Wordsmiths, get nervous.
OpenAI, the artificial-intelligence outfit backed by Elon Musk, recently opened up access to GPT-3, a new AI language model.
It’s so powerful that it can produce text that’s practically indistinguishable from human work.
Psh. Robots can’t write
That used to be true. People fed books and scripts to these programs and generated hilariously bad endings to episodes of Game of Thrones.
But while earlier models mimicked a human’s vocabulary and writing style, GPT-3 is able to analyze context. One guy used GPT-3 to write an entire blog post, and he was surprised by the quality of the results.
And that’s not all GPT-3 can do
How do I get my hands on this?
The API currently is in beta, and you have to request access to use it. When the API becomes commercially available, OpenAI will use the proceeds to fund further research.
We at The Hustle will be nervously watching for future developments. Last week, Microsoft cut dozens of full-time MSN staffers in favor of news curation that’s handled by AI.
Sam, look away.
Trung’s take
We asked Trung Phan, one of our Trends analysts, what he thought of this powerful new toy. Here’s what he told us — not aided by GPT-3 (or was he?).
In the classic sci-fi movie Terminator 2, the villainous AI Skynet becomes self-aware at 2:14am ET on August 29th, 1997.
That prediction may have been 23 years too soon, but here we are. *nervous laughing*
More seriously, the potential threat of general AI is grounded in something more (seemingly) benign than killer robots (read: The Paperclip Maximizer article and — for a deeper dive — this book).
The purpose of OpenAI is to ensure the development of AI “benefits all of humanity.” Fingers crossed.