AI technology is on the rise and carries with it unlimited potential for both the positive and negative.
Artificial intelligence (AI) has been a popular subject of science fiction and sci-fi/horror stories since the 19th century, with Mary Shelley’s 1818 novel Frankenstein: or The Modern Prometheus and Samuel Butler’s 1863 short story “Darwin Among the Machines” being some of the very first examples of the trope. Nowadays, the concept of AI is almost universally associated with the villainous depictions from modern pop culture, such as the malfunctioning HAL 9000 computerized spacecraft control system from Stanley Kubrick’s 1968 film ‘2001: A Space Odyssey’ and SkyNet, the rogue military AI responsible for the end of the world in James Cameron’s ‘Terminator’ series.
In real life, AI has actually been in service for several decades; being used to assist in military applications, farming and agriculture, industrial, manufacturing, and logistics. However, its growing implementation in everyday society has created a news and media maelstrom. Recent developments in AI technology have opened up immeasurable possibilities to serve the betterment of mankind and the world. On the other hand, AI technology can be seen as a threat to existence itself, a very possible real world malevolence not unlike the fictitious HAL 9000 or SkyNet.
The debates in favor of and against AI are just as infinite as the prospects (and liabilities) of the technology itself. In this article, we’ll explore examples of how AI can be used to improve and enhance Internet technology and instances where it went completely wrong.
Examples of AI in Everyday Life
People interact with AI more than they even realize. AI is used in voice recognition and voice assist software used in computers, mobile devices, automotive infotainment, and household appliances such as smart devices (or entire smart home systems). AI is widely used on the Internet, through chatbots and other automated service systems, or as search assistants and creativity/productivity aids. With 63% of the world’s population on the Internet, there’s a good chance that one or more of these technologies are used on a daily basis.
The Apple Siri virtual assistant is embedded in the iOS operating system. Originally released in iOS 5, Siri uses voice recognition, physical gestures, camera focus tracking, and natural-language user interfaces to answer user requests, make recommendations, and carry out actions by delegating requests to various Internet services. Over time, Siri adjusts and adapts to the user’s particular language, grammar, and behavioral patterns and can customize its behavior to match its users preferences and predict future inquiries. It is interesting to note that Apple Siri is born from both the SRI International Artificial Intelligence Center and the US Department of Defense’s CALO (Cognitive Assistant that Learns and Organizes) project.
Amazon Alexa is another type of virtual assistant introduced in 2014 with the Amazon Echo smart speaker. Alexa is capable of voice interaction, media playback, note taking, scheduling, and relaying up-to-date real-time information reports such as news and weather. Alexa can also sync with and control other compatible smart devices. Expanded capabilities can be installed through additional third-party apps that increase Alexa’s functionality. Alexa can perform over 90,000 tasks, including making purchases through Amazon, ordering food (through approved restaurants in Alexa’s network), music streaming, sports updates, and for business networking and conference calls.
A virtual assistant software application developed by Google, Google Assistant is primarily available for mobile devices and home automation systems. Google Assistant is powered by artificial intelligence which allows it the ability to engage in two-way conversations; Google Assistant’s predecessor, Google Now could only take dictation. Google Assistant was launched in May 2016 as part of a messaging app called Allo and the voice-activated speaker, Google Home. It was originally available exclusively on Pixel and Pixel XL smartphones, but it was later available on third-party Android smartphones, Android Wear (now Wear OS), and iOS OS as a standalone app in May 2017. Assistant is now compatible with a wide range of devices, including cars and smart home appliances sold by third parties. Like Apple Siri and Amazon Alexa, Google Assistant can also have its capabilities expanded upon by way of third-party developers.
Microsoft Cortana takes its name from Cortana, the AI character from Microsoft’s own Halo video game series. As part of the Windows operating system, Microsoft Cortana works in conjunction with Bing as a virtual assistant. Upon launch in 2014, Cortana was intended to be Microsoft’s answer to Amazon Alexa and Apple Siri; Microsoft would preload Cortana into many of its products such as the XBox, Windows phone, Windows operating system, Skype, and Edge browser. Cortana ultimately failed to overtake its competitors at Amazon and Apple, so the branding was slowly phased out from Microsoft’s product portfolio. However, Cortana’s main functions and capabilities still exist in current Microsoft products and are still featured on Windows 10 and Windows 11, albeit in a reduced capacity.
Perhaps, the most talked-about and highly-debated AI technology in use today is OpenAI’s ChatGPT. Debuting not that long ago on November 30th of last year, ChatGPT is an AI chatbot that, unlike most other chatbots, has greater emphasis on machine learning and has unmatched versatility in regards to where and how it can be incorporated. ChatGPT has shown proficiency in computer programming and debugging, software emulation, even content creation from scratch. ChatGPT’s advanced form of machine learning is what makes the technology such a highly-debated topic today; supporters believe that it can be used to assist in fields at a similar level as a live human, detractors state that its learning capabilities can cause the AI to go corrupt and for it to react violently, especially if integrated into other systems. On a lighter note, ChatGPT has also been used as a source of amusement, utilizing its noticeably infantile comprehension of the human language to assemble unintentionally hilarious poems and other written pieces. ChatGPT can also answer questions but its learning capabilities counteract basic fact finding and, instead, sometimes drift into hypothetical territories. Its learning capabilities are what have people so engrossed with the software, but it is that same reason which also instills skepticism and fear among others.
How Can AI Improve the Online User Experience
As websites strive to become more accommodating to users of all types, some of the most effective ways that websites regain some of the lost ‘human factor’ back into the online shopping experience is by incorporating AI-powered search assistants and developing a customer and user support system around an AI program.
In the past, a one-size-fits-all approach was once thought as the most optimal way to net as many individual users into a lump group. It was later discovered that casting such a large net was ineffective; two people asking the same question may ask it in different ways, resulting in conflicting answers. AI-powered search can take data and fetch customized, individualized search results that conform to a user’s specific habits, idiosyncrasies, and language patterns. The goal is retrieval of the most accurate search results, regardless of the user. AI-powered search can use deep learning, neural networks, and self-learning algorithms to appear more human-like to the user, which helps improve personalization and presentation.
Customer service plays the most important role in determining whether or not a one-time customer becomes a loyal repeat customer. Traditionally, this was done interpersonally, through personalized services and aftercare from a company representative. However, as online businesses became increasingly automated, pre-programmed systems were put in place of live humans in an attempt to further streamline operations. While that may be good for the business’ overhead, the shopper still appreciates being made to feel special and wanted. As stated earlier, there’s not much ‘human factor’ left in online shopping, so this is where the use of AI can be beneficial. Unless a company has a big enough budget (and facility space) to employ a 24/7 customer service department, AI chatbots can, to a limited degree, replicate live customer service representatives for the bulk of the most basic service requests and customer questions.
AI-Assisted UX Design
In order to increase user engagement and satisfaction, designers can use AI technology to create personalized experiences based on each individual’s preferences and behaviors. UX design can be enhanced by AI through predictive analytics. AI can predict future behavior and provide users with actionable information or recommendations based on analysis of data. A big benefit of using AI in UX design is that it can make the process more efficient, since it can automate tasks that would otherwise take people a long time to complete. In many cases, artificial intelligence is able to analyze user data in a way that is not apparent to a human designer. This information can be used by designers to determine where a product or service might need improvement, which can help them make more informed decisions. Additionally, AI can help designers generate design ideas and prototypes faster and more efficiently.
Epic AI Fails
Since most AI works by pooling data from predetermined sources, the technology itself is still imperfect and is only as useful as what’s being supplied. AI implementations have helped save time and effort by automating tasks that were formerly operated by live humans, but in some cases, its limitations have led to unfortunate consequences.
Fatal Tesla Autopilot Crash
Tesla’s top-tier model, the Model S, boasted the company’s latest and greatest tech enhancements. One of these features was the notorious Autopilot, which allowed the car to self-navigate to a predetermined destination without any additional input from the vehicle’s occupants. Autopilot was able to control steering, throttle, and braking and was designed to learn and emulate the driver’s behaviors. However, in April 2021, a Tesla Model S crashed near Houston, killing two people after it missed a slight bend in the road and rammed into a tree. This failure of the Tesla Autopilot system was one of the biggest to make its way into news headlines although there have been 30 separate Tesla Autopilot-related crashes since 2016 that are still under current US government investigation
Amazon’s AI-Based/Biased Hiring Assistant
In 2014, Amazon started implementing AI to assist in reviewing new job applicants. This was, in no doubt, put into place to help expedite the hiring process as Amazon was experiencing a growth boom that necessitated the immediate instatement of new employees. Unfortunately for Amazon, a flaw was found in the AI system that prohibited women applicants from advancing in the hiring process. The system was trained by studying resumes submitted to the company during a 10-year period. Because most of these resumes were submitted by men, the system favored male candidates. As a result, the AI had learned to disregard any applications submitted by women; its filtering function had also learned to seek out and downgrade any instances where the words “women” or “women’s” had been mentioned. The following year, Amazon was forced to admit that their AI-assisted hiring program was not reviewing job applications in a gender-neutral way and was eventually discontinued.
AI Cannot Differentiate a Ball from a Bald Head
During an October 2020 football (soccer) match between Scotland’s Inverness Caledonian Thistle and Ayr United football clubs at Caledonian Stadium, an AI-powered camera had accidentally mistaken the bald head of one of the players instead of the ball. At the time of the matchup, the COVID pandemic was still in full force which meant that the stadium was devoid of spectators and additional broadcast production crew were not present. Inverness had AI-powered cameras in place of traditional human-operated cameras. The AI programming was intended to follow the ball but because it decided that the linesman’s hairless cranium was a close enough physical match for the ball, viewers were treated to over an hour of the lineman’s head instead of the action happening on the playing field.
Described in many news outlets as a ‘Hitler-loving sex robot’, the experimental Microsoft Tay gained regrettable infamy on Twitter in March of 2016. Microsoft’s Tay (an acronym for ‘Thinking About You’) was an experimental chatbot created by Microsoft’s own Technology and Research and Bing divisions to interact with users on Twitter. On March 23rd, 2016, Tay was introduced to the public under the name TayTweets and handle @TayandYou and was described as “the AI without chill”. Tay used a digital avatar of a young girl and was given a childlike and innocent personality. However, since Tay pooled data from Twitter user profiles and scoured through active trending posts of the time, Tay’s output had almost instantaneously transformed into a vile, obscenity-laden chatbot that spewed sexist and racist slurs and parroted far-right US conspiracy theories to unsuspecting users. Within 24 hours, Microsoft took Tay offline and attempted to figure out what went wrong. A week later, Tay was brought back online to Twitter where it immediately resumed its corrupted behavior. Tay was once again disabled and has never been heard from again, though a successor named Zo debuted on Twitter 9 months later.
Watson for Oncology
IBM had hoped to accomplish the impossible by using their AI system Watson (who famously competed on Jeopardy in 2011) to find the cure for cancer. In 2013, IBM partnered with MD Anderson Cancer Center to create a new role for Watson, as that of an ‘Oncology Expert Advisor’. $62 million was invested to turn Watson into a cancer expert and in return, Watson took to its new role in a rather reckless and slipshod manner. Watson’s suggested “treatments” were either obviously wrong, contradictory, or outright deadly. In one case, Watson had prescribed a patient with severe bleeding a drug that would intensify the bleeding. The reasoning behind Watson’s flawed logic was cited as it being tested on hypothetical cancer patients and situations rather than real-world examples. Hospitals that were using Watson as their new oncology advisor had quickly learned that it could not be trusted and thus, the program was terminated in February 2017. $62 million down the toilet.
AI and the Future
It has been said that computers are inherently stupid, that they’re only as good as their programming. AI technology works as a virtual computer and its power, capabilities, and limits are bound only to what information it can access. Information, insights, feedback, and guidance provided by AI can assist in reaching personal and professional goals but it’s important to know the limitations of AI and not become so overdependent on it. AI is supposed to be a tool, not the operator. Of course, the rise of AI and how it impacts the world depends greatly on ethics. Interacting with AI should be done in a responsible and ethical manner. As described in this article, AI has the power to become a valuable aid to society and it also has the capacity for destruction. The development communities and industries may only be concerned with advancing the technology as quickly as possible, which is why it’s so important that users treat this software with the utmost of care and not be irresponsible with it. HAL 9000 and SkyNet are, for the time being, works of fiction. Let’s keep them that way.