Rockets @ Suns Tuesday Game Thread, 4-7-2026 ON PEACOCK

[leclerc]

I asked AI to quiz me on 90's NBA players and I had to correct it time and time again. It was a year+ ago so maybe it has gotten better, but if we stop maintaining quality facts and information, it won't get better at all.

 

[Ouchie-Z-Clown]

It's interesting you mentioned AI. I've found so many mistakes made by AI when it comes to the NBA. I guess it's only as good as it's sources. So much bad information out there.

Google might be the smartest though. Gemini voice is really good although I haven't talked basketball with it.

More than the sources, it’s only as good as the prompt. That’s what most people are failing to understand and blaming it on the AI. It’s primarily the prompt that’s the failure.

 

[Mainstreet]

More than the sources, it’s only as good as the prompt. That’s what most people are failing to understand and blaming it on the AI. It’s primarily the prompt that’s the failure.

I have a lot of respect for AI, especially Google/Gemini but I have taken some others down to the mat in sports where they finally admit they were wrong.

It's not the flaw in the AI but rather the sources.

I do agree. Ask a bad question and one can get a bad answer.

 

[Hoop Head]

I have a lot of respect for AI, especially Google/Gemini but I have taken some others down to the mat in sports where they finally admit they were wrong.

It's not the flaw in the AI but rather the sources.

I do agree. Ask a bad question and one can get a bad answer.

Sports AI is iffy. I use it to verify more than research. It works well to double check your work, sort of like a calculator.

 

[Phrazbit]

More than the sources, it’s only as good as the prompt. That’s what most people are failing to understand and blaming it on the AI. It’s primarily the prompt that’s the failure.

From my experience, most AIs will rarely answer in way a way that does not that affirm your position, it will rarely challenge you, it wants to agree, even if you propose something absolutely loony to it.

Sure, the prompt will often have something to do with that, fishing for compliments, but you can take a some insane positions and it is very rare that AI won't try to find some way to agree with you.

 

[Mainstreet]

Sports AI is iffy. I use it to verify more than research. It works well to double check your work, sort of like a calculator.

And it's okay for a quick overview, but I usually fact-check details.

 

[Ouchie-Z-Clown]

From my experience, most AIs will rarely answer in way a way that does not that affirm your position, it will rarely challenge you, it wants to agree, even if you propose something absolutely loony to it.

Sure, the prompt will often have something to do with that, fishing for compliments, but you can take a some insane positions and it is very rare that AI won't try to find some way to agree with you.

That’s not because it’s looking to agree with you. It’s pattern recognition and completion. Most people don’t understand this about AI. It doesn’t want to do anything. It’s not “trying to find” anything. It’s taking the next logical step on pattern completion based its training. It’s algorithmic. So if you create a prompt that tells it how to “behave” it will do so. But you need to be explicit. Tell it to consider contrarian perspectives or to be incompletely unbiased with explicit direction and to remove flowery language and it will do so. Because it interacts with an LLM that feels human people are way too loose with it with their prompts if they want better reaponses.