Hi guys.
I’m having an hard time with something that should have been super simple.
If I ask chatgpt4
Browse this investor website [website] and tell me which investment they made (3 max) a. Answer me in json and name them company1,company2,company3. Do not add anything else. If you can’t find any investment, answer NO INVESTMENTS.
The answer is amazingly perfect
{
“company1”: “inGain”,
“company2”: “Groover”,
“company3”: “Scorestars”
}
But if ask the same thing to OpenAI using the same engine
{
“model”: “gpt-4”,
“prompt”: “Browse this investor website [website] and tell me which investment they made (3 max) . Answer me in json and name them company1,company2,company3. Do not add anything else. If you can’t find any investment, answer NO INVESTMENTS”,
“temperature”: 0.7,
“max_tokens”: 150,
“top_p”: 1,
“frequency_penalty”: 0,
“presence_penalty”: 0
}
The answer is always NO INVESTMENTS or some company names totally invented.
Is there a way to fix this?
Why ChatGPT answer are always perfect while OpenAI APIs are so less precise?
Is because OpenAI APIs can’t access the internet?