OpenAI Launches Its Latest Multimodal AI Model GPT-4

OpenAI continues its advancement in deep learning by releasing its latest milestone GPT-4.

March 15, 2023

While the buzz around ChatGPT generative AI platform is still around, OpenAI recently unveiled a strong new image and text comprehending AI model, GPT-4, which they call “the latest milestone in its endeavor to scale up deep learning.”

GPT-4 is a large-scale, multimodal model that can take inputs of both text and images and generates text outputs. A transformer-based model, GPT-4, has been pre-trained to anticipate the next token in a text. The process of post-training alignment leads to better performance on tests of factual accuracy and adherence to intended behavior.

OpenAI’s paying customers can currently access GPT-4 through ChatGPT Plus (with a utilization limit), and developers can join a waitlist to gain access to the API. GPT-4 is expected to perform at the “human level” on various professional and academic benchmarks, generate text, and accept both text and image inputs, an upgrade over GPT-3.5, which only accepts text.

As per the reports, 1,000 “prompt” tokens (about 750 words) cost $0.03, and 1,000 “completion” tokens cost $0.06. (again, approximately 750 words). The word “fantastic” would be broken up into the tokens “fan,” “tas,” and “tic” to represent raw text. While the content produced by GPT-4 is known as completion tokens, prompt tokens are the word fragments supplied into GPT-4.

For instance, GPT-4 successfully completed a mock bar exam with a score in the top 10% of test takers, while GPT-3.5 received a score in the lowest 10%.

“To understand the difference between the two models, we tested on a variety of benchmarks, including simulating exams that were originally designed for humans,” the company wrote in a blog post. “We proceeded by using the most recent publicly-available tests or by purchasing 2022–2023 editions of practice exams. We did no specific training for these exams. A minority of the problems in the exams were seen by the model during training, but we believe the results to be representative.”

See More: How Can AI Developers Avoid Ethical Landmines?

The business claims to have spent six months iteratively aligning GPT-4 utilizing learning from an adversarial testing program and ChatGPT, notably how people engaged with the chatbot. The company is supported by a $10 billion Microsoft investment and off the success of ChatGPT. The “best-ever results” on factuality, steerability, and refusal to go outside of OpenAI-defined guardrails as a result of this presented challenges for GPT-3.5, the model upon which ChatGPT is based.

“GPT-4 generally lacks knowledge of events that have occurred after the vast majority of its data cuts off (September 2021), and does not learn from its experience,” OpenAI wrote. “It can sometimes make simple reasoning errors which do not seem to comport with competence across so many domains, or be overly gullible in accepting obvious false statements from a user. And sometimes it can fail at hard problems the same way humans do, such as introducing security vulnerabilities into code it produces.”

There is a lot to discover about GPT-4. Nonetheless, OpenAI is moving forward at full speed, clearly confident in its improvements.

“We look forward to GPT-4 becoming a valuable tool in improving people’s lives by powering many applications,” OpenAI wrote. “There’s still a lot of work to do, and we look forward to improving this model through the collective efforts of the community building on top of, exploring, and contributing to the model.”

Let us know if you enjoyed reading this news on LinkedInOpens a new window , TwitterOpens a new window , or FacebookOpens a new window . We would love to hear from you!

Image source: Shutterstock

MORE ON CHATGPT

Shruti Umathe
Shruti Umathe

Senior Associate - Content Operations , Spiceworks Ziff Davis

As a senior associate - content operations at Spiceworks, Shruti manages the industry engagement program covers many topics, including cloud, security, devops, etc. When not at work, she spends her time with her dogs and loves sketching. She is a big Marvel and Harry Potter fan! For editorial queries, please reach out: shruti_umathe@swzd.com
Take me to Community
Do you still have questions? Head over to the Spiceworks Community to find answers.