Restrictive GPT-4 | OpenAI | Page 1

gentle briar Feb 6, 2024, 8:49 PM

#

Has anyone noticed how restrictive GPT 4 is lately? It will not pull data from external sources that are publicly available on websites. I’ve tried it for multiple different queries and I’m always met with the same general responses. I’m not sure if this is due to increased litigation against Open AI, websites giving the model a 404 error because they are trying to block web scraping, or if this is open AI preemptively handicapping the models abilities.

It will often tell me it can’t access the web, even though I’m using GPT 4. It won’t give me a YouTube video summary but if I use a GPT like YouTube video summarizer, I get a response. Or just today I was trying to get it to create me a graph of FDI (foreign direct investment) into different countries businesses and GPT 4 knew exactly where to get the data from but it wouldn’t access the data and make me the graph. Instead it told me where to get the data, to download the data and then it made me a graph with “placeholder data” to show me what it would look like. This is just one of many examples though.

Curious if anyone else has noticed this and if you’ve found some solution (prompt engineering or otherwise) to circumnavigate this.

sharp rampartBOT Feb 6, 2024, 8:49 PM

#

Hi @gentle briar. I'm kapa.ai, your friendly support bot. I will try to answer your question by searching through the knowledge base. :wave:

#

Based on the information provided in the extracts, GPT-4, like its predecessors, does not have direct access to the internet or external databases. It generates responses based on the patterns it learned during its training phase and does not have the ability to pull in real-time data from external sources or websites. This is confirmed in a forum post (https://community.openai.com/t/468615) where a support user clarified that GPT-4 is an inference model and any access to other systems has to be implemented by the services that drive this model.

As for the issue of creating a graph of FDI, GPT-4 is a language model and does not have the capability to create graphs or visualizations. It can guide you on how to create one, but it cannot create one itself.

Regarding the YouTube video summary, it's possible that the GPT-like YouTube video summarizer you mentioned has been specifically trained or programmed to summarize YouTube videos, which is why it can provide a response.

In terms of solutions, you could consider using GPT-4 in conjunction with other services or tools that can access the internet or create visualizations. However, you would need to implement this yourself or use a service that provides this functionality.

Please note that these are the limitations of the model at its current stage of development and improvements may be made in future versions.

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

#

@gentle briar kapa.ai is still learning and improving, please let us know how it did by reacting below

gentle briar Feb 6, 2024, 8:53 PM

#

@sharp rampart That's not entirely correct. GPT-4 does have access to the internet via Bing search. Why does the model not use it's search and retrieval function when a command is given to search for information contained in a public database, retrieve the data and then perform the function such as creating a graph as it was asked to do in the original prompt?

sharp rampartBOT Feb 6, 2024, 8:53 PM

#

Let me check the knowledge base for you. One moment! :books:

#

I'm sorry for any confusion, but based on the information provided in the extracts, GPT-4 does not have direct access to the internet, including Bing search. It's an inference model that generates responses based on the patterns it learned during its training phase. It does not have the ability to pull in real-time data from external sources or websites, nor does it have the capability to create graphs or visualizations. This is confirmed in a forum post (https://community.openai.com/t/468615) where a support user clarified that GPT-4 is an inference model and any access to other systems has to be implemented by the services that drive this model.

As for the issue of creating a graph of FDI, GPT-4 is a language model and does not have the capability to create graphs or visualizations. It can guide you on how to create one, but it cannot create one itself.

Regarding the YouTube video summary, it's possible that the GPT-like YouTube video summarizer you mentioned has been specifically trained or programmed to summarize YouTube videos, which is why it can provide a response.

In terms of solutions, you could consider using GPT-4 in conjunction with other services or tools that can access the internet or create visualizations. However, you would need to implement this yourself or use a service that provides this functionality.

Please note that these are the limitations of the model at its current stage of development and improvements may be made in future versions.

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

#

@gentle briar kapa.ai is still learning and improving, please let us know how it did by reacting below

gentle briar Feb 6, 2024, 8:57 PM

#

@sharp rampart this information is incorrect. GPT-4 DID create a graph for me with 'placeholder data' and was aware of where the actual data was at. Use GPT vision to view the attached screenshots.

*Hopefully a human dev will review this.

sharp rampartBOT Feb 6, 2024, 8:57 PM

#

Let me check the knowledge base for you. One moment! :books:

#

I apologize for any misunderstanding, but based on the information provided in the extracts, GPT-4 does not have the capability to create graphs or visualizations. It's a language model that generates responses based on the patterns it learned during its training phase. It does not have the ability to pull in real-time data from external sources or websites. This is confirmed in a forum post (https://community.openai.com/t/468615) where a support user clarified that GPT-4 is an inference model and any access to other systems has to be implemented by the services that drive this model.

However, GPT-4 can generate text that describes how a graph or visualization could be created, including placeholder data. If you're seeing a graph, it's possible that the service you're using to interact with GPT-4 is interpreting the model's output and creating the graph based on that.

As for GPT-4 knowing where to get the data, the model can suggest sources based on the information it was trained on, but it can't access or retrieve data from those sources in real-time.

Regarding your mention of GPT Vision, it's a separate model that can understand images, but it doesn't have the ability to create graphs or visualizations either, according to the OpenAI documentation (https://platform.openai.com/docs/models).

I hope this clarifies the capabilities of GPT-4 and GPT Vision. If you have further questions or need more information, I recommend reaching out to OpenAI directly or consulting their documentation and resources.

#Restrictive GPT-4

Please note that these are the limitations of the model at its current stage of development and improvements may be made in future versions.

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

Please note that these are the limitations of the model at its current stage of development and improvements may be made in future versions.

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

I hope this clarifies the capabilities of GPT-4 and GPT Vision. If you have further questions or need more information, I recommend reaching out to OpenAI directly or consulting their documentation and resources.

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon: