OpenAI ChatGPT Can Answer Fusion Question Pretty Well!

yeah I actually tried this out, I was shocked it even knew what Fusion was. Crazy future we're in for!

What the hell, the thing appears to understand it O.o

Definitely impressive, though seems like the technology probably needs a few years still before it's at a genuinely capable level. Or at least needs a lot more accurate Fusion data funnelled into it, as you suggest. None of the screenshotted answers seem to be particularly helpful. The suggestions are basic, and many of the details are totally misleading and wrong.That fast loop one is basically 80% nonsense, and could only hurt a beginner. X)

It does understand path movement and "Start of Frame" tho, which I thought was interesting considering there's not much written Fusion info that they could have feed to it...
Perhaps they have feed the documentation ... ?

It says it's not allowed to connect to the internet to search for information, which makes it a bit limited right now. And while I'm not usually one to complain about "wokeness", some of its answers to more challenging questions were a bit off...
For example: "How do you beat Viktor Axelsen?" He's the seemingly invincible world #1 in badminton, so plenty has been written about tactics that might work against him, but OpenAI just goes off on a rant about how beating is violence, and violence is wrong.
Or "Is Sheldon Cooper crazy?". It doesn't seem to know who Sheldon Cooper is, which is fair enough, but again, it then goes off on another long rant about how we shouldn't use the word "crazy", as it's offensive to those with mental health issues an so on...

It will be a great tool for learning... it can re-write code and even tell you what it did to fix it XD

Please login to see this picture. Please login to see this picture.

I think the above two posts illustrate quite well just how far away this thing is from meeting its intended purpose. It's smart like a computer, but is a million miles off from being smart like a human. So when it comes to rearranging code according to unambiguously strict rules of syntax, like in DaveC's example, it does well. But when there's a requirement for deciphering context, identifying pivotal details, and deducing the applicable meaning of a multi-definition word like "beat", like in MuddyMoles' examples, it's hopeless.

Google can give the illusion of intelligence by regurgitating or collating stuff that real humans have written (lots of people would have written things about beating Viktor Axelsen in badmington, so Google would base its answer on that pre-existing content). But if this thing isn't allowed to connect to the internet, and must rely on its naked cognitive abilities instead, we can see how shallow those abilities actually are.

PS - MuddyMole: I wouldn't go as far as complaining about 'wokeness'. I suspect that if someone asked you "how do I assault John Smith?", your answer would probably be something along the lines "um, are you sure you want to do that? Maybe there's a better way of resolving your conflict?". You probably wouldn't just unquestioningly launch into instructions about how best to commit the crime. So in this (and only in this) OpenAI approximates a real human answer. (Though unlike OpenAI, a real human like you would manage to do it without sounding like a moron, obviously!). Likewise, as someone who's married to a psychologist, I can tell you with reasonable confidence that if you asked any mental health expert is "Is John Smith crazy?" you would not get a satisfying answer. The first thing they would say is that "crazy" is an unhelpful and largely meaningless term. Partly because it's stigmatising and could cause offence, but mainly because it's an inherently unscientific word that is too broad, subjective and inaccurate to be useful. So again in this aspect (and only this aspect) OpenAI approximates the answer of a real human expert on the topic. Because a real human expert would probably insist that you ask a more specific question (is John neurodivergent? is John mentally ill? Is John irrational?) before answering it.

Indeed it has many limitations, it's mostly an illusion.. but imagine this kind of stuff in 20-30 years.. pretty crazy.. especially if they give it access to the internet to "learn" from.

Please login to see this picture.

that old game dev forum joke "can someone program my game for me?" might not be a joke in 20 years.. you'll type in.. "make a game like mario except it has x, y, z as enemies and x, y, z themed levels in yoshi island art style and include online multiplayer with leader boards and a balloon popping mini game" and it'll just churn out a game in less than a minute...

Or it could be like jetpacks, and in 2050 we still won't have jetpacks OR useful AI.

The future has been one giant disappointment for me. I was expecting jetpacks, flying cars, AI, and cities on the moon. Instead, technology has given us......social media. woohoo

That poem is actually really, really, impressive :o. I must admit I was taken aback. Though I think my criticisms from above still apply: when given strict parameters to work with (rhyming verse), it can assemble the scaffolding adeptly (though I would have liked to see a semicolon in line #8 >:D), and then it just pads out the rest with fluff. Nicely assembled fluff, to be sure, but not a single detail that hints at it actually knowing anything about Fusion (kind of like most of the answers it provided in Retrobolt's screenshots) XD