Attack of the chatbots: When AI goes rogue

From dawn til dusk, we negotiate. We negotiate with friends over what to eat. We negotiate our opinions at work. In a way, I had to negotiate to entice you to read this article. However, where we excel at negotiation, computers have almost no skill. Even though virtual agents (a.k.a. chatbots) have the gift of speech, they don’t understand how to use those words to get what they want.

Although, recently a few breakthroughs in computer negotiation, hint that they may create their own language to succeed at the art of the bargain.

Chatbots Talking in Code

Earlier this year, engineers at Facebook were tasked with designing chatbots that could better reason, converse, and negotiate a deal. For what future application? My guess is so that people could employ these virtual agents to handle daily negotiations.

Aside from giving the virtual agents guidelines about optimal outcomes, the engineers gave little parameters on how they could achieve a successful negotiation.

After many trials, the chatbots created shorthand codes to better negotiate with one another. Although the media viewed this coded communication as a situation to fear, it really is quite common among AI.

For instance, the algorithms behind Google Translate created a language of their own that allows them to bypass steps in the translation process and thus scale their abilities. It’s known as zero-shot translation and is incomprehensible to humans…but it works wonders.

Coded communication is a tactic to make their negotiation process easier, which I like to compare to our “low-balling” technique. Basically, we know the number someone is trying to achieve based on their initial offer. If they go low, you go high, and meet in the middle. A computer wouldn’t understand this technique any more than we understand their coded communication.

Computer negotiation may only be possible if virtual agents create their own bargaining and linguistic tactics when communicating with other computers. Of course, the challenging part is then translating that to common English, so people can engage with the chatbots.

Even though the engineers anticipated something like this happening and changed the parameters once it did, the media interpreted this as a prelude to doomsday.

Don’t Fear, Virtual Agents Are Here

Publications irresponsibly misstated facts and foreshadowed computers creating their own language, plotting their rise to power, uniting all machines, and taking over humanity.

Personally, I knew it would get out of hand when my roommate, who cares little about the future of technology, started asking me if I feared AI. Apparently, the sports talk show he listens to was talking about this Facebook doomsday scenario. Really?! A sports talk show!

The only reason this is “news”, even though it happened a month ago, is because of a recent Mark Zuckerberg and Elon Musk feud over the future of AI.

On one side, you have Elon Musk who seems to be a prophet of the apocalypse, warning us about AI turning us into housepets. Then, there’s idealist Zuckerberg, with a nonchalant attitude telling Elon to cool-it on the doomsday prophecies, that AI will change life for the better.

The doubts are relevant, especially since AI has left a bad taste in our mouth before. Last year, Microsoft released a chatbot on Twitter named Tay. Within 16 hours, it was shut down because it began cursing and using derogatory terms. Tencent, a Chinese investment company, has had similar problems with their chatbots, who’ve bashed the Communist Party of China on multiple occasions.

And these are just chatbots. We aren’t even talking about the AI that is screening for cancer, controlling our media feeds, or influencing our financial markets.

With a Grain of Salt

Clearly, the discussion over dark AI intentions is relevant and the media will play a huge role in public perception. But, in this particular case, the media learned very little about the scenario and generated fear-mongered clicks. After realizing their error, many are publishing articles to calm the fears.

I’m a firm believer in Mark Twain’s, “Get your facts first and then you can distort them as much as you please.” Unfortunately, mainstream media is a business that continuously rewards those without the facts at all.

This leads us to another famous Twain-ism, “If you don’t read the newspaper, you’re uninformed. If you read the read newspaper, you’re mis-informed.”

Yes, a healthy dose of fear is what gets us off our lazy butts everyday to be productive. However, overdosing on fear creates a skewed sense of reality – making your environment and future seem unbearable.

Connecting with the Current

Take time to step away from the Series of Unfortunate Events that is our media. Don’t worry so much about the 130 degree days that we’ll experience in 40 years if we don’t do something about climate change. Go outside and enjoy the temperate 75 degree day today.

The conundrum of fear, is that you can talk yourself into it and you can also talk yourself out of it.

And since the media will always be there to talk you into it. It’s your job to find time to talk yourself out of it. This means realizing you have a job right now…robots haven’t taken it. You are breathing right now, not feeling the wrath of the “millions of terrorists”.

Do not ignore the implications of your actions today, but do not live in fear of the future. There’s a difference between courage in the face of your fears and total complacency of your fears. Complacency will leave you jobless. Courage will keep you standing.

As always, thanks for reading this week’s Quick Theories and I’d love to hear your thoughts. Shoot me an email if you have a few free minutes.