Microsoft has received an AI license. Scientists consider this model “very dangerous”

Microsoft has been licensed for a language model that generates the most realistic texts in the world. Now they can harness the power of AI in their services.

Microsoft has announced an exclusive license for the OpenAI GPT-3 text generation model. It can generate very realistic text and can be embedded in any application or service. The company previously invested $1 billion in a San Francisco-based startup and became an exclusive partner of OpenAI. Now Microsoft will have access to technology and will be able to use it in its services.

OpenAI released the GPT-3 model for a limited group of developers, but the researchers “highly appreciated its capabilities.” The company noted that the model can not only generate automatic text, it is also capable of more creative tasks – AI can write poetry, computer code, or create memes.

That being said, the model is so powerful that OpenAI initially refused to publish even a limited version of the previous GPT-2 model. Scientists feared it could be used to write realistic false news. Despite its huge size and power, the model can be used by almost anyone.

First, OpenAI collects huge amounts of data from the Internet and analyzes it for stylistic patterns that allow it to realistically predict which letters or words will be written next.

When a user feeds a word, a phrase, or a longer piece of text into the generator, he improves the tool. The results can be used to create stories, dissertations, play chess, solve math problems, or create text scripts.

Experts warned that the system generates texts very realistically, and scammers can use it to publish fake news. The researchers originally announced the system in February 2019 but declined to publish the full version of the program. They feared that it would be used to spread fake news, spam, and disinformation.

Tags: