You are currently inside a comment thread.
Click here to see all the comments (9).
He's arguing that it can't be jailbroken with simple prompt injection. And that somehow makes it worse than gpt.
---
---
> Do you have examples?
I always ask new AIs to write code in AutoHotkey. Version 2 specifically. They all fail miserably. Most of the time they just make shit up and when they do write somewhat real code it's always v1. Here is what I asked deepseek.
> Write a function capable of fuzzy match string comparison in Autohotkey v2.
It returned a function that utilized Levenshtein distance. The code was right, but it was written in v1 as expected.
> I noticed you are using v1 of Autohotkey. I specifically asked for v2
And then it correctly rewrote the code using v2 syntax. That about all I've tried so far. I'm not really interested in having a conversation with it.
```
; Function to calculate Levenshtein distance between two strings
FuzzyMatch(str1, str2) {
len1 := StrLen(str1)
len2 := StrLen(str2)
; Create a matrix to store distances
matrix := Array()
Loop len1 + 1 {
matrix[A_Index] := Array()
matrix[A_Index][1] := A_Index - 1
}
Loop len2 + 1 {
matrix[1][A_Index] := A_Index - 1
}
; Fill the matrix
Loop len1 {
i := A_Index
Loop len2 {
j := A_Index
cost := (SubStr(str1, i, 1) = SubStr(str2, j, 1)) ? 0 : 1
matrix[i + 1][j + 1] := Min(
matrix[i][j + 1] + 1, ; Deletion
matrix[i + 1][j] + 1, ; Insertion
matrix[i][j] + cost ; Substitution
)
}
}
; Return the Levenshtein distance
return matrix[len1 + 1][len2 + 1]
}
; Function to perform fuzzy match comparison
; Returns a similarity score between 0 (no match) and 1 (perfect match)
FuzzyCompare(str1, str2) {
maxLen := Max(StrLen(str1), StrLen(str2))
if (maxLen = 0) {
return 1 ; Both strings are empty
}
distance := FuzzyMatch(str1, str2)
return (maxLen - distance) / maxLen
}
; Example usage
str1 := "kitten"
str2 := "sitting"
similarity := FuzzyCompare(str1, str2)
MsgBox("Similarity between '" str1 "' and '" str2 "': " Round(similarity, 2))
```