OpenAI appears to be on a roll ever since the launch of its ChatGPT endeavor.
And now, it’s launched a GPT-4 which is the newest version or shall we say the next-generation variant of the famous chatbot that gained viral fame.
A recently published blog post went on to detail more about how it would feature some better offerings including enhanced performance. But out of all the offerings, the biggest one that really caught our attention included the tool answering questions that combine text as well as graphics.
GPT-4 would function over a wider range of domains right now and entails so many documents with both text and pictures as well as screenshots. It has a lot of similar capabilities to the previous version that only used to deal with text-only inputs.
Moreover, it can be argued further how this new product can be further augmented using test-time procedures that were produced solely for the purpose of text-only language models.
But it’s not all fun and games. The bad news here has to do with image inputs not being available for use by the general public as it’s in the mode designated research preview. And as far as when it’s going to be up for grabs for all others, well, there’s no definite word on that.
Instead of the usual ChatGPT with the classic tone, style, verbosity, and more- developers will be getting the chance to avail the AI style and task by taking on directions seen across the system’s message. The latter allows all API users to customize the entire user’s experience within limits.
In the same way, the blog is showing off some great performance charts that showcase the mega improvements at stake. Moreover, there happens to be one case where OpenAI gives the chatbot plenty of simulated exams.
They claim to be using the most advanced tests of them all and they didn’t even make use of any form of specific training either for such exams. So the results published showed how GPT-4 did much better in terms of exam results when compared to its previous counterpart.
As can be seen with the older version of the app, this tool is again limited in terms of getting the latest data regarding various topics. And that is due to the fact that it can only go back as far as 2021.
Moreover, OpenAI claims this new version really does a great job at reducing hallucinations as compared to its previous counterparts. Yet, we can’t call it one that’s fully reliable.
For now, this new offering is only up for grabs to selected developers. They can gain access to it after signing up for it through a waitlist. Meanwhile, the general public may check this out by signing up for ChatGPT Plus which comes at a price tag of $20 each month.
There are even some reports about OpenAI launching a new subscription tier that paves the way for greater levels of access.
Read next: ChatGPT Smashes Records: Surpasses 1 billion Visits in February 2023
And now, it’s launched a GPT-4 which is the newest version or shall we say the next-generation variant of the famous chatbot that gained viral fame.
A recently published blog post went on to detail more about how it would feature some better offerings including enhanced performance. But out of all the offerings, the biggest one that really caught our attention included the tool answering questions that combine text as well as graphics.
GPT-4 would function over a wider range of domains right now and entails so many documents with both text and pictures as well as screenshots. It has a lot of similar capabilities to the previous version that only used to deal with text-only inputs.
Moreover, it can be argued further how this new product can be further augmented using test-time procedures that were produced solely for the purpose of text-only language models.
But it’s not all fun and games. The bad news here has to do with image inputs not being available for use by the general public as it’s in the mode designated research preview. And as far as when it’s going to be up for grabs for all others, well, there’s no definite word on that.
Instead of the usual ChatGPT with the classic tone, style, verbosity, and more- developers will be getting the chance to avail the AI style and task by taking on directions seen across the system’s message. The latter allows all API users to customize the entire user’s experience within limits.
In the same way, the blog is showing off some great performance charts that showcase the mega improvements at stake. Moreover, there happens to be one case where OpenAI gives the chatbot plenty of simulated exams.
They claim to be using the most advanced tests of them all and they didn’t even make use of any form of specific training either for such exams. So the results published showed how GPT-4 did much better in terms of exam results when compared to its previous counterpart.
As can be seen with the older version of the app, this tool is again limited in terms of getting the latest data regarding various topics. And that is due to the fact that it can only go back as far as 2021.
Moreover, OpenAI claims this new version really does a great job at reducing hallucinations as compared to its previous counterparts. Yet, we can’t call it one that’s fully reliable.
For now, this new offering is only up for grabs to selected developers. They can gain access to it after signing up for it through a waitlist. Meanwhile, the general public may check this out by signing up for ChatGPT Plus which comes at a price tag of $20 each month.
There are even some reports about OpenAI launching a new subscription tier that paves the way for greater levels of access.
Read next: ChatGPT Smashes Records: Surpasses 1 billion Visits in February 2023
