I just thought you should know, that both GPT 3.5 and GPT 4 give far more desirable and to the point, responses when web search is off.
This does not make sense, much. Supposedly a current web search would give the model , more data , for more accurate responses. This is far from every day reality.
In fact responses with web search enabled are so bad, that I would rather do the web search myself than bother asking GPT.
On the contrary , responses without web search enabled are awesome especially with the latest GPT 4 . GPT 3 has some hiccups sporadically.
This I address as a question and also publish as a feedback.
They do not. That’s an extension (WebChatGPT), not ChatGPT native product. It literally says it there. The way this extension responds to your queries totally depends on the developer of it. It’s not OpenAI’s product.
Plus Bing Chat is not GPT 3.5. It’s GPT-4. They literally stated it yesterday.
Extension or not, there is a way for it to communicate with the web. And it does not make results any better. Plus , as far as I tested bing chat, it produces about the same reponses as chat gpt 3.5 when connected to the internet and the same responses as gpt 3.5 when you specifically ask from it not to search.(that is , it also makes the same mistakes).
So, at least until now, I don’t know what they stated, but GPT-4 is far superior in its responses(when not connected to the web) and bing chat looks like GPT-3.5 , actually dumbed down a bit compared to GPT 3.5 .
So, for me personally, right now it is either GPT-4 or personal human search of the web.
Yeah, that is part of the issue, I am sure you will agree @prodromosregalides
@AgusPG is both technically correct in his replies to you and he has read the OpenAI docs; so he is simply trying to help you understand something you do not understand.
RC et al - Thanks for your posts. I think I’ll try “talking” with GPT-4 (re: my subjects of interest) without Bing’s “AI” help. Why? Although Bing-AI is very nice for ordinary web-search on questions about “best apps for…” (etc.), I like too like my results best, re: complex sociocultural, psychological, bio-ethical and political subjects. After all, Bing-AI still lacks sufficient training (and knoledge) to simulate actual intelligence and real wisdom.
BTW, RC, would you like to help develop an ethical geo-political AI-Advisor app (with an appropriate database, K-base, inference engine, etc.)?
