A trick I do with ChatGPT 3 is type in student questions.
One thing it is generally bad at is recognising bad questions. Someone who knew the frameworks would probe a little and figure out that the student was looking up a dead end.
I think it’s great, and sometimes good enough for me to use as the basis of my answers.
I feel like anyone who claims chatgpt is helping them work is just telling on themselves as trying to do things out of their depth and producing output they don't understand well enough to vet.
ChatGPT sets a baseline. If my answer isn’t better than ChatGPT there would be no point in employing me. But students use it all the time. They use it to cheat. They use it to brainstorm.
I have plenty of criticism for chatGPT. It attempts to fake nuance by creating shallow syntheses of conflicting info. It doesn’t understand anything it talks about systemically. It has no insight into the user.
But it can regurgitate a list that you can use to structure your answer to a student’s question. It’s interesting to me because people do want me to believe it’s coming for my job. It’s a way off yet.
It’s somewhat useful to brainstorm, it really helps with the empty page problem but it’s not good and the resultant nudge could be in a very generic direction. Instead of promoting creativity it pushes you toward sameness.
It's just a tool to be used--didn't say it solves every problem. I think of it like a better version of a search engine. ChatGPT basically just summarizes the best (i.e. most popular) content out there. So if you search "Singapore notary/certification laws" you get a better/faster 3 sentence summary than if you were to click through Google results. I find it useful for things, but only so far as it points you in the right direction and provides some useful starting point to then expand on manually.
I think companies are worried about how unpredictable and abusable this currently is. Imagine how parents feel when their 12 year old jailbreaks an npc and starts having explicit conversations with it.
It depends on how it’s implemented. Maybe you could train mmo npcs in this and make it fun but I agree I expect this to simply generate a dumpster fire experience
Thank god that interest rates are no longer zero, which propped up all kinds of over hyped, half baked crap in the last 20-30 years.
It will take care of hypsters and their proven 'raise gobs of cheap cash on hype and capture market as fast as possible to hide half baked product' strategy.
With falling interest rates it was possible to burn money in a furnace and still show an accounting profit on your operation. That’s how screwed up the incentives were
23 comments
[ 4.6 ms ] story [ 57.6 ms ] threadOne thing it is generally bad at is recognising bad questions. Someone who knew the frameworks would probe a little and figure out that the student was looking up a dead end.
I think it’s great, and sometimes good enough for me to use as the basis of my answers.
I have plenty of criticism for chatGPT. It attempts to fake nuance by creating shallow syntheses of conflicting info. It doesn’t understand anything it talks about systemically. It has no insight into the user.
But it can regurgitate a list that you can use to structure your answer to a student’s question. It’s interesting to me because people do want me to believe it’s coming for my job. It’s a way off yet.
Certainly I find that it reminds me of what I know that it doesn’t.
Thinking about it more fully it’s like it pulls the “Cunningham’s Law” trigger in me that a lazy question just can’t.
It will take care of hypsters and their proven 'raise gobs of cheap cash on hype and capture market as fast as possible to hide half baked product' strategy.
Apple version of GPT-n will certainly find its way to users, but will most likely have some serious bumpers that will take time to roll out.
If Apple has shown anything it’s that they are ruthlessly patient