It might look like skepticism about AI, but I don't think Chat-GPT will be the winner among the AIs to come in the future.
Hallucination
To prevent this, you need to invest a lot of resources (Grok, Gemini Pro) but Chat-GPT can't do that.
Although it's still performing well... as the system becomes more of a liar day by day, it's becoming impossible to use
Today's question: So when cross_vol + di_break are applied simultaneously, it's 1.5 x 1.5, right?
ChatGPT's answer:
• No, the current implementation is not 1.5 × 1.5 = 2.25
**"only one largest multiplier is applied"**.
That is, currently:
- Basic entry multiplier (e.g., cross_vol 1.5) is applied
- Then the overlap multiplier applies only the maximum value
→ If cross_vol + di_break are simultaneous, only 1.5 overlap multiplier is added
So the result is not 1.5 × 1.5 = 2.25
1.5 × 1.5 = 2.25? ← No
1.5 (entry) × 1.5 (overlap max) = 2.25 correct.
(The overlap "one maximum value" means applying only the larger of 2.0 and 1.5.)
To summarize, when cross_vol + di_break, the actual applied multiplier is 2.25.
The reason for answering this way is simple. It means not seeing the full context. Even if the context window is large, if it's going to save resources this way, it's unusable. And I'm paying 300,000 won a month at that.
Isn't it time to let go of ChatGPT now?
