You’re summoned to court where a ton of evidence including phones calls, handwritten notes, emails and chat logs undeniably prove your complicity in a crime.
The problem is, none of that evidence belongs to you. They have been masterfully crafted by Artificial Intelligence algorithms that have meticulously profiled you.
This is not an excerpt from Blade Runner. It’s the here and now. Developments in Artificial Intelligence are helping revolutionize many fields such as conversational commerce. But the same technology can serve as a tool to invade privacy and commit acts of fraud.
Here are some of the ways fraudsters may put AI to ill use in the future.
In the old days, imitating hand writings and signatures was a feat that required skill and practice. Not anymore, according to an AI algorithm developed by researchers at University College London (UCL).
Titled “My Text in your Handwriting,” the algorithm only needs a paragraph’s worth of script to learn a person’s handwriting. It can then write any text in the person’s handwriting. This is the most accurate replication of human script to date.
The innovation has positive uses such as helping stroke victims formulate letters without the concern of illegibility. It can also help in translating comic books while preserving the author’s original writing style.
However, evil actors can also take advantage of the technology. Given its accuracy, it can become instrument in forging legal and financial documents—or maybe changing history. The researchers were able to reproduce the handwriting of as Abraham Lincoln, Frida Kahlo, and Arthur Conan Doyle.
The researchers claim that forensics experts could tell the difference. But that will become harder as the software develops and becomes more advanced.
Chatbots are finding their way into more and more domains. And thanks to artificial intelligence, they’re succeeding in providing increasingly natural experiences. But what happens when they become too real?
Last year, messaging app company Luka created a chatbot that impersonated the cast of HBO’s Silicon Valley. The app’s neural networks ingested the script from the first two seasons of the show to learn the characters’ language patterns. It then created bots that talked like the actual fictional characters.
Two seasons’ worth of dialogue is not enough to create an efficient chatbot, but the idea behind it was very real.
A few months later, the company used the same technique to virtually bring the dead back to life. By feeding the algorithm with a history of text messages, social media conversations and other sources of information, Luka’s engineers succeeded in creating a chatbot based on the company’s deceased co-founder.
Luka wants to create bots that mimic real-life people. And with newer generations of people creating even more digital content, that goal is becoming achievable. Such chatbots can have some very productive uses — as long as they are within your control.
But fraudsters can put the same technique to malicious uses. For instance, spear phishers usually spend weeks and months to learn and mimic the habits of their targets. Will they use AI and machine learning as a shortcut?
TNW recently ran a report about Lyrebird, an AI company that synthesizes speech in anyone’s voice with a one-minute recording. The samples published on the company’s website are rather rudimentary.
Google’s Wavenet provides a similar functionality. It requires a much bigger data set, but it sounds eerily real. The technology behind it is, as you guessed it, neural networks.
The point is, the technology is advancing at an accelerating pace. And as Lyrebird’s founders warn, copying the voice of someone else is possible and audio recordings might no longer be a trusted source of evidence.
When put together, voice, handwriting and conversation forgery can do an awful lot of good — or evil. We might be heading toward an era where guarding your every bit of data will become critical.
This post is part of our contributor series. The views expressed are the author's own and not necessarily shared by TNW.