OpenAI, the capped for-profit company founded by Elon Musk that is developing the massive GPT-3 language model, is unveiling a new version and opening up its model to customization. Developers will be able to use high-level artificial intelligence at a lower cost and with higher levels of accuracy.
The GPT-3 model is becoming increasingly open to customization. Indeed, OpenAI recently announced the possibility of creating custom versions of the language model. This model is capable of generating text or even basic code. Among its possible uses, GPT-3 can summarize texts, create content from basic and simple ideas, translate text to a very good standard or even answer questions using web pages. This model is particularly accurate in its use of human language and can also summarize large texts.
Previously in private beta, OpenAI is now offering developers the opportunity to train the language model themselves by using it. They will be able to fine-tune the algorithm to create models adapted to the content of their application or the service they wish to provide. The undeniable advantage of this would be to obtain better results with greater speed than basic algorithms and with many more possibilities.
"According to Gartner, 80% of technology products and services will be built by those who are not technology professionals by 2024. This trend is fueled by the accelerated AI adoption in the business community, which sometimes requires specifically tailored AI workloads," an OpenAI spokesperson told VentureBeat.
A game powered by artificial intelligence
If Gartner's predictions are accurate, companies could benefit from this new technology to train artificial intelligence models quickly and at an unbeatable cost. GPT-3 is an autoregressive language model that uses deep learning to produce human-like text.
Given any text prompt, such as a phrase or sentence, GPT-3 returns a natural language text complement. Developers can "program" GPT-3 by giving it just a few examples -- or "command prompts." About 300 applications are currently using the model in fields such as industry, productivity, education and gaming. A video game has even been designed based on the model. The game, called "AI Dungeon," is powered by GPT-3 algorithms. From a simple adventure proposal, the language model adapts and improvises around the player's vision of the story. It will generate a virtual game master whose texts will be entirely written by the model to drive the story.
And there are other examples. Fable Studio creates interactive adventures and uses the model to help power story-driven characters, or "Virtual Beings." The studio even received a 2019 Emmy Award in the "Outstanding Innovation in Interactive Media" category for crafting a virtual-reality experience. "GPT-3 has given us the ability to give our characters life," said Edward Saatchi, CEO of Fable Studio on the OpenAI blog. "We're excited to combine an artist's vision, AI, and emotional intelligence to create powerful narratives, and believe that one day, everyone will know a Virtual Being." Language models have a bright future ahead of them. In the future, many applications may well use deep learning to provide users with increasingly advanced services.
