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

71–80 of 103 posts

Re: Show HN: AI Playground by Vercel Labs

#71

Post by carrolldunham »

vertis wrote: post #45
alehlopeh wrote: post #44

Earlier quoted context omitted.

But you didn’t ask it to pitch a new book. At one point, Pattern Recognition was actually his next book, and there’s probably a lot of text out there discussing it as such.

Sure absolutely, but would a human make this mistake? You can't pitch something that already exists (I mean in this context we're pretending, so maybe we can pretend we're pitching that book). But a human (and GPT-4 does) assumes you want an original book. It's still an assumption. I was playing around with GPT-4 making it take the questions for an Economics exam the other day and it was missing a lot of context that…

It's not a mistake. It's just ambiguous - you and the OP (rather than "humans") share some contextual assumptions that to you make it imply one meaning. If someone asked me with no context to 'be william gibson pitching your next book' I might answer just the same as gpt did

Re: Show HN: AI Playground by Vercel Labs

#72

Post by jondwillis »

Trystanr wrote: post #41
atleastoptimal wrote: post #31

How much does it cost to query these models all the time? It's offering a lot for free.

There is a rate limit.

Even still. GPT4 in particular gets expensive at scale.

Vercel is probably noticing that the LLM community is using NextJS (and Vercel) by default a lot of times, and that supporting the community and building more goodwill will pay off.

Re: Show HN: AI Playground by Vercel Labs

#73

Post by zurfer »

Thanks for setting up a limited free version. The UX is great.

I had some failures for huggingface and cohere. Replicate produced an output that was not rendered (but returned an answer in the network tab).

It would also be great to add: https://docs.aleph-alpha.com/docs/introduction/luminous/

Re: Show HN: AI Playground by Vercel Labs

#74

Post by greyman »

capableweb wrote: post #69
greyman wrote: post #67

Earlier quoted context omitted.

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…

I use both ChatGPT and API. Mostly I would say I use it for work and when I learn Chinese, and Claude is in my mobile via Poe app, which I use for more casual talk or when I know the answer doesn't require advanced reasoning. Then I also use phind.com website, which I would say is like a better Bing (at least for programming queries).

Re: Show HN: AI Playground by Vercel Labs

#75

Post by vertis »

andybak wrote: post #57
vertis wrote: post #45

Earlier quoted context omitted.

Sure absolutely, but would a human make this mistake? You can't pitch something that already exists (I mean in this context we're pretending, so maybe we can pretend we're pitching that book). But a human (and GPT-4 does) assumes you want an original book. It's still an assumption. I was playing around with GPT-4 making it take the questions for an Economics exam the other day and it was missing a lot of context that…

> but it lacked the ability to step back and recognise it had taken a nonsense path. Errr. Yes! We're at the stage where you should be amazed how capable these things are considering their fairly brute force nature. Being surprised at the limitations is a peculiarly - dare I say? - human response? Usually the workflow involves: 1. Write a naive prompt 2. Gain some insight into AI's behaviour 3. Improve prompt and rep…

It's all astonishing. I don't think it's so much surprised at the limitation as trying to explore the edges. Enjoying finding some of the emergent behaviours.

Re: Show HN: AI Playground by Vercel Labs

#76

Post by wut42 »

jondwillis wrote: post #72
Trystanr wrote: post #41

Earlier quoted context omitted.

There is a rate limit.

Even still. GPT4 in particular gets expensive at scale. Vercel is probably noticing that the LLM community is using NextJS (and Vercel) by default a lot of times, and that supporting the community and building more goodwill will pay off.

Vercel is so overpriced on their hosting plans that they can afford that to attract new customers, I guess.

Re: Show HN: AI Playground by Vercel Labs

#77

Post by vertis »

carrolldunham wrote: post #71
vertis wrote: post #45

Earlier quoted context omitted.

Sure absolutely, but would a human make this mistake? You can't pitch something that already exists (I mean in this context we're pretending, so maybe we can pretend we're pitching that book). But a human (and GPT-4 does) assumes you want an original book. It's still an assumption. I was playing around with GPT-4 making it take the questions for an Economics exam the other day and it was missing a lot of context that…

It's not a mistake. It's just ambiguous - you and the OP (rather than "humans") share some contextual assumptions that to you make it imply one meaning. If someone asked me with no context to 'be william gibson pitching your next book' I might answer just the same as gpt did

It's a fair point, and some of the context that's missing for the LLM is just present in the background for humans. You wouldn't need to be told, it would be whether you were in an acting class or creative writing class (these are weak examples).

We get so many context clues that aren't present in current LLM text interactions.

Re: Show HN: AI Playground by Vercel Labs

#78

Post by jozvolskyef »

neom wrote: post #17

I dunno what Cohere is, but it's answers are super weird. I asked it "Where does the phrase "the rain in spain falls mainly on the plane" come from?" and it said: Does anybody know a green song? What happens if a girl doesn't have a twin brother? What musical instrument did larry black sing 'jealous heart' on? Is country music more country than country music. I'm confused.? What is the difference between a copyright…

You didn't tell it to act as an assistant so it just assumes that your input is some list of questions that it's supposed to complete.

Re: Show HN: AI Playground by Vercel Labs

#79

Post by palmdeezy »

zurfer wrote: post #73

Thanks for setting up a limited free version. The UX is great. I had some failures for huggingface and cohere. Replicate produced an output that was not rendered (but returned an answer in the network tab). It would also be great to add: https://docs.aleph-alpha.com/docs/introduction/luminous/

Investigating now. Thanks for the feedback!
Post reply on HN