昨天阅读1小时，累计1122小时
Two researchers(Joseph Bullock, Miguel Luengo-Oroz) at Global Pulse, an initiative of the United Nations, showed how easy now it is for anyone to forge convincing misinformation with machine learning.

In paper \"Automated Speech Generation from UN General Assembly Statements: Mapping Risks in AI Generated Texts\", they used only open-source tools and data to show how quickly they could get a fake UN speech generator up and running. They used a readily available language model that had been trained on text from Wikipedia and fine-tuned it on all the speeches given by political leaders at the UN General Assembly from 1970 to 2015.
(paper1:What Drives the International Development Agenda? An NLP Analysis of the United Nations General Debate 1970-2016
paper2:Regularizing and Optimizing LSTM Language Models)
Thirteen hours and $7.80 later (spent on cloud computing resources), their model was spitting out realistic speeches on a wide variety of sensitive and high-stakes topics from nuclear disarmament to refugees.

(Wikitext-103 Dataset，fast.ai ，AWD-LSTM model, NVIDIA K80 GPU，13 hours,$7.80)