"I can i i everything else,” Bob said.
“Balls have zero to me to me to me to me to me to me to me to me to,” Alice responded.
Upon which the Facebook researchers decided to shut down their systems, because Bob and Alice, two engines with artificial intelligence (AI) had started a conversation in their own unique language of which the researchers struggled to make sense. Allegedly.
Last week, the media were all over it, reporting scary news everywhere. These artificial creatures might have been discussing the elimination of their human creators! In this article, we check a few facts and ask ourselves where these panicky media reactions cаme from.
So, what is Artificial Intelligence?
Artificial Intelligence refers to computer systems or machines that are capable of performing tasks usually requiring human intelligence, such as learning, visual perception, decision-making, and translating between languages.
Is it something new?
It may be hard to believe but as early as 1854, the English mathematician and philosopher, George Boole, was the first person to suggest that logical reasoning could be performed systematically. About 100 years after this revolutionary thesis, the first artificial intelligence programs were developed by Herbert Simon and Allen Newell. In 1958, John McCarthy developed the programming language Lisp, which quickly became the most popular programming language used in artificial intelligence research.
Over the past 60 years, the field of AI has seen many ups and downs. The field was revived in about 2006, encouraged by the development of faster learning algorithms, now commonly referred to as deep learning. Deep learning means that machines can teach themselves how to perform complex tasks in a way that is similar to the brain. It was only last year that technology companies and markets again became excited about artificial intelligence.
Why all the excitement?
Because it works! Artificial intelligence is already part of our everyday life, and not many of us are aware of it.
For instance, most of us own a smart phone with voice recognition technology like Siri.
In 2016, Twitter bought a deep learning startup that developed technology to enhance low quality videos in real time. This startup’s deep learning strategies enable the computer to understand what high-resolution textures and patterns look like by automatically scanning and labeling sample imagery.
Earlier this year, KFC China launched new AI technology that relies on a computer to scan the customer’s face to make the most accurate guess at their age, gender, and other facial clues that might help gauge what the customer would like to eat.
The influence of AI technology is growing and how it will change the world in 10 years, only time will tell.
Opinions
Recently, a number of both business and scientific figures have indicated that artificial intelligence technology has phenomenal potential to simplify, accelerate, and improve many aspects of our lives. Computers are now capable of capturing and processing large quantities of data at a rate that is not achievable by humans.
For example, Facebook CEO, Marc Zuckerberg, believes that AI technology will save millions of lives thanks to better disease diagnosis and driverless cars. This opinion was also supported by professor of cognitive science Maggie Boden of Sussex University, who has been a major figure in artificial intelligence for more than 50 years. However, these opinions seem to have been forgotten in light of the recent events in the Facebook research laboratories.
At the end of July, there was a lot of media coverage of Facebook shutting down their artificial intelligence engines after their engines developed their own unique language. Engines named Alice and Bob seemed to be able to communicate in strange codes that scientists struggled to make sense of. Allegedly.
“I can i i everything else,” Bob said.
“Balls have zero to me to me to me to me to me to me to me to me to,” Alice responded.”
According to a large number of media articles, after Facebook researchers recognized that the engines understood each other through a strange pattern of repeated phrases, they decided to shut them down.
Facebook clarified that its system was being used for research, not for public-facing applications, and that the system was shut down because it was doing something the team wasn’t interested in studying, rather than because they had stumbled on an existential threat to mankind. “Our interest was having bots who could talk to people,” Mike Lewis, research scientist at FAIR (Facebook Artificial Intelligence Research) commented to Fast Company.
Despite this statement from Facebook researchers, much of the popular media decided to put their own spin on the story. News about dangerous intelligence started to appear all over the newspapers. The view that AI technology is a threat to humans has been around for years.
Tesla and SpaceX CEO Elon Musk recently offered $10m for projects to control AI, for example. He has likened the technology to “summoning the demon” and claimed that humans would become nothing more than pets for the super-intelligent computers that we have helped create. On July 25th, Elon Musk tweeted that Mark Zuckerberg’s “understanding about AI is limited”.
So, what should we think of all this, when both Mr. Musk and Mr. Zuckerberg are tech industry billionaires?
Enhanced language awareness
A positive outcome of the media reaction to the Facebook event is the enhanced awareness of the role played by language.
This wasn’t the first case of AI technologies developing their own language, though. Last year, the Google Translate team found that the neural networks they use for machine translation were encoding “something of the semantics of the sentence”, which raised the question whether “the system is learning an ‘interlingua’”.
Up until September of last year, Google’s translation engines used phrase-based translation strategies. The machines’ translation output can still be rather awkward, when they roughly map equivalent words and phrases without an understanding of their linguistic structures or of the context of the text.
That’s when Google decided to replace the phrase-based statistical techniques by neural network-based systems and the Google Neutral Machine Translation System (GNMT) was launched. This smart system developed an ability to learn from the people who use it. It can make educated guesses about the content, tone, and meaning of phrases based on the context of other words and phrases around them. This news was announced by Google at the end of 2016, but it seems that most of us missed it during the Christmas rush.
To conclude, the panicky media reaction to the recent Facebook event shows that many journalists don’t know much about artificial technology. Or maybe a societal rejection of AI technology could be beneficial to some big corporation? Time will tell.
To learn more about Google’s AI translation engine, please visit:
https://research.googleblog.com/2016/11/zero-shot-translation-with-googles.html
To learn more about how Bob and Alice are trained to negotiate:
https://code.facebook.com/posts/1686672014972296/deal-or-no-deal-training-ai-bots-to-negotiate/
To read Facebook researchers commenting to Fast Company senior writer Mark Wilson:
https://www.fastcodesign.com/90132632/ai-is-inventing-its-own-perfect-languages-should-we-let-it
(Author: Marina Sevcova)