Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Show HN: AI Playground by Vercel Labs

play.vercel.ai

Post reply on HN
Print view View original

61–70 of 103 posts

Re: Show HN: AI Playground by Vercel Labs

#63

Post by ImaCake »

phnofive wrote: post #55
IanCal wrote: post #49

Earlier quoted context omitted.

Here's gpt4, it's formatting is better but this is the text. Though this isn't really the right way to use it, single responses miss a large amount of the potential imo. Startup Idea: SmartUpCycle Overview: SmartUpCycle is a disruptive startup aimed at revolutionizing the way we handle waste and promote sustainability through upcycling. The platform uses advanced AI and machine learning algorithms to intelligently id…

> The platform uses advanced AI and machine learning algorithms to intelligently identify and match discarded materials with potential new uses, connecting individuals and businesses with creative solutions for repurposing waste. In other words, GPT-4

I get this is in jest, but I think the actual solution is to use image classification with a framework of normal code to glue it together. GPT-4 could help you code it…

Re: Show HN: AI Playground by Vercel Labs

#65

Post by VierScar »

pastacacioepepe wrote: post #62

It's completely broken. I couldn't even get one full response from one of the AIs, it always gets stuck near to the completion of the first response. After like 5 tries it told me that I used all my requests. lol.

Same :( I just wanted to see what Google's Flan would give to "77 + 3 * 11" and it says "Something went wrong" and now I'm rate limited

Re: Show HN: AI Playground by Vercel Labs

#66

Post by 0vermorrow »

pastacacioepepe wrote: post #62

It's completely broken. I couldn't even get one full response from one of the AIs, it always gets stuck near to the completion of the first response. After like 5 tries it told me that I used all my requests. lol.

Well it's not broken, I'd assume they limit the `max_tokens` to minimize costs.

Re: Show HN: AI Playground by Vercel Labs

#67

Post by greyman »

spaceman_2020 wrote: post #61

GPT-4 is miles ahead of the competition. It's not even remotely close. I asked it to help me design an app for a specific use case. It came up with a list of features, the tech stack, and even designed an initial database.

Yes and no. Or, yes for such use cases you mentioned, also it is miles ahead in reasoning capability, translations from Chinese, etc. etc. But still, for more casual talk, like bouncing ideas, brainstorming etc., I prefer Claude. It's kind of more "friendly", less formal, and, which is sometimes also important, much quicker.

Re: Show HN: AI Playground by Vercel Labs

#68

Post by osteele »

osteele wrote: post #33

Very nice! If you want to add a feature, feel free to make use of https://github.com/osteele/prompt-matrix.js , which I (well, ChatGPT and I) created for use in my own playground so that I could enter e.g. "Write a program to look for anagrams, in the style of ".

To be clear, that’s a link to a library meant for use in projects like yours. I’m not promoting a competing project (and haven’t provided a link to one).

Re: Show HN: AI Playground by Vercel Labs

#69

Post by capableweb »

greyman wrote: post #67
spaceman_2020 wrote: post #61

GPT-4 is miles ahead of the competition. It's not even remotely close. I asked it to help me design an app for a specific use case. It came up with a list of features, the tech stack, and even designed an initial database.

Yes and no. Or, yes for such use cases you mentioned, also it is miles ahead in reasoning capability, translations from Chinese, etc. etc. But still, for more casual talk, like bouncing ideas, brainstorming etc., I prefer Claude. It's kind of more "friendly", less formal, and, which is sometimes also important, much quicker.

Have you used GPT-4 via the API or via ChatGPT only? The difference is huge, as with the API you can control the parameter + you can set your own system prompt (basically steer the AI better).

With the API, you can basically customize the session ranging from providing batshit bananas replies (by setting high temperature) or almost deterministic replies (by setting low temperature) with personality controlled by the system prompt, so you can make it "friendlier", "less formal" or anything else.

It is still as slow via the API as via ChatGPT though, which kind of sucks when you're using it for longer passages. I do wish it was just about 100% faster.

Post reply on HN