Good day.
OpenAI released the latest models: GPT-5 and GPT-5 Thinking.
I usually on GPT-4o for casual banters about anything. Most of the idle times, about films and telly shows I watched randomly.
For instance, why did people in a jungle expedition (in those films) wear tank tops knowing they'd be, at least, in a jungle? Shouldn't safety be the first consideration? Not that we can't wear tank tops in the jungle. But when the story is about gigantic venomous, aggressive spider, reptile, other muppets — it means safety is numero uno. Not one of them brought pesticide, none. Lalala, let's stroll around here, let the wind carries us away.
At least, I'd expect, once they met the beasts along the way, ran away in panic — oh hey, let's not go there.
AND.... they did the opposite. 😂🤷 Proper, that. We don't munch popcorn to watch professional pest-control blokes do their job, do we?
Oi, Bob! What's that popcorn for? Any good film?
Termites.
What does that mean?
Blimey, "Termites" is an actual film. 🤣 Bob! It's your fault. But, then again, how is it possible for termites to change diet? Classic mutation trope. Never mind the overhaul of their biology and society. Let's mind that. Like we, the omnivores, suddenly chewing beams? 🤣🤦 Oh, that's caused by chemical spill. The teeth and such... We live in burrows now. Take that, Bayerische Motoren Werke. BMW would mutter, Was habe ich getan? 🤔
Aside from the "oops" above, GPT-4o could go hilarious most of the times, I was in tears for its antics. But then again, I did spark it first with my own daft observations and inquiries. Thus we went proper nutters.
By the way, I am a Plus subscriber on ChatGPT.
GPT-4o and OpenAI o3-mini high
For coding and logic, GPT-4o could go unexpected, loopy and such. If that happened, I went (changed the model) to o3-mini high for more precise, more advanced evaluation about the "problem". And yes, in my experience, 99% of the time, o3-mini high did the job.
GPT-5
Since this morning, I tried GPT-5 because the options are only those two models.
GPT-5 outputs shorter and more concise response compared to GPT-4o. Even in casual banter. It's a bit... off. 🤷
Like we're chatting with a bloke with too many IQ points. Or just a bloke. With encyclopaedia in his head.
But when I tried Bootstrap v5.3 "problem", wow, it went like o3-mini high — it solved it like a heavyweight boxing champion against a 6-foot-tall soda can. 🤔 Though I did it with multiple prompts until it found the correct solution and alternative. But marvellously, it didn't go swirling back and forth in a maze like GPT-4o usually did.
So, well, it's a marvel nonetheless, with the short-response quirk.
Like, What's this movie called? [placing snapshot] ➡️ It's called [movie name and details]. Well, that actually answers the question.
Or, Oi mate, what's this bloke doing? [placing snapshot] ➡️ [precise anatomical description]. Hm, so that's what it's called.
And we should say, Where's the charm there? The wit? ➡️ Eh. 🤣
Sam Altman should add another name to that, like Ctrlman. As in "Control" man. Thus he would have Ctrl key, then Shift. The entire QWERTY gang awaits.
3-4 seconds for that reply — GPT-5 Thinking. They should branch it, only for actual analysis which will take... oh, they probably did. 😂
ChatGPT web developers are like mug-smashing frontend lads, trail-blazing backend wizards of sleepless realms, software architects of caffeine land, certified Scrum necromancer, pixel-perfect alchemists of CSS, Docker druids, Kubernetes conjurers of the cloud, and so forth.
From GPT-5.
Can you spot the cloned people in the image above? 🤔😶
Conclusion
Thus to me, indeed, GPT-5 was noticeably different from the usual GPT-4o. Yet when we bantered about logic, its precision was nothing short of remarkable. So I don't mind about its brevity, that much.
Comments
Post a Comment