somewhat confused by that interchange, and would be misled into doing it wrong. Meanwhile, I can type in "i2c schematic" to google and get thousands of correct schematics back to look at.