By Jude Huck-Reymond

Artificial Intelligence is evolving at a rate that is hardly possible to comprehend by the human mind. By now, you’ve probably heard of the capabilities of OpenAI’s ChatGPT which is built on their Artificial Intelligence called GPT3. If you have created an account and played around with it, you probably have an idea of its abilities and the utility it can provide.

OpenAI is rumored to release GPT4 during the week of March 12th, 2023, about the time this article is published. Here, we’ll be discussing the capabilities of GPT3 and then comparing them to the potential capabilities of GPT4.

The Abilities of GPT3

GPT-3 stands for Generative Pre-trained Transformer 3, which is a language model developed by OpenAI. It was released in June 2020 and is currently one of the most advanced AI language models available. GPT-3 has a massive number of parameters, with the largest version having 175 billion parameters, which is several times larger than its predecessor, GPT-2.

The AI uses unsupervised learning techniques to analyze and process large amounts of text data, allowing it to generate human-like text, answer questions, and perform a variety of language-related tasks. GPT-3 has been trained on a wide range of text data, including web pages, books, articles, and more, allowing it to generate text that is often coherent, grammatically correct, and semantically meaningful.

Some of the abilities of GPT-3 include language translation, summarization, answering questions, content generation, chatbot development, and more. It can also be used for a variety of creative applications, such as generating poetry, writing stories, and creating music.

In general, GPT-3 works by analyzing input text and generating output text based on its understanding of the language and context. It uses a neural network architecture known as a transformer, which is capable of processing long sequences of text and capturing complex relationships between words and phrases.

Overall, GPT-3 is a powerful tool that has the potential to revolutionize many aspects of natural language processing and AI. Its vast range of abilities and high level of accuracy makes it a valuable resource for developers, researchers, and businesses in many different fields.

The Potential of GPT4

While GPT3 is already multimodal, meaning its fluent in multiple inputs such as text, code, and creative language, GPT4 will expand the modes by which users can input commands. GPT4 will potentially be able to analyze and manipulate images, video, and audio, while improving the already existing modes in GPT3.

For example, you could give GPT4 an image and it could describe to you in words what is in the image, or it could manipulate the image into something completely different depending on what the user wants. It could also generate entirely new images based on user input such as text and speech. You could say “give me a picture of a blue banana on a plate with a chimpanzee” and it would give you its best estimation of what that scene may look like.

Likewise, it could use video input such as a movie or show and give descriptions of the plot, characters, or themes. You could also potentially ask it to generate a video of its own based on the input you give it.

As for audio, it will be revolutionary for an AI of this caliber to be able to recognize human speech. GPT4 could become the receptionist at every business, hostess at every restaurant, or fluently speak to customers on support lines.

Comparing GPT3 and GPT4

Both of these models are built on a vast set of data using parameters to interpret inputs and generate outputs. GPT3 uses around 175 billion parameters, while GPT4 could use anywhere from 1 to 20 trillion. It is fair to say that the quality of the AI is very directly proportional to the number of parameters it is able to use. So GPT4 could be around 100x more useful than GPT3.

The most significant differences are the modes by which input can be used and which output can be generated. GPT3 is only text-based, using language, code, and creative text input to generate language and code output. Meanwhile, GPT4’s addition of visual and auditory modes will expand the utility of the technology to many walks of life, especially realms like entertainment, business, and social media.

Conclusions

Ultimately, it is quite hard to imagine the entire scope of how this technology can be used and how it will continue to evolve. I have heard rumors that GPT4 will likely be used to write the parameters used by GPT5. If that is the case, then we are only at the beginning of a steep exponential growth curve tracking the intelligence of this technology. GPT3 was released in mid-2020 while GPT4 will be released in early 2023. Based on Moore’s law, GPT5 could be upon us as early as late 2024. The iterations will only become faster over time, and eventually, we could be updating our devices with the new GPT software weekly or daily.

How long will it take before humans give up control of this growth entirely? How fast can this technology iterate upon itself? What does this development mean for human civilization?