Source: Deep Learning on Medium
Did you hear about the latest breakthrough?
The researchers at OpenAI built a machine learning app that can generate coherent paragraphs of text and achieves state-of-the-art performance on reading comprehension, machine translation, question answering, and summarization.
This is huge, because it means we can finally build AI’s that can write novels, analyze reports, chat with humans, browse and understand the web, write news articles, and hundreds of other roles that require reading and writing skills.
Check this out, here’s a sample prompt fed into the network and the corresponding AI answer:
Prompt: Legolas and Gimli advanced on the orcs, raising their weapons with a harrowing war cry.
Response: The orcs’ response was a deafening onslaught of claws, claws, and claws; even Elrond was forced to retreat. “You are in good hands, dwarf,” said Gimli, who had been among the first to charge at the orcs; it took only two words before their opponents were reduced to a blood-soaked quagmire, and the dwarf took his first kill of the night. The battle lasted for hours until two of the largest Orcs attempted to overwhelm Aragorn. When they finally stopped, they lay defeated and lifeless for miles and miles.
“I take nothing,” said Aragorn. “But I give my word, at my peril and mine, that I will never forget this day of horror. None of us will forget. Ever!”
You can see that the app correctly picked up a ‘Lord of the Rings’ theme from the input prompt and just ran with it. And the story stays reasonably coherent until the very end, which is remarkable for an AI.
In fact, this app is so good at generating text that OpenAI is keeping it under wraps for now. Malicious users could easily use it to smear political opponents, or create fake amazon reviews, or write fake news articles.
Just look at this one:
Prompt: Miley Cyrus was caught shoplifting from Abercrombie and Fitch on Hollywood Boulevard today.
Response: Shoplifting: Miley Cyrus was caught shoplifting from Abercrombie and Fitch on Hollywood Boulevard today (pictured). The singer was wearing a black hoodie with the label ‘Blurred Lines’ on the front and ‘Fashion Police’ on the back
The singer was also wearing a pair of black-rimmed glasses, a black jacket, black jeans and black sandals. She was carrying a pair of black and white striped gloves and a small black bag.
That looks somewhat believable, doesn’t it? And the AI even understands that celebrity magazines tend to obsess over what Miley is wearing.
Now imagine thousands of articles like this, all designed to bring Miley down. You could even train a GAN to generate thousands of fake pictures to go with the fake articles.
Suppressing this technology is probably not the answer because it’s going to get out eventually.
A better solution in my opinion is to add digital signatures to pictures, news articles, and comments, so that it’s always clear to us who the author of a piece of media is.
And then we can decide which authors we want to trust.
What do you think? Do you believe we should release this AI to the world?
Write a comment and let me know!