OpenAI, the company behind the viral chatbot ChatGPT, has released a new version of its artificial intelligence (AI) chatbot — GPT-4.
GPT-4 is a large language model (LLM), a neural network trained on massive amounts of data to understand image and text inputs and generate text outputs.
On Twitter, OpenAI CEO Sam Altman described the model as the company’s “most capable and aligned” to date, designed to follow human ethics. But “it is still flawed, still limited, and it still seems more impressive on first use than it does after you spend more time with it,” he wrote in the tweet.
New features – How is GPT-4 different from GPT-3.5?
“GPT-4 is more reliable, creative, and able to handle much more nuanced instructions than GPT-3.5,” the San Franciso-based company said In a blog post.
The new version outperforms its predecessors at tasks that require reasoning and problem-solving. OpenAI tested both GPT-3.5 and GPT-4 on a variety of tests designed for humans, including a simulation of a lawyer’s bar exam, the SAT, and the GRE for college graduates. GPT-4 achieved human-level scores on many tests and consistently outperformed its predecessor. It passed a “simulated bar exam with a score around the top 10% of test takers; in contrast, GPT-3.5’s score was around the bottom 10%.”
OpenAI says the system can read both text and image inputs but responds in text. The company claims that the successor to GPT-3.5, the model behind ChatGPT, can interpret and output longer blocks of text: more than 25,000 words at once.
In a livestream demo of GPT-4, OpenAI co-founder and president Greg Brockman showed some fascinating uses, including the ability to transform a hand-drawn mockup of a website into a functional site in a matter of seconds.
GPT-4 limitations and risks
- It is not fully reliable, can “hallucinate facts and makes reasoning errors”, and hence strictly requires human review and additional context.
- GPT-4 generally lacks knowledge of events that have occurred after September 2021 and does not learn from its experience.
- It can sometimes make simple reasoning errors and can also be confidently wrong in its predictions.
- GPT-4 poses similar risks as previous models, such as generating harmful advice, buggy code, or inaccurate information. However, it incorporates additional safety and has a lower tendency to respond to sensitive requests such as medical advice and self-harm).
How to access GPT-4?
Currently, GPT-4 is available to ChatGPT Plus subscribers ( $20 per month) who can access it on chat.openai.com with a usage cap.
You could also access GPT-4 with the API, but you have to sign up for the OpenAI waitlist.
How to use GPT-4 for free?
There is no waitlist on Microsoft Bing, so now anyone can sign up and use the GPT-4-powered chatbot. To access it, go to bing.com/new and hit the “join waitlist” button. Just sign in with your Microsoft Account and immediately access GPT-4.
