The title of this blog post reads like the start of a lame nerdlinger joke. Actually let’s see how 3.5 and 4.0 finish this potential belly buckler…
3.5’s attempt:
Booo! Garbage. And yes, I tried other prompts and role playing. If you want to test for ChatGPT, it’s simple…test for humour. You wouldn’t invite 3.5 to a casual dinner party..funeral, maybe.
4.0’s attempt:
Wait a minute……I smiled at this….what the?…then a smily face and a door to illustrate its crafty joke construction…cheeky monkey…:)
And herein lies the difference between 3.5 and 4.0. It comes across smarter and more adept at swerving through your meddlesome prompts. I have been amazed more by 4.0. 3.5 is already dead to me. Faster, yes, but dead.
There’s another example I want to include here and it’s one that involves coding. Something I tried and failed miserably at in my first university experience at the turn of the millenium. I asked both 3.5 and 4.0 to generate code in Python to add two numbers together.
3.5
4.0
Now, 4.0 is infinitely more useful here because I had no clue how to even run Python code, so a win right there out of the box. Having downloaded Python and renaming the file I ran the program. However, both programs closed after inputting the second number.
Why? I asked both bots. Actually, I asked “when i hit return after the second humber, the application closes without showing me the sum, what could be wrong?”
3.5 told me to: “To fix this issue, you can modify the program to remove the newline character from the user input using the strip() function”.
Did not work. Told it so, it then gave me another long list of debugging options. No thanks. 4.0 where art thou?
4.0 response to my initial plea for help:
Boom, result. It worked. 4.0 is better than 3.5
In other areas, it feels like 4.0 just “listens” to prompts better:
Prompt: “act as an art history influencer on youtube, provide an informative but fun analysis of Gustave Caillebotte’s Paris Street; Rainy Day. Include the artist’s main points from his biography and the historical context of the painting”
3.5
It basically ignored my “fun” command. Boorrriiinnggggg….
4.0
Much better 4.0, bit more character, a bit more fun. It even gave a name for our channel..I already have a name for it, thank you very much.. Overall, a lot less bland and dare I say it…a lot less ChatGPT-ish…
Which leads to another prompt engineering point: the one of role play. The prompt to get more flavour and character to ChatGPT generated content. We might explore that more at a later date…I have explored retelling the history of Ireland through different authors/Quentin Tarantino. MUST SEE!
….I’ll leave it at this. 3.5 has evolved into 4.0 and sprung several new wings. It reacts more effectively to prompts and is overall more helpful from the get-go. ChatGPT 5.0 will most likely, in all seriousness, start replacing jobs…