OpenAI Withholds Powerful Text Generator Over Misuse Fears
OpenAI declines to release its highly advanced GPT-2 text generator due to concerns about fake news and spam. The decision sparks backlash from researchers who accuse the organization of breaking its open research promises.
OpenAI unveils a powerful new natural language model called GPT-2, but the non-profit refuses to release the full version to the public. The system trains on 40 gigabytes of internet text to predict the next word, allowing it to produce coherent and convincing paragraphs that easily adapt to the user's chosen style and topic.
While the technology shows promise for improving dialogue bots and speech recognition, OpenAI worries about malicious applications like generating fake news, impersonating real people, and automating spam. For example, when researchers prompt the AI with a statement opposing recycling, the system immediately outputs a highly persuasive, flawlessly written essay arguing that recycling harms the environment and the economy.
This unprecedented capability leads OpenAI to release only a scaled-down version of the model, a decision that sparks immediate controversy. Critics accuse the organization of breaking its founding promise to share its research openly, highlighting a growing tension between the advancement of artificial intelligence and the need to prevent dangerous misuse.