[Introduction to New Wisdom]After GPT-4o, ChatGPT received another update. This time, the data analysis capabilities have been taken to a new level, and will support the uploading of online documents from Google and Microsoft, as well as real-time interaction and customized charts. More importantly, netizens have already tested the new model behind it in grayscale.
ChatGPT is stronger…
Just now, OpenAI has released another big move – ChatGPT can directly open online data files to complete real-time data analysis.
New enhancements, including:
– Upload the latest versions of files directly from Google Drive and Microsoft OneDrive
– Interact with tables and charts in new expandable views
– Customize and download diagrams for use in presentations and documents
Now, ChatGPT can support single/multiple document uploads, and data can be analyzed by writing Python code.
No matter how complex the data task is, it can easily handle it, such as merging and processing data sets and creating charts to provide data insights.
“A consulting analyst somewhere in the world just shed a tear of happiness.”
Some netizens said, “OpenAI's move will kill many startups again.”
This powerful capability of ChatGPT is said to be achieved with the support of the new model.
Also today, netizens have posted about the new model ADA V2 (GPT-4) they tested in grayscale, and its coding capabilities are very powerful.
It has only been three days since the GPT-4o press conference. OpenAI continues to surprise and shock everyone, and a more powerful “GPT-5” may be waiting for us.
Some free users have grayscale tested GPT-4o
Flowers, the whistleblower, said, “GPT-5 will add a brand new model that no one has considered so far.”
Microsoft and Google’s “Office Family Bucket” provides one-click analysis of file uploads
How is ChatGPT's data analysis capabilities?
Add files from Google Drive and Microsoft OneDrive
In the past, if you wanted ChatGPT to analyze data, the only entrance was to upload local documents on your computer.
At this time, you need to download the file locally and then upload it to ChatGPT.
A single file is fine, but if there are hundreds of files, it will be troublesome.
Now, people can associate Google Drive and Microsoft OneDrive with ChatGPT. After successfully logging in, they can select online files.
Including Microsoft Excel, Word, PPT, and Google Sheets, Docs, and Slide, all can be uploaded with one click.
As shown in the demonstration below, a Google account is associated.
Then, select a form file in your online files, and then you can ask ChatGPT for any information.
Process forms in real time
Additionally, ChatGPT creates an interactive table when you upload your dataset.
You can then select any part of the table to interact with ChatGPT.
For example, select the “Category” column in the table below and let ChatGPT find out which category in the table costs the most.
Then, ChatGPT immediately analyzes the results and can also create a visual table.
In addition, you can also select data from ChatGPT's suggested prompts for more in-depth data analysis.
Customize presentation-ready charts
The third new ability is to customize charts for presentation.
In other words, you can let ChatGPT generate bar charts, line charts, pie charts, scatter charts, etc., and interact with them.
For example, ChatGPT was asked to “chart retention rates by group and write a few brief bullet points.”
ChatGPT then produces the chart shown.
You can also choose any color for the curve labels in the settings in the upper right corner and download the chart for use in presentations.
The last thing everyone is worried about is the data privacy issue. OpenAI said that they will not use the data of the ChatGPT team and corporate customers for training. ChatGPT Plus users can choose to turn off the data for training through the “data control” option.
The new model ADA V2 is in grayscale testing
Just a few days after getting started with GPT-4o, ADA V2 (GPT-4) appeared again, and the entire Internet was looking for clues about this new model.
Some netizens said that ADA stands for Advanced Data Analysis V2, which is the main reason that makes ChatGPT's data analysis powerful.
Someone also analyzed the code and concluded that the gpt-4-ada-v2 model has 128k context, is marked as “confidential”, and appears to be an internally developed prototype.
This appears to be an unexpected release.
Maybe gpt-4-ada-v2 is still an embedded model.
Training LLM using Reddit data
OpenAI’s data landscape has expanded again.
A few days ago, OpenAI just announced its cooperation with Stack Overflow, but it encountered opposition from community members.
Some people even want to delete all their answers one by one.
Early this morning, OpenAI once again announced its cooperation with Reddit, another well-known AI community, and is expected to use Reddit content to train ChatGPT.
The OpenAI official blog also found a very high-level reason for its use of data:
Reddit is a large and active community that has long been an important space for conversation on the Internet.
Keeping the Internet open is critical, and part of that means making Reddit's content accessible to those who are committed to human learning and research on how to build community, belonging, and empowerment online.
Not only that, using LLM, ML and AI technology can help Reddit improve the user experience.
This cooperation will bring many benefits to the user community:
– OpenAI will bring improved Reddit content to ChatGPT and new products to help users discover and participate in the Reddit community.
To do this, OpenAI will access Reddit's Data API, which provides real-time, structured and unique content from Reddit. This will allow OpenAI's artificial intelligence tools to better understand and surface Reddit content, especially content about the latest topics.
“In short, I want to use your data, clean and filter it, to train my own model.”
– The partnership will also enable Reddit to bring new AI-based features to Redditors and moderators. Reddit will be based on OpenAI's artificial intelligence platform to realize its powerful vision.
“I use my AI capabilities to exchange data with you, so that I can empower Reddit with new AIn capabilities.”
– Finally, OpenAI will become an advertising partner of Reddit.
Brad Lightcap, chief operating officer of OpenAI, said, “We are very excited to work with Reddit to enhance ChatGPT with unique timely and relevant information and explore the possibility of enriching the Reddit experience with AI-driven features.”
At the same time, the co-founder and CEO of Reddit also expressed his views on this cooperation:
Reddit has become one of the largest open archives of real, relevant human conversation on the Internet, about anything and everything. Its inclusion in ChatGPT aligns with our belief in connectivity, helping people more easily find the information they want and helping new users find community on Reddit.
Some netizens said that Google’s visits to Reddit have now increased by 500 million.
Google is now flooded with more and more Reddit search results, ranking in the top 3. This is part of the deal to get more users to use Reddit, so Google gets more data.
Now, all this data must be handed over to OpenAI.
I wonder how the AI community researchers reacted to this this time?
References:
https://openai.com/index/improvements-to-data-analysis-in-chatgpt/
https://x.com/OpenAI/status/1791227287569932368
https://x.com/OpenAI/status/1791205420142670250
https://x.com/nanulled/status/1791232438334566783