GPT-4 is the latest advance of OpenAI to scale up deep learning. GPT-4 is a large multimodal model (accepting image and text inputs, emitting text outputs) that exhibits human-level performance in various professional and academic markets.

OpenAI announced that GPT-4 is available to paid ChatGPT+ subscribers and within the OpenAI API, which has a waitlist.

On the first day after it was unveiled, GPT-4 impressed many users in early tests with its ability to write court cases, pass standardized exams, and build a functional website from a hand-drawn sketch.

GPT-4, released a week ago, is an updated version of OpenAI’s language model, which is trained on large amounts of online data to generate complex responses to user requests. It is now available through a waiting list and is already present in some third-party products, including Microsoft’s new search engine – Bing. Some users with early access to this tool are sharing their experiences and highlighting some of the most compelling uses.

Five things GPT-4 can do

Analyze more than text

The biggest change in GPT-4 is its ability to work with photos that users upload.

A user, for example, uploaded an image of a website design and then copied the resulting code into a preview window that showed how it could become a functional website.

Another example was a picture of the inside of a refrigerator, that the New York Times showed to GPT-4 and asked it to come up with a meal based on those ingredients.

Programming just got even easier

GPT-4 can write code in all major programming languages, according to OpenAI and also similar models can be used to create marketing content around game previews, generate news articles, and even moderate game discussion forums.

Passing exams with distinction

Although OpenAI said the upgrade is “less capable” than humans in many real-world scenarios, it exhibits “human-level performance” on various professional and academic tests. The company said that GPT-4 recently passed a simulated US Bar exam with a score around the top 10%.

Providing more accurate answers

Compared to the previous version, GPT-4 is able to deliver longer, more detailed and more reliable written answers, according to the company.

The latest version can now give answers of up to 25,000 words, previously it was about 4,000.

Rationalizing work in various industries

Joshua Browder, CEO of legal services chatbot DoNotPay, said his company is already working on using the tool to generate “one-click lawsuits” to process calls made by robots, in a first indication of GPT-4’s vast potential to change the way people work in different fields. Imagine someone calls the office and Imagine that someone calls the office and the voice message is transcribed and a 1,000-word lawsuit is immediately generated.

What the future holds

Although OpenAI has made major improvements to its AI model, GPT-4 has similar limitations to previous versions. That is, it lacks knowledge of events that occurred after its database deadline (September 2021) and does not learn from experience. It can also make “simple errors in reasoning” or be “overly naive in accepting obviously false statements from a user,” and does not do a double-check job.

For now, OpenAI said that GPT-4 users should be cautious using the tool, particularly “in high-risk contexts.”