GPT-4: How to use, new features, availability, and more
“Recently, ChatGPT has gained widespread popularity, with one million users in its first week of launch, a faster adoption rate than popular platforms like Instagram and Facebook.”
ChatGPT-4 has been formally announced, confirming long-rumored enhancements to OpenAI’s ChatGPT’s already extraordinarily remarkable language skills.
The latest version of OpenAI’s powerful AI represents the next step in what the company’s technology is capable of, from passing a variety of tests to taking instructions from images.
According to the company’s release, the technology still has limits.
Here are some critical facts about GPT-4.
Availability
GPT-4 was formally revealed on March 13, as Microsoft had previously indicated, however the actual date was unknown. Nevertheless, it is currently only available as part of the ChatGPT Plus premium membership.
The current free version of ChatGPT will continue to be based on GPT-3.5, which is less accurate. GPT-4 will also be offered as an API “allowing developers to build applications and services.” Duolingo, Be My Eyes, Stripe, and Khan Academy are among the companies that have already implemented GPT-4.
The first public demonstration of GPT-4, showcasing some of its additional capabilities, was also live-streamed on YouTube.
What’s new in GPT-4?
GPT-4 is a new language model developed by OpenAI that can generate text that is comparable to human speech. It advances the technology utilized by ChatGPT, which is presently based on GPT-3.5. GPT stands for Generative Pre-trained Transformer, a deep learning system that employs artificial neural networks to write like a person.
According to OpenAI, this next-generation language model is more sophisticated in three crucial areas: originality, visual input, and lengthier context.
In terms of creativity, OpenAI claims that GPT-4 is far superior at both producing and interacting with people on creative projects. Examples include music, scripts, technical writing, and even “understanding a user’s writing style.”
GPT-4 accepts visual inputs
One of the most noticeable differences between GPT-4 and the GPT-3.5 that powered ChatGPT when it was released to the public in November is that it allows visual inputs in addition to text.
Users can now ask inquiries based on picture queries thanks to this upgrade.
The chatbot is shown an image of a few baking materials and asked what may be baked with them in the example supplied on the GPT-4 website. It is still unknown whether video can likewise be utilised in this manner.
GPT-4 is acing exams
The performance of OpenAI’s technology on exams such as the SAT and Advanced Placement-level tests was already generating headlines. The new version outperforms its predecessor.
GPT-4 passed a simulated bar exam with a score in the top 10% of test takers, whereas GPT-3.5 passed with a score in the bottom 10%.
GPT-4 earned 710 out of 800 on the SAT reading and writing exam, 40 points more than GPT-3.5. GPT-4 earned 700 on the SAT math exam, up 110 points over GPT-3.5.
GPT-4 received a 5 on AP Biology, up from a 4 on GPT-3.5. One of the most notable variances on tested AP examinations was in AP Calculus BC, where GPT-4 received a 4, a huge improvement over GPT-3.5’s 1.
Yet, GPT-4 continues to struggle in other testings. GPT-4 and GPT-3.5, for example, received just 2 on the AP English Language and AP English Literature tests.
GPT-4 can change its personality
GPT-4 improves the AI’s “steerability,” allowing it to change its behavior on command.
“Rather of the classic ChatGPT personality with a defined verbosity, tone, and style, developers (and soon ChatGPT users) can now prescribe their AI’s style and task by providing such directions in the ‘system’ message,” OpenAI added.
For example, when set as a “socratic tutor” and a user asked how to solve a linear equation, GPT-4 responded by saying, “Let’s start by analyzing the equations. Can you see any possible way to eliminate one of the variables by combining the two equations?” rather than spitting out the result.
Limitations
Although detailing GPT-4’s additional possibilities, OpenAI also mentions some of the new language model’s limitations. According to OpenAI, the current model, like earlier versions of GPT, has issues with “social biases, hallucinations, and aggressive prompts.”
OpenAI also admitted that the model may have “different biases in its output,” a point expressed by opponents when generative AI technologies penetrate the public sector.
In other words, it’s not ideal, but according to OpenAI, these are all concerns that the business is attempting to resolve.
Does Bing Chat use GPT-4?
According to Microsoft, the new Bing, or Bing Chat, was more powerful than ChatGPT. Because OpenAI’s chat employs GPT-3.5, there was speculation at the time that Bing Chat would use GPT-4.
Microsoft has recently verified that Bing Chat is built on GPT-4. Nevertheless, capabilities like visual input were not available on Bing Talk, so it’s unclear which features have been integrated and which have not.
Nonetheless, Bing Chat has certainly been enhanced with the ability to obtain current information over the internet, which is a significant increase over the present version of ChatGPT, which can only draw on the training it received until 2021.
Aside from internet connectivity, the AI model utilized for Bing Chat is substantially faster, which is critical when brought out of the lab and incorporated to a search engine.
Thank you for reading!
If you found this article useful, you can follow me for new ones every week. If you have any further questions or would like to discuss further, please leave a comment below.
