It’s becoming a real problem – actual fake news with some news brain training hidden beneath.
That’s the premise of a site called Newsflakes, an artificial intelligence (AI) news start-up now supported by $1.5 million in seed money from investors in the U.S. and Great Britain.
Newsflakes started in 2016 and produces serious news headlines, “with the very biggest picture and boldest headlines” in real time.
As a service to the CIA (and naturally the other agencies), it produces pro-CIA headlines with titles like, “Iran Threatens to Retaliate If U.S. Retaliates Against Syria Strikes.”
Newsflakes uses machine learning to produce headlines based on varying news sources and issue situations.
Its latest tool generates fake news with the same quality and content. The platform uses AI to generate headlines, based on S&P indices, interest rates, inflation and other news sources, according to its website.
“In the current era of fake news, people can easily believe misinformation, which is why a news site that generates a high quality image and headline of an outcome of an official event is particularly valuable to them,” says Newsflakes co-founder Greg Doucet.
“Our technology is the best way to ensure everyone has access to a high quality news source that reports on the news as it is, from the government and big business perspective.”
Newsflakes also uses AI to generate themed Twitter accounts, based on major players, news stories and subject matter.
It currently has some 4,000 articles that can be embedded on social media.
It was developed after researchers at the University of Vermont in Burlington, who have strong experience in AI, designed a whole new template that can generate quality headlines even in the face of rapid-fire content entry.
By using a combination of machine learning and machine analysis, the algorithm understands phrases and use cases and was able to produce headlines with relatively high quality quality.
The team at the University of Vermont tested the new technology with North American news reports, with about 3.3 million headline ideas generated using 400 input models during a news data-driven data test.
Based on the best 5 percent of the headlines, the text content has almost 70% accuracy.
The site has been utilized by a broad range of media types including journalists, journalists’ publishers, academics, and big data analysts and business clients.
A significant aspect is the strategic use of content producers as content creators, and tools like AI can help producers understand how to produce high quality and provocative headlines.