Microsoft Corporation-backed startup OpenAI has begun rolling out GPT-4, a powerful artificial intelligence (AI) model that leverages the technology behind the wildly popular ChatGPT.
GPT-4 is “multimodal”, meaning it can generate content from both image and text prompts.
What is the difference between GPT 4 and GPT 3.5?
GPT 3.5 only takes text prompts, while the latest version of the larger language model can use images as input to recognize and analyze objects in the image.
GPT 3.5 is limited to responses of approximately 3,000 words, while GPT 4 can generate responses of over 25,000 words.
GPT4 is 82% less likely to respond to disapproved content requests than its predecessor and scores 40% higher on certain reality tests.
This will let developers decide the tone and wording of their AI. For example, GPT 4 may assume a Socratic style of discourse and answer questions with questions. Previous iterations of the technology had a fixed tone and style.
OpenAI said that soon ChatGPT users will have the option to change the chatbot’s tone and response style.
What can ChatGPT 4 do?
The latest version has surpassed its predecessors in the American Bar Exam and the Graduate Record Examination (GRE). GPT 4 can also help people calculate their taxes, a demonstration by OpenAI President Greg Brockman showed.
Hand-drawn pencil drawing -> website (https://t.co/4kexpvYAgV).
— Greg Brockman (@gdb) March 15, 2023
Prompt: "Write brief HTML/JS to turn this mock-up into a colorful website, where the jokes are replaced by two real jokes." https://t.co/zQ4smwqGVo pic.twitter.com/cunT74HO5l
The demo showed that it can take a hand-drawn mockup for a simple website and create a real image.
Be My Eyes, an app that caters to the visually impaired, will provide a virtual volunteer tool powered by GPT 4 on its app.
Limits of ChatGPT 4
According to OpenAI, GPT-4 has the same limitations as its previous version and is “less capable than humans in many real-world scenarios”.
False responses known as “hallucinations” have been a challenge for many AI programs, including GPT 4.
OpenAI said that GPT 4 can compete with human propagators in many domains, especially when working with a human editor.
He cited an instance where GPT 4 came up with proposals that seemed plausible, when asked how to make two parties disagree with each other.
Sam Altman, chief executive officer (CEO) of OpenAI, said that GPT4 was “the most capable and aligned” with human values and intent, although “it still has flaws.”
GPT 4 generally has no knowledge of events occurring after September 2021, when the majority of its data was decommissioned. It also does not learn from experience.
Access to GPT 4
Although GPT 4 can process both text and image input, only the text input feature will be available to ChatGPT Plus subscribers and software developers with a waiting list, while the image input capability is not yet publicly available. .
The subscription plan, which offers faster response times and priority access to new features and improvements, was launched in February and costs $20 per month.
GPT 4 powers Microsoft’s Bing AI chatbot and some features on the subscription tier of language learning platform Duolingo.
