Have you ever played along with one of those posts on Facebook where you type in a few suggested words then use your phone’s suggestion feature to complete the sentence? The results can be nonsensical, surreal, mundane, and/or hilarious. Last week, the non-profit artificial intelligence research company OpenAI announced that they had built a neural network language model, GPT-2, that is so good at this game that OpenAI felt it would be irresponsible of them to make it available to others. About 50% of the time, GPT-2 can generate text that mimics the style and content of its input sample, as well as successfully automate question-answering, summarization, translation, and other language-based tasks.
- OpenAI built a text generator so good, it’s considered too dangerous to release [TechCrunch] “With every good application of the system, such as bots capable of better dialog and better speech recognition, the non-profit found several more, like generating fake news, impersonating people, or automating abusive or spam comments on social media.”
- The AI Text Generator That’s Too Dangerous to Make Public [Wired]
“Clark says OpenAI hopes that by voicing its concerns about its own code, it can encourage AI researchers to be more open and thoughtful about what they develop and release. ‘”
- Artificial Intelligence Is Getting Good at Fake News [Bloomberg Opinion] “Stringing together a few formulaic passages from a set of numbers is a mechanical job. Inventing a fake news story on a random subject requires imagination; not every human is up to it.”
- An AI helped us write this article [Vox] “This is the future, and the field of AI is going places we could only have guessed at.”
From the Ohio Web Library:
- Knight, Will. “Reinforcement LEARNING.” MIT Technology Review, vol. 120, no. 2, Mar. 2017, pp. 32–35.
- Miller, John W. “The Creeping Ethical Challenges of Artificial Intelligence.” America, vol. 219, no. 11, Nov. 2018, p. 20.
- Boden, Margaret A. “But Is It Intelligence, Really?” Artificial Intelligence: A Very Short Introduction. Oxford University Press, 23 August 2018, pp. 106-129.