OpenAI says its new model GPT-2 is too dangerous to release (2019)
Why it matters: OpenAI's decision to withhold GPT-2's full release highlighted the immediate challenge of responsibly developing powerful AI.
- OpenAI developed GPT-2, a language model capable of generating coherent and versatile prose, but released only a smaller version due to safety concerns.
- News outlets like Metro U.K. and CNET published alarming headlines, suggesting GPT-2 was so powerful it needed to be locked up for humanity's good.
- Machine learning experts debated whether OpenAI's claims about GPT-2's danger were exaggerated, sparking a broader discussion on handling potentially dangerous AI algorithms.
OpenAI's 2019 announcement of its powerful text-generation model, GPT-2, sparked both alarm and debate, with the organization withholding the full algorithm due to "safety and security concerns." While headlines from outlets like Metro U.K. and CNET sensationalized its danger, experts in the machine learning field questioned whether OpenAI's claims were exaggerated.
