GPT-4 unveiled: OpenAI’s latest language model takes creative leap

OpenAI’s new model – GPT-4 – is capable of carrying out more complex instructions, has advanced reasoning capabilities and is, according to its creators, more creative.

During the roll-out period, OpenAI will be capping usages of GPT-4 to 100 messages every four hours.

OpenAI’s GPT-4

When I asked the GPT-4 model how it would describe itself, the chatbot suggested I use this as a headline: OpenAI Unveils Groundbreaking GPT-4: A New Era in Artificial Intelligence.

It also wanted me to say: “Building on the success of its predecessor, GPT-3, this cutting-edge artificial intelligence (AI) system takes natural language processing (NLP) and understanding to unparalleled heights.”

Well, there you have it, straight from the horse’s mouth; or should I say, the chatbot’s cortex.

GPT-4 versus ChatGPT-3

OpenAI says the latest model draws from a larger training dataset and has advanced algorithms enabling it to better understand context, nuance and semantics in human communication.

It is also supposedly better at conversation and can engage “in more in-depth, coherent and context-aware” discussions, making it “an invaluable tool for applications requiring human-like interaction”.

Most notably, however, is the upgraded knowledge base.

Even though the knowledge cutoff is still capped at September 2021, GPT-4 can now contextualise information for a more accurate interpretation.

GPT-4 key features

Key features include multilingual support, reduced bias and increased ethical considerations.

“OpenAI has worked diligently to address biases and ethical concerns by implementing improved techniques and guidelines for minimising unintended consequences.”

With all its shiny new features, GPT-4 is poised to “revolutionise various industries and applications, from content creation to customer service”.

If you are worried about AI taking your job, there are measures to make yourself indispensable. But nonetheless, I’ve strayed from the point.

So, how could GPT-4 be used? Glad you asked.

Chatbot applications

The new language model is perfect for content creation, including blogs, social media and SEO optimised content.

OpenAI said it could also produce accurate news articles, but we haven’t put that feature to the test yet.

The new model could also be used as a virtual assistant for businesses, since it is capable of “handling a broad range of customer inquiries and issues”.

It even has educational applications, such as creating personalised tutoring tailoring to individual needs “for a more effective learning experience”.

If you’re a game creator, GPT-4 will assist you with enhanced storytelling and dynamic world-building.

RELATED: Yo Eskom, this is how ChatGPT says AI could solve load shedding

Ethical issues and biases

OpenAI said it spent the past six months “making GPT-4 safer and more aligned”.

In its latest iteration, the chatbot is 82% less likely to respond to requests for disallowed content and 40% more likely to produce factual responses.

It’s courteous, too:

That said, the company admits there are still “many known limitations” which they are working to address, such as:

• Social biases,
• Hallucinations,
• Adversarial prompts.

I’ve only had a few hours to play around with GPT-4 and will share my thoughts in due course.

The only downside to the current model is the slower response time on Plus.

I’ve grown accustomed to GPT 3.5’s lightning fast response, but I digress.

For now, GPT-4 is available to Plus members. Alternatively, users interested in taking it for spin can join the API waitlist.

RELATED: ChatGPT now available in Microsoft’s Azure OpenAI service