OpenAI has published the text-generating AI it said was too dangerous to share (2 minute read)
OpenAI has released the full version of GPT-2, a text-generating AI system. When it was first announced, there were concerns that it could be misused to spread fake news, spam, and disinformation. OpenAI has since released smaller versions of GPT-2, but has seen no strong evidence of misuse. GPT-2 can generate coherent text from minimal prompts, for example, it can create an article from just a headline. Its limitations include having trouble with long-term coherence, for example, sticking to a single subject in a news article. There is an ongoing debate with AI researchers about the ethics of releasing cutting-edge AI tools that can enable malicious actors. OpenAI has created systems that can spot GPT-2's output with around 95 percent accuracy. A link to a web version of GPT-2 is available.
In 2020, Some Americans Will Vote On Their Phones. Is That The Future? (4 minute read)
The US is experimenting with Internet voting as a means to increase turnout. While there are some concerns, advocates claim that the project is necessary as the current voting system is outdated. Each state has the freedom to choose how voting is conducted, including choosing what voting technologies are used. Some states have been experimenting with voting apps for overseas and military voters. The main concern for online voting is security. Voting technology is usually owned by private companies and kept secret, and this lack of transparency may decrease trust in the voting process. The difficulty of the current voting system has been blamed for low voter turnout. Mobile phone voting will encourage more people to vote, and also help those who can't physically vote.