- OpenAI has announced GPT4, the latest version of its large language model that powers key applications like ChatGPT and the new Bing.
What is ChatGPT?
- ChatGPT is an artificial-intelligence chatbot developed by San Francisco-based AI research company OpenAI in 2022.
- It is a trained model which interacts in a conversational way. The dialogue format makes it possible to answer followup questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.
- It can have conversations on topics from history to philosophy, generate lyrics and suggest edits to computer programming code.
- The technology that underlies ChatGPT is referenced in the second half of its name, GPT, which stands for Generative Pre-trained Transformer.
- Transformers are specialized algorithms for finding long-range patterns in sequences of data.
- A transformer learns to predict not just the next word in a sentence but also the next sentence in a paragraph and the next paragraph in an essay. This is what allows it to stay on topic for long stretches of text.
Shortcomings of the Previous Models
- The chatbot isn’t always accurate: Their sources aren’t fact-checked, and they rely on human feedback to improve its accuracy. They can also get facts mixed up and produce misinformation.
- GPT-3 and ChatGPT’s GPT-3.5 were limited to textual input and output, meaning they could only read and write.
- GPT-3 and GPT-3.5 only operated in one modality, text, meaning users could only ask questions by typing them out.
What is GPT-4?
- GPT-4 is a large multimodal model which means it can encompass more than just text and accepts images as inputs.
- GPT-4 also exhibits human-level performance on various professional and academic benchmarks.
- For example, it can answer tax-related questions, schedule a meeting among three busy people, or learn a user’s creative writing style.
- GPT-4 is also capable of handling over 25,000 words of text, opening up a greater number of use cases that now also include long-form content creation, document search and analysis, and extended conversations.
How is GPT-4 different from GPT-3?
- GPT-4 can see images now: GPT-4 can be fed images and asked to output information accordingly whereas GPT-3 and ChatGPT’s GPT-3.5 were limited to textual input and output.
- GPT-4 can process a lot more information at a time.
- GPT-4 has an improved accuracy: GPT-4 significantly reduces hallucinations relative to previous models and scores 40 per cent higher than GPT-3.5 on factuality evaluations.
- Hard to trick: It will be a lot harder to trick GPT-4 into producing undesirable outputs such as hate speech and misinformation.
- Better at understanding languages other than English: GPT-4 is more multilingual which means that users will be able to use chatbots based on GPT-4 to produce outputs with greater clarity and higher accuracy in their native languages.