4 ms·
Googles: Custom versions of Google. Anyone can easily build their own Google. No coding is required. You can make them for yourself, just for your company’s in
by nojvek 3y ago
Googles: Custom versions of Google.
Anyone can easily build their own Google. No coding is required. You can make them for yourself, just for your company’s internal use, or for everyone. Creating one is as easy as starting a conversation, giving it instructions and extra knowledge, and picking what it can do, like searching the web, making images or analyzing data.
The whole point of ChatGPT was go to one single place for all your knowledge needs.
The whole point of Amazon is the largest collections of things you can buy and have it delivered to your doorsteps in a few days.
I don't want many GPTs. I want one GPT that can reliably digest all information available on the internet, understand it, organize it and allow me to do useful things with it.
It's the same enshittifaction on Whatsapp that Meta is doing with Celebrities AI like SnoopDog AI that are gimmicks.
Please don't build gimmicky features. Leave that to the community via integrations.
- jes5199 3y agofor a while, adding a custom google search to your website was considered a great feature
- BoorishBears 3y agohttps://search.brave.com/goggles/create https://search.brave.com/goggles/create https://developers.google.com/custom-search/v1/overview https://developers.google.com/custom-search/v1/overview https://learn.microsoft.com/en-us/bing/search-apis/bing-custom-search/overview https://learn.microsoft.com/en-us/bing/search-apis/bing-cust...
- vunderba 3y agoThey still have that - it's just regular GPT-4. One immediate application about this one is that it makes it trivial to create a fine tuned version of GPT based on your data, where you can upload a series of documents that can basically act like a set of embeddings that augment the regular GPT trained data.
- nomel 3y ago> create a fine tuned version of GPT No. It's unclear fine-tuning is happening. Many are guessing it's RAG.
- vunderba 3y agoMy bad, I wasn't intending fine tune to mean fine tuning the model itself. And leveraging vectorized embeddings as I mentioned IS FUNCTIONALLY a rag-style architecture.
- gustavopch 3y agoBeing able to have multiple personae could be very useful. One persona may not give you the answer you're looking for, but another one may. I think maybe they should require the GPTs to have human names though so people intuitively understand that. Like, Paul can't help me with this task, so let me ask Monica instead. They could even interact.
- nomel 3y agoA good example is creative/idea work vs fact work. You don't want creative facts, and you don't want fact bounded creativity. You either have a prompt ready to paste, to prime the conversation/context, or you can use a personality. One "über" AI is great, but it requires guidance into the context you're interested in, including yourself. For example, the default ChatGPT will assume you're uneducated about any topic you ask about. I think this all fits perfectly into what Sam Altman talked about in the Lex Friedman podcast: people want an AI that fits their own worldview/context. Custom instructions, and "about yourself" are good starts, but sometimes you want to talk to a chef, and sometimes a scientist.
- gfodor 3y ago> The whole point of ChatGPT was go to one single place for all your knowledge needs. That's just your own perception. OpenAI is trying to build AGI. You entered into the storyline at a specific junction and jumped to conclusions based on the limits of your own imagination, or something.
- nojvek 3y agoRight. All I'm sayin is I want the God level AGI, the king of kings of AGI. Allowing me to customize ChatGPT is the same itty bitty intelligence, like putting a new mask on ChatGPT. OpenAI changed their values to 'AGI Focus'. This seems like OpenAI losing focus.