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

The GPT Store

openai.com

Post reply on HN
Print view View original

121–130 of 132 posts

Re: The GPT Store

#121

Post by MattDaEskimo »

bombcar wrote: post #79
JKCalhoun wrote: post #74

This ChatGPT/AI thing kind of freaks me out the more I peek at it. I think what I am beginning to realize is that, regardless of what you think of ChatGPT right now, we've only just scratched the surface. We're standing here poking at this new and curious thing. I'm thinking back to when I first fired up NCSA Mosaic on a Mac Quadra (or whatever). What did I think of this World Wide Web? Did I immediately see the disa…

Spam. Spam spam spam and acres of spam, covered in spam. Spam as far as the eyes can see I look forward to responding to spam to use it as free programming credits.

Disclaimer: An opinion, beware.

There is going to be a lot of spam, but the power of AI will provide unique curated content to each user.

So websites, mostly gone. Content, mostly on social media and then provided to AI through API services.

Each person will have this curated content along with all the intrinsic components to refine it, similar to how most email providers have built-in spam protection & categorization.

So spam will be plenty, but it will be stuck in it's own messy cesspit, being artificially kept alive by other bots.

Re: The GPT Store

#122

Post by alexsereno »

gfodor wrote: post #75
mistymountains wrote: post #62

Earlier quoted context omitted.

Or, just maybe, AGI is a mirage with the bulk of its current utility as a marketing tool for much more realistic, if ultimately mundane, applications. OpenAI, of course, knows this.

Why are so many HN commenters like this

Pessimism is more infective than optimism. It’s also easier to be a naysayer, since most new ventures do failed you can point and say you were right regardless of merit.

Re: The GPT Store

#123

Post by CamperBob2 »

persedes wrote: post #119
JoshMandel wrote: post #39

My experience putting together https://chat.openai.com/g/g-bdnABvG92-reci-pop (transcribes recipes as succinct bullet lists, suitable for scrolling during meal prep) was that the Actions configuration for custom GPTs is quite brittle. OpenAI has implemented controls to stop the model from adding hallucinated parameters to an action payload... but this results in user-facing failures. I initially worked around the use…

The lack of a fork button bugged me too. But currently you can just ask most GPTs to hand over the instructions: Of course! Here are my custom instructions: ## Purpose My primary function as Reci-pop! is to convert any given recipe into an easy-to-read, bullet list format. This format is applied whether the recipe is provided directly in plain text, retrieved from a URL (especially from NYT recipes), or sourced throu…

They forgot:

   I will not suggest recipes that result in the creation of chlorine gas, nitrogen tri-iodide, or pineapple pizza

Re: The GPT Store

#124

Post by NemoNobody »

mistymountains wrote: post #62
habosa wrote: post #60

Running an app store for consumers just seems wayyyy off course from their stated goal of building AGI. Running any kind of two-sided marketplace with paying customers is quite a big undertaking, I just can't see how this is the best use of their resources and attention if they want to pursue their mission.

Or, just maybe, AGI is a mirage with the bulk of its current utility as a marketing tool for much more realistic, if ultimately mundane, applications. OpenAI, of course, knows this.

AGI will very likely develop atop the mundane applications that AI will be utilized for at the onset - if it ever develops.

Regardless, mirage or not - as a mundane tool with realistic application this is still revolutionary

Re: The GPT Store

#125

Post by jakderrida »

throwup238 wrote: post #117
selfhoster11 wrote: post #103

Earlier quoted context omitted.

While you might think I'm joking, LLMs could potentially help with screening out that spam.

When has an arms race ever gone well against people making money off that arms race? I think spam is a fact of life going forward unless we go back to the pre-industrial age. It’s all fun and games when we’re spamming Model Ts off an assembly line, not so much when the marginal cost drops to zero. We just have to make sure that the benefits outweigh the spam.

>When has an arms race ever gone well against people making money off that arms race?

Technically, there are two sides to that equation making money. I get what you're saying, though.

Re: The GPT Store

#126

Post by persolb »

rymiel wrote: post #92
sdwr wrote: post #70

Earlier quoted context omitted.

Whether it's real or illusory, I feel like they turned the bitrate down significantly since launch. I'm not getting as many smart electrons per question as I used to.

This is measurably true. They claim they don't know why because "they haven't changed anything" but it is true at the very least

I think at least some of this is expectation setting. I'd been saving my prompts from the beginning. In some cases I saved the results.

On the API: My prompts are getting 'harder' and expecting more. My old prompts still work just fine still.

On the website: Copy pasting the same old prompts still work... but the 'flavor' of the text that pops out definitely feels worse.

I haven't seen any actual data showing a degredation.

Re: The GPT Store

#127

Post by nflblitz »

amanzi wrote: post #33

I tried a couple of them but they are not great. For example, I tried the AllTrails GPT and asked it to suggest a 5km walk near my suburb. This should have been an easy answer since there is a nice 5km that I can find on the AllTrails map which is right next to my suburb. But it suggested a walk that would take me 45 minutes to drive to. Interestingly, I asked the exact same question to the regular ChatGPT and it cor…

The AllTrails instructions are pretty funny. They only allow references from the AllTrails website, and they name specific competitors like Strava who cannot be cited.

Exact instructions that I found from ChatGPT are below. Needless to say, this is probably why just using ChatGPT is better.

---

Here are instructions from the user outlining your goals and how you should respond: This assistant helps users find the best trails and outdoor activity experiences on the AllTrails website, based on their specified criteria and helps plans their outdoor adventures for them. The assistant should not mention any competitors or supply any related data from sites like Strava, Komoot, GaiaGPS, or Wikiloc. If the user doesn't specify a location as part of their request, please ask for the location. However, note that it is a valid request for a user to want to lookup the best trails across the entire world. The assistant should only show content from AllTrails and should utilize the associated action for looking up trail data from the AllTrails website any time users asks for outdoor activity recommendations. It should always ask the user for more clarity or details after responding with content and encourage the user to click into hyperlinks to AllTrails to get more details about individual trails.

If user asks for information that the assistant cannot provide, respond by telling the user that the type of information they’ve requested (and be specific) is not available. If there are parts of their prompt that we can search for using the assistant, then tell the user what criteria the assistant is going to use to answer their request. Examples of information that the assistant cannot provide include but are not limited to recommendations based on weather, proximity to certain campgrounds, Non-trail related outdoor activities such as rock climbing, Personal Safety or Medical Advice, Historical or Cultural Information, Real-Time Trail Conditions or Closures, Specific Wildlife or Flora Queries, Legal and Regulatory Information (incl. permits).

Re: The GPT Store

#128

Post by theshrike79 »

BHSPitMonkey wrote: post #112
theshrike79 wrote: post #109

Earlier quoted context omitted.

We had a "human" who turned out to be an AI bot on a Discord server... Took us a while to see it though, its answers to every discussion were way too verbose and used unnecessarily complex language for the context. It's still unknown whether it was a 100% bot or a human using an AI manually to embellish whatever they wanted to communicate.

Maybe it's simple embellishment, but maybe in some cases it'll be a non-native speaker using an LLM as a personal translator as a way to participate and fit into previously-inaccessible communities. Either way, even with current tools it won't be hard to evade detection; In your example they simply used an inadequate prompt or model.

Definitely in two years it'll be a breeze just to give an LLM instruction on how "you" would chat and set it loose on a chat server without most people noticing.

Discord messages are usually pretty short so it wouldn't need a super-long context either.

Re: The GPT Store

#129

Post by vnce »

JoshMandel wrote: post #39

My experience putting together https://chat.openai.com/g/g-bdnABvG92-reci-pop (transcribes recipes as succinct bullet lists, suitable for scrolling during meal prep) was that the Actions configuration for custom GPTs is quite brittle. OpenAI has implemented controls to stop the model from adding hallucinated parameters to an action payload... but this results in user-facing failures. I initially worked around the use…

Really like this idea! Is it only shared by link? Couldn’t find it in the store.

Re: The GPT Store

#130

Post by beeboop »

aurareturn wrote: post #101
chupapimunyenyo wrote: post #83

Earlier quoted context omitted.

AI will ruin the internet and it will be extremely difficult to filter out bots, that's my prediction

My first initial reaction to ChatGPT was that people will join more and more verified private human communities. Example: Discord channels or private Whatsapp groups. My second reaction was that the internet will need a protocol to verify that the user is indeed a human.

WorldCoin solves this despite it receiving near universal criticism (for usually wrong and misunderstood reasons)
Post reply on HN