What are the main differences between GPT-3 and GPT-4?

On the 14th of March the OpenAI officially introduced the new version of GPT language artificial intelligence. GPT-4, which is more improved than the old version, GPT-3, is not limited to recognizing text commands. Understand the main differences between the two versions.

GPT-4 understands visual information, although it has not yet expanded to video. The user can upload an image and the AI ​​will respond. With the integration of text and visuals, GPT-4 can revolutionize several industries, such as advertising, design and e-commerce.

ADVERTISING

The new version of GPT also has better memory than its predecessor. The machine can remember up to 25.000 words, which allows the model to generate more coherent and contextually relevant responses. 

GPT-4 can also identify and answer questions in 26 languages, up from 24 in the old version. To get a sense of the power of the new Chat, it can understand languages ​​such as Latvian, Welsh and Swahili. The update is also smarter than its last version, the AI ​​passed 20 of the most common knowledge tests, including mathematics, physics and chemistry.

ChatGPT-4: new version has old problems
ChatGPT-4: new version has old problems

Another positive point of the new version of ChatGPT is that it is less likely to generate harmful/sensitive content. The machine has been trained on a large number of malicious text prompts, including hate speech, cyberbullying and trolling content, so it can more easily identify possible dangerous commands. 

ADVERTISING

GPT-4, like its previous version, cannot be called perfect. Its knowledge is still limited to September 2021. Another fact is that for some functions AI is still naive.

ChatGPT, Dall, Midjourney: What is an AI? | Newsverso Glossary
ChatGPT, Dall, Midjourney: What is an AI? | Newsverso Glossary
Scroll up