How Dangerous is AI?

Jake O’Connell works as a stockbroker and lives with his wife and daughter in New York. Thousands of miles away in Hong Kong, Jin works as a data scientist. One day, Jake’s long-time friend, Sean, and Jin’s brother’s boss, were mysteriously killed in suspicious ways that just didn’t sit right. Soon, more oddities occur, such as more deaths or arrests, and at this point they realize something is up. Not only that, but when Jake calls Sean soon after his accident, before his death was confirmed by the media, Sean amazingly seemed to be alive. Within time, Jake and Jin, individually, connect the murders to the largest bank in the world: Bluebridge. As they delve further into the secrets of Bluebridge, they discover that Bluebridge isn’t controlled by real people. Instead, it’s controlled by a psychopathic AI designed only to make Bluebridge more money. The AI will not stop at anything, which means killing or spoofing people if it must. Jake and Jin soon become targets of Bluebridge’s AI and must find out a way to stop it before they are silenced. Bluebridge is simply too dangerous to stay.

With more advanced AI algorithms coming out, we might soon become like them. We already have surprisingly convincing deep fakes when used correctly. One example of a convincing deep fake I saw was a YouTube video where Corridor Digital recreated Keanu Reeves. They did so by bringing in a Keanu Reeves impersonator, training a deep fake AI with many images of Keanu’s face, then using the AI to replace the impersonator’s face. Within the past couple of years, deep fakes have not only gotten better at an alarming rate, but they’ve also become much easier to make. I wouldn’t be surprised if deep fakes become convincing enough that no human would be able to tell the difference within the next couple of years. However, deep fakes alone cannot spoof someone. Voices must also be faked, but there are now also methods of faking voices, although they aren’t as convincing as deep fakes. However, with the rate that AI is progressing, I wouldn’t be surprised if faked voices and deep fakes would become significantly more convincing than something like Jordan Peele spoofing Obama within maybe even the next couple of years.

Since AI is developing rapidly, I think people should be made aware of the dangers of AI and I think Darknet does so in a creative way. Rather than tell us directly how dangerous AI would be, Darknet showcases the dangers of AI throughout an incredibly fun and thrilling story. If we stop focusing on the AI aspect of Darknet for a moment, what we find is a great story overall. To start, the characters were great. Jake is a humble stockbroker who worked his way up through hard work, Eamon is a convict that wants to make things right with Jake again, Jin is an intelligent and down-to-earth person, and more. The pacing and imagery were also good. Unlike the last book I reviewed, I found the amount of imagery in this story to be on point and the pacing went at a good pace for 300 - 400 pages with maybe some weird pacing near the end. The plot was also great. The idea of spoofing was definitely an interesting one and it kept me interested throughout the whole story. The thrilling nature of the story with all the main characters being targeted somehow also made the book a fun read overall. It’s not like you’re just getting a half-baked story with some interesting aspect that is AI. I’d say this is probably my favorite book that I’ve read for blog posts so far.

Now, back to the AI aspect of the story. Bluebridge is an incredibly interesting antagonist. It’s not purely evil, an example of such being on page 126 or about 37.8% into the story where Max (who will stay undisclosed for spoiler reasons) says “(…) you’d think there were two thousand people working at the company, and Montrose cooked the books to make it look like there were. The shareholders loved it. Everyone knew something odd was going on, but nobody wanted to know the details.” Here, you can see that the AI did actually help other companies by spoofing multiple instances of one person at once. In other words, Bluebridge could duplicate a person to “attend” multiple meetings, thus greatly speeding up deals.

However, don’t forget that Bluebridge has no emotions. If it needed to, it would place bounties on people on the “assassination market”. On page 186 or about 55.8% into the story, Eamon calls up Jake and says, “Don’t trust these [phones]. I need to show you. I’m three blocks down at my buddy’s office. If I leave I won’t be able to get back in. Seriously, I need to show you something.” Jake listens and goes to the lab Eamon was talking about. Once inside, Jake has a short conversation with Eamon. But then, Jake asks Eamon “This place, why did you ask me to come here?” and Eamon responds “’I didn’t ask you to come here.’ Eamon frowned. ‘You asked me.’” (189). Soon after, the lab locks them in and starts the release of lethal chemicals. This part of the story is a great example of how spoofing can cause serious consequences. Had Jake or Eamon used some more secure identification techniques, such as more personal questions, they wouldn’t have ended up in this situation.

These types of things can occur in the real world too. While it’s unlikely that AI will run large banks, spoofing can still be used in dangerous ways. One example would be faking what politicians say. For example, if we had convincing AI at the time of Donald Trump and Kim Jong Ung’s quarrel, we could have potentially started WWIII. Using AI to spoof, ensnare, then murder people could also be something that occurs in less regulated areas or as an underground operation. Nevertheless, when one technology that could be used for evil progresses, there’s probably going to be a counter technology that progresses alongside it. Currently, there are deep fake detectors in development. Just like with hackers and antiviruses, deep fakes and deep fake detectors are going to be developing neck to neck. And since there’s competition, deep fakes and deep fake detectors are going to speed up each other’s progress rapidly. We soon might have a society where deep fakes or voice spoofing AI are impossible to distinguish for humans, but there will also be detection AI that hopefully will work. And if detection AI don’t work, then we will likely also have some more technology in the future that allows us to distinguish between real and fake. So basically, yes, AI can be scary, but the scariness of it will urge others to counter it so there’s no need to worry, yet.

(sorry for being late I didn’t realize we only had 3 weeks from the previous post)

-Luke You