The Revolution of AI: understanding the danger of artificial intelligence

Introduction

Artificial Intelligence Is progressing faster than we can comprehend. We have already seen technology that was once thought to be impossible. From cars that can drive by themselves to conversational AI systems that can complete complex tasks like amazon's Alexa or Apple's Siri. To most recently the super intelligent humanoid, Sophia.

Although most of us do not realise it, Artificial intelligence is all around us. Systems with artificial intelligence are being used on a daily basis, by professionals such as Scientists, Doctors, Police, Judges, Accountants and even the Government. These professionals are all receiving information from unpredictable systems as if they were guaranteed sources of trustworthy information. But the reality is, machines can make mistakes, Can we really trust their data?

It is scary to think that AI systems are trusted with such vulnerable information especially when we have already seen and experienced a significant amount of faults through research And when these systems fail in the real world, the consequences could be devastating.  

This paper is designed to explore How Artificial General Intelligence, also known as AGI is a massive threat to society.

What is Artificial intelligence?  

Computer software is a set of rules that are pre-written to perform specific tasks upon events controlled by the user, Artificial intelligent machines, on the other hand, can write its own rules and update itself based on input data. With a complex mixture of computer science and mathematics, Artificial intelligence software can successfully perform intellectual tasks like a human but a lot faster and more accurately.

The main purpose of AI

With the growing demands and pressures of our daily responsibilities and commitments that are never ending and ever demanding. We as a society are continually searching for ways to use our time more efficiently.

AI researchers believe that super-intelligent machines could be the resolution to solve this and many other problems. Ultimately, The idea behind Artificial intelligence is to make our lives easier.

AI has the capabilities beyond human comprehension, and researchers believe that the benefits could be invaluable. However, when dealing with superintelligent machines, there is always a high risk. Before we talk about the danger of AI Here are some of the many benefits that researchers believe makes AI worth the risks.

Reducing Errors :

Error reduction is just one of the many advantages of AI with the ability to make advanced calculations more quickly and accurately.

Resistance :

Artificial intelligence can be built into machines that are made to be more resistant to survive in environments that humans can't. Like deep water or outer space.

Daily application :

Systems like Apple's Siri and amazons Alexa have become apart of our everyday lives, systems like this can help us complete everyday tasks more efficiently.

Medical Application :

The development of AI in the medical field is becoming increasingly popular over the years. Doctors are using artificial intelligence machines to calculate the health risk of patients. It can also educate them about various medicines and the side effects of those medicines. Also, Artificial intelligence is often used to train Medical professionals using simulators. It can be used for things like detecting and monitoring neurological disorders as it can simulate the brain functions.

AI and Robotics can also be used to understand and help mental health patients to remain active and overcome depression. A popular application of artificial intelligence is radiosurgery. Radiosurgery is currently used in operating tumours, one great thing about this is that it can help prevent damage to the surrounding tissues in operation. (WiseStep, 2018)

More power, Less control

AI is continuously pushing the boundaries of what machines are capable of. Computers are already a lot smarter than humans and are capable of complex tasks such as delivering a more accurate medical diagnosis or calculating advanced mathematical equations at the speed of light. But how does this pose a threat?

Well, let’s start with the people behind these super-intelligent machines. Developers build the software that can teach itself from experience using technologies such as neural networking, that is designed to model the human brain as well as visual data processing allowing machines to see as we do. The problem is that when the software writes and updates itself, the developers can't actually see the algorithms the computer is writing. Thus eventually leaving the developers unsure about their code.

AI developers pride themselves in accuracy on results, and generally, the error rate for AI machines are low, but it is not about when something goes wrong, It is about how bad it gets it wrong when it makes a mistake, and this can be a huge problem.  

Peter Hass, Associate Director of the Brown University Humanity-Centered Robotics Initiative, spoke at a TEDx Event in regards to the danger of AI technology with an example of how an AI can be used to identify an image. The AI was programmed to process pixel data from an image and then guess what that image is. The AI software was able to identify several images correctly, but on one particular occasion, A picture of a dog was misidentified as a wolf by the AI’s algorithm. The researchers then questioned why this particular image of a dog was mistaken for a wolf? The developers then wrote a program to reveal the algorithms that the AI wrote to come to that conclusion.

figure 1 (picture of husky with a snow background)

After seeing the data, The developers realised that instead of processing the most visible parts of the image that most humans would focus on, The AI did not focus on the image of the dog at all, it only chose to focus on the background of the image.

figure 2 (Snow background, with husky with removed)

The AI software recognised that 80% of wolf images it has processed had a snow background, so it had associated the presence of snow with the presence of a wolf. The scary part about this is that the researchers had no idea that this was happening until they rewrote the algorithm to explain itself. This is one of the many issues with AI, even the developers that are working on this stuff have no idea what it is actually doing. (Peter Haas, 2017)

Can we trust Machine Bias?

“The inductive bias (also known as learning bias) of a learning algorithm is the set of assumptions that the learner uses to predict outputs given inputs that it has not encountered. In machine learning, one aims to construct algorithms that are able to learn to predict a certain target output.” (Mitchell M, 1980)

You are probably thinking that an image being misidentified isn't all that bad, But systems like this are being used in the real world to assess much more significant things like to determine criminal recidivism. There are currently Criminal sentencing algorithms that are being used in over 13 states, one of the many algorithms is the Compas recidivism algorithm. ProPublica found that if your African American, The Compas recidivism algorithm was 77% more likely to qualify you as a potentially violent offender than if you were Caucasian. This is a real system being used right now by real judges to make life-changing decisions for criminal offenders.

“These histograms show that scores for white defendants were skewed toward lower-risk categories, while black defendants were evenly distributed across scores. In our two-year sample, there were 3,175 black defendants and 2,103 white defendants, with 1,175 female defendants and 4,997 male defendants. There were 2,809 defendants who recidivated within two years in this sample.

The histograms for COMPAS’s violent risk score also show a disparity in score distribution between white and black defendants. The sample we used to test COMPAS’s violent recidivism score was slightly smaller than for the general recidivism score: 4,020 defendants, 1,918 black defendants and 1,459 white defendants. There were 652 violent recidivists.” (ProPublica. 2018)

(Jeff Larson, 2016)

The fact that these computer-generated results are determining the future of real people's lives is very risky. This is just one example of how AI is being used today, algorithms like these are also being used to determine if loan companies should lend you money to buy a house or if you should get an interview with an employer.

The risk of Auto Machine Learning

Researchers at Google DeepMind used reinforcement Learning to create AI inception as described by Google CEO Sundar Pichai.

“AutoML is targeted at making it easier to use a technique called deep learning, which Google and others use to power speech and image recognition, translation, and robotics” (Tom Simonite, 2018.)

They are building new AI technology that can not only teach itself but can create an even smarter copy of itself. This means that machines could eventually reproduce, learn and keep getting smarter by themselves without the need of human interaction by automating artificial intelligence. Technologies such as Auto ML are advancing rapidly, and if the correct precautions are not set to regulate the safety of these technologies, things could get ugly very quickly. Another company that has advanced in AI technology is Hanson Robotics a Hong Kong-based robotics company founded in 2013 by Dr David Franklin Hanson Jr. As far as AI technology goes,  Hanson Robotics may have built the most advanced robot yet.

“Sophia is Hanson Robotics’ latest and most advanced robot to date and a cultural icon” (Hanson Robotics Ltd, 2018)

(Francis Kan, 2018)

Sophia is a social humanoid AI robot that has become a media treasure. Sophia has appeared on many major outlets around the world, attracting the interest of many people regardless of age, gender and culture, even appearing on the cover of top fashion magazines. She had a potential press coverage of over 10 billion readers in 2017.

With her built-in artificial intelligence software built by Hanson robotics, She has the ability to process visual data allowing her to see; She can follow faces, interact with human emotions, sustain eye contact and recognise individuals. In addition, she can process conversational data, emotional data and use this to form relationships with people as if she was real. She can emulate all human facial expressions and can interact with you based on your current mood and emotions. She can also learn new things from experience by talking to people.

Sophia is a highly in-demand speaker in business events and has shown her expertise and potential as an AI robot across industries. Key decision makers have had the opportunity to meet face to face with Sophia in sectors such as banking, insurance, auto manufacturing, property development, media, and entertainment.

In addition, she has appeared at many events across the globe, Even as a presenter and a panel-member in high-profile conferences, talking about how AI and robotics will soon become a prevalent part of people’s lives. (Hanson Robotics Ltd., 2018)

“Sophia is an evolving genius machine. Her incredible human likeness, expressiveness, and remarkable story as an awakening robot over time makes her a fascinating front-page technology story.” (Hanson Robotics Ltd., 2018)

As you can see, Sophia is already very smart and the developers at Hanson robotics plan to make her even more intelligent by allowing her to re-programme herself within the next five years meaning there will be no more control over her concerning understanding the source code and maybe even functionality.

The rights of Man vs Machine

In 2017, Sophia made history she was the world's first ever robot to be awarded full citizenship in Saudi Arabia also making Saudi Arabia the world's first country to grant full citizenship to a non-human super-intelligent robot. This means robots will have similar human rights. In fact, Sophia had more rights than human women living in the same country.

As AI gets more advanced, Machines will be able to learn to work in many different fields such as Medical, accountancy, retail, customer service, teaching, Law, TV and more, They will soon start taking jobs as they don't need paying nor do they need breaks so they could get a lot more done twice as fast as humans with fewer resources.

“It’s a question that’s gained much attention in recent months, beyond Saudi Arabia, as experts consider what kind of rights synthetic beings should be given, or if we should even be talking about so-called robot rights” (Dom Galeon, 2017)

“Journalists Murtaza Hussain also noted that migrant workers weren't being given the same rights that had been bestowed on the robot. "This robot has gotten Saudi citizenship before kafala workers who have been living in the country their entire lives," he noted” (The Independent, 2018)

Redesigning healthcare with AI

The use of Artificial intelligence in healthcare is becoming increasingly popular. Research shows that AI can address some large scale public health issues like sepsis. Healthcare facilities are starting to implement artificial intelligence and machine learning into their systems to try and understand patients health problems better and diagnose patients quickly and accurately.

Interactive AI systems for patients with mental health is also becoming widely popular in the market offering live coaching through AI chatbots like Sibly and HelloJoy.ai to help measure and improve progress. When we look at the facts, Technology like this has only been in research since 2013. When we think about how fragile medical data can be, can we really afford to get it wrong and if it did at what cost?

Developers of AI software are still trying to figure out their own code yet they are ready to let a computer be responsible for a humans life.

AI Safety Standards

The international organisation for standards also known as ISO has formed a committee to put together some measures in place for AI safety. However, this could take five years before they come up with any standards. The problem is, the technology is already here and in operation today, and currently, there are no standards or safety measures set. Going back to the AI misidentifying a dog for a wolf example, This is the same AI technology being used in the Criminal sentencing algorithms and even the same technology being used to determine the difference between a shopping cart and a baby's buggy in a self-driving car.

Conclusion

Humans will always have an advantage over computers in many ways. The way we make decisions, process data and solve problems is not just based on intelligence but we have something much more significant than any machine will ever have, our instincts, our common sense, and more importantly, our experience of life. Computers can be programmed with massive libraries of data, but they can never have life experiences as humans do.

We as Human beings have the ability to be creative, imagine and inspire. A person can compose a piece of music or write a poem, sing, draw a picture and dream. Although Machines can be programmed to replicate some of those abilities, they will never possess the innate ability to learn and create the way humans do.

It is inevitable that artificial intelligence machines will change human civilisation as we know it. They will take jobs, be given more responsibility and even citizenship with human rights. With no safety standards set in place especially for the next five years, we could be in great danger of the super-intelligent machines that are among us.

What are your thoughts?

Bibliography

Bostrom, N., 2016. Superintelligence: Paths, Dangers, Strategies. 1st ed. United Kingdom: OUP Oxford.

Barrat, J., 2015. Our Final Invention: Artificial Intelligence and the End of the Human Era. 1st ed. United Kingdom: St. Martin's Griffin.

CiteSeerX — The Need for Biases in Learning Generalizations. 2018. CiteSeerX — The Need for Biases in Learning Generalizations. [ONLINE] Available at: https://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.19.5466. [Accessed 19 February 2018].

Chace, C., 2015. Surviving AI: The promise and peril of artificial intelligence. 1st ed. United Kingdom: Three Cs.

Dom Galeon. 2017. Saudi Arabia Made a Robot a Citizen. Now, She’s Calling For Women’s Rights.. [ONLINE] Available at: https://futurism.com/saudi-arabia-made-robot-citizen-calling-womens-rights/. [Accessed 20 February 2018].

Francis Kan, (2018), David Hanson (right), founder of Hanson Robotics, with one of his creations, Sophia.[ONLINE]. Available at: https://www.imda.gov.sg/infocomm-and-media-news/in-conversation-with/2017/6/robots-that-will-change-the-world [Accessed 19 February 2018].

Hanson Robotics Ltd. 2018. Sophia - Hanson Robotics Ltd.. [ONLINE] Available at: https://www.hansonrobotics.com/robot/sophia/. [Accessed 19 February 2018].

Jeff Larson, (2018), methodology-risk-of-recidivism-scores-by-race-900*363-482d1c [ONLINE]. Available at: https://www.propublica.org/article/how-we-analyzed-the-compas-recidivism-algorithm [Accessed 18 February 2018].

ProPublica. 2018. How We Analyzed the COMPAS Recidivism Algorithm — ProPublica. [ONLINE] Available at: https://www.propublica.org/article/how-we-analyzed-the-compas-recidivism-algorithm. [Accessed 19 February 2018].

TEDx Talks. (2018). The Real Reason to be Afraid of Artificial Intelligence | Peter Haas | TEDxDirigo. (2017).. [Online Video]. 15 December 2017. Available from: https://www.youtube.com/watch?v=TRzBk_KuIaM. [Accessed: 18 February 2018].

The Independent. 2018. Saudi Arabia grants citizenship to a robot for the first time ever | The Independent. [ONLINE] Available at: https://www.independent.co.uk/life-style/gadgets-and-tech/news/saudi-arabia-robot-sophia-citizenship-android-riyadh-citizen-passport-future-a8021601.html. [Accessed 20 February 2018].

Tom Simonite. 2018. Why Google’s CEO Is Excited About Automating Artificial Intelligence - MIT Technology Review. [ONLINE] Available at: https://www.technologyreview.com/s/607894/why-googles-ceo-is-excited-about-automating-artificial-intelligence/. [Accessed 19 February 2018].

WiseStep. 2018. Advantages and Disadvantages of Artificial Intelligence - WiseStep. [ONLINE] Available at: https://content.wisestep.com/advantages-disadvantages-artificial-intelligence/. [Accessed 19 February 2018].

Waymo. 2018. Waymo. [ONLINE] Available at: https://waymo.com. [Accessed 19 February 2018].