This article was written by Aadil Zia Khan, a member of Educative's technical content team.
Artificial Intelligence (AI) has been at the forefront of technological advancement for quite some time now and is expected to continue to gain prominence in the future. This article gives an overview of how this field evolved over time and touches upon some of the significant developments and applications of AI in recent years. The reader will get a glimpse of what the future of Artificial Intelligence will look like and also gain insights into the various trends that we should follow.
Artificial Intelligence and the 4th Industrial Revolution
AI is a critical driver of the fourth Industrial Revolution. The ever-increasing volumes of data being generated, rapidly changing societal patterns, greater interconnectivity, and crosscutting of diverse technologies that characterize the current industrial era are dependent on machine intelligence to derive insights and automate responses.
The global AI market size is expected to grow from around $119.78 billion in 2022 to $1.59 trillion by 2030. The private investment in AI in 2021 exceeded $90 billion. To put this figure into perspective, the invested amount was more than the GDP of 149 countries. Some of the biggest companies are already heavily invested in AI. For instance, Microsoft is investing $10 billion in OpenAI, the creators of the artificial intelligence tool ChatGPT. The expectations from the consumers are moving markets. Recently, shares of Alphabet slid more than 7% during an event in which the company announced Bard, their chat technology to rival ChatGPT—investors were expecting better, and they were expecting more.
Evolution of AI: Turning science fiction into reality
AI has fascinated minds since antiquity. Concepts of artificial life and robots were present in ancient myths as far back as 700 BC. One such myth by Hesiod describes Talos, a giant bronze man built by Hephaestus, the Greek god of invention, to protect the island of Crete from invaders. It was fueled by a life force, ichor, which flowed through a tube running from his head to one of his feet.
The word "robot" comes from an old Slavonic word, robota, for "drudgery." It was first used by the Czech playwright Karel Čapek in his 1920 play "Rossum’s Universal Robots" about a company that mass produces workers to perform mundane tasks humans would prefer not to do.
The transition from fiction to the real world started with Alan Turing’s seminal work "Computing Machinery and Intelligence." In that paper, published in 1950, Turing proposed the Imitation Game, now popular as the Turing Test, to see if a machine can think like a human. In this game, a human interrogator asks questions from a human and from a machine and, from the responses, tries to determine which of the two is a human. Some research labs and discussion groups were already exploring the concept-Turing and made it formal. This was followed by the first AI-based game of checkers created by Arthur Samuel in 1952.
John McCarthy coined the name "Artificial Intelligence" in 1956.
The next two decades saw work on powerful concepts like using search algorithms to replicate reasoning and creating semantic nets to explain the relationship between concepts. The limited processing power of the time hampered the growth and widespread adoption of AI. But that did not stop researchers. This period saw the advent of expert systems, an AI-based system that tries to copy the decision-making skills of a human expert. The first one, Dendral, created in 1965, identified compounds from spectrometer readings. An expert system to diagnose blood diseases, MYCIN, was developed later in 1972. The next leap was in the 80s, which saw improvements in neural networks. The theoretical foundations for these were laid decades earlier, but the processing power available now enabled practical work and swift progress. In 1997, IBM's Deep Blue defeated the grandmaster Gary Kasparov in chess. It was time for AI to take the next leap. The consumer was ready.
The modern age of AI
Recent years have seen an explosion in data volumes (2.5 quintillion bytes of data created per day), very high processing powers (ChatGPT has 175 billion parameters, trained using 570 GB of data obtained from books, Wikipedia, and other written content from the Internet—such scales made possible only by high processing powers), and availability to the masses of affordable, off-the-shelf hardware and open-source libraries. The reduction in training costs of, for example, image classification systems (63.6%) and improvement in training time (94%) since only 2018 give a good idea of how far humans have come in AI technology.
Consequently, AI has seen rapid progress as more and more people adopt it. Major players, such as IBM Watson and OpenAI, recognized the potential impact AI can have on businesses and people’s lives early on. The current AI wave impacts a diverse set of domains, including healthcare, education, e-commerce, manufacturing, virtual assistance, and entertainment, to name just a few. Other areas, such as replicating human creativity and modeling consciousness, are loftier and yet, distant goals— the Holy Grail of AI. The state of AI has yet to make much of a dent there.
If you want to have a beginner-level introduction to the field, you can have a look at Mastering Machine Learning Theory and Practice.
Seeing the impact that AI has had and will continue to have, one can't afford to ignore or overlook it. So, let’s get started. Let's explore this fascinating field and see how it impacts various domains. Let's see how deep the rabbit hole goes.
Entertaining the masses
AI in entertainment has a long history. Initially, the focus was on gaming, and there is a lot that has happened in this domain. From Nimtron, an electromechanical machine that played Nim in the 40s, to early chess and checkers AI players in the 50s to Deep Blue defeating the chess grandmaster Kasparov in 1997, AI in games has come a long way.
Initially, AI in games was restricted to algorithms like pathfinding and decision-making. There is much more to it now. For instance, in games like Spore, AI is used for the procedural generation of content like game assets, terrain, music, and storyline. This saves time and also allows a different and unpredictable experience each time. Companies also mine player data to understand player behavior, identify the interesting parts, and prevent bits that might discourage players. Mining can bring useful insights because players spend millions of person-hours playing, and a 30-minute survey can't capture all this information. Companies like Supercell, creators of Clash of Clans, use this information to improve gameplay and their monetization strategy. Several games also use AI to control the difficulty level based on the player’s skill.
Another application of AI is in detecting cheating. For instance, League of Legends uses multiple algorithms to detect fraudulent behavior like maphacks (a technique by which a player is able to see those parts of the game map that would otherwise be hidden) and smurfing (when a high-ranked player uses a low-ranked account to play in a low-skill bracket).
AI in entertainment is not limited to gaming. It has several other uses as well. After all, the market for AI in media and entertainment is worth $13 billion—and it's still growing. An important focus is on personalized user experience. Social media, such as Twitter and Facebook, use AI to better recommend feeds and accounts to follow. They also use AI to detect spam and hate content to improve their customer experience. AI is at the core of companies like YouTube and SoundCloud, where it helps curate playlists and recommend content based on users' interests and activity history. This translates to more users spending more time on the platform and hence more revenue. SoundCloud acquired Musiio, whose AI product tags audio, curates playlists, and predicts songs a listener would be interested in.
Personalized teacher for every student
The biggest issue teachers face (other than the boredom of grading exams, of course) is the large variation in the capabilities of students in a classroom. There is no one-size-fits-all solution. Each student learns in a different way.
The learning styles can be broadly categorized into visual, auditory, read/write, and kinesthetic. On top of multiple learning styles, each student also learns at a different pace. Students, especially in a multicultural classroom, may have different levels of comprehension and language skills. Students may also have learning disabilities, such as ADHD, autism, and dyslexia, to name a few. A teacher can’t customize their lessons for all students. Doing so for a particular group of students would be unfair to the other groups. But here, AI can help.
Let’s imagine what an AI-assisted classroom would look like. The teacher delivers a bite-sized lecture using a video-conferencing application. Students who have hearing impairments can enable the live audio transcription feature. Someone comfortable in a different language can also enable translation to a language of their choice. If the teacher has a rich vocabulary and is fond of using it, the app can include a feature that recognizes a difficult word and replaces it with a more common synonym. (This idea of translation and transcription is not science fiction— OpenAI provides an API, based on their open-source Whisper model, that does this accurately.) Once a lecture is finished, an adaptive testing application can be used to ascertain if the student has grasped the concept and to what extent. An application for spaced learning can then chime in to determine the frequency of revision needed to make sure the transition from a student’s short-term memory to long-term memory is smooth. The revision itself can be supervised by an application that presents concepts according to the student’s learning styles.
The buck doesn't stop here. Data-driven career counseling should then steer a student’s career path. To prevent burnouts and ensure good mental health, optimal workloads and study times can be determined by an AI expert. And, in order that the teacher doesn’t lose sanity from the amount of work they have to do, an automatic grader can be used, which would significantly save teachers’ time. The list goes on.
There is however a darker side that needs to be addressed. With the advent of generative AI tools like ChatGPT, the line between unfair means and that which is ethical would be blurred. Education would have to adapt to this changing scenario.
There is so much room to play and experiment, it's fascinating. The global market for artificial intelligence in education accounted for $1.55 billion in 2020. Recently, Duolingo, a popular language learning application, introduced a new feature that allows users to chat with and get guidance from a GPT-4-based chatbot. Squirrel AI, another company in the domain of adaptive learning, provides personalized educational plans. Brainly, an online space for peer-to-peer learning, uses AI to filter out low-quality content and also to match students needing help in a domain with other students who perform well in that domain. Content Technologies Inc. analyzes course materials to create study guides and multiple-choice questions. There is something for teachers too. Gradescope offers AI-assisted grading tools to save time.
If you are interested in the vibrant ecosystem of OpenAI and want to get a foot in, our course Using OpenAI API for Natural Language Processing in Python is a good start.
Saving lives one algorithm at a time
The global market for artificial intelligence in healthcare was around $15.1 billion in 2022. AI’s impact runs the gamut from vaccine development to diagnoses to personalized care.
What better use of AI than to save lives? Let's have a look at some of the ways AI is helping save lives and providing comfort to the vulnerable.
In the recent Covid pandemic, the quick development of Covid vaccine surprised the public—but not the researchers. There was a plethora of data gathered over several years that covered previous epidemics, and there was the availability of AI-based techniques to gather insights and automate processes. Moderna employed AI for protein sequencing, and interestingly, they also used AI for resource planning when the vaccines were rolled out, and phone calls were coming in. There are many other companies that use AI for drug research. One such company, Deep Genomics, has an AI platform that helps experts discover and develop genetic medicines.
Another area benefiting from AI is disease diagnosis. Work on disease diagnosis and prediction is not new. One of the earliest expert systems, MYCIN, was diagnosing blood diseases as early as 1972. The recent explosion in data and availability of processing power at much lower costs has enabled much more accurate and robust techniques. With the advent of deep learning, we are in a better position to identify diseases like breast cancer, lung disease, and heart ailments from X-ray and MRI images. One success story is when DeepMind and Google Health were able to create a solution that could detect breast cancer more accurately than human radiologists.
Personalized care is another area where AI is leaving its footprint. Wearable devices monitor vital signs to predict the onset of a serious condition. MySense uses data from wearable devices to make a profile of the user and determines when the daily activities match a state of well-being. Products like Diabeloop determine the insulin dosage needed for a patient and automatically deliver it over the day as needed. Fall detection solutions at home help alert emergency services in case of a dangerous fall. This is especially useful for people living alone and especially critical for old people more prone to losing balance. For remote care, companies like Babylon Health use AI to match patients to the relevant doctors remotely and also to give insights into healthier living.
It does not stop here. Mental health and suicide prevention are other areas that benefit from AI. AI helps practitioners sift through patient data, mining available information and guessing patients’ emotional states from their social media posts, thus predicting mental ailments and alerting authorities if someone is vulnerable to suicide. It helps practitioners plan more effective therapy sessions and enables out-of-clinic activity to improve mental health.
Creating Shangri-La or stepping into an Orwellian future
AI has the potential both to make the world safer and to change it into a much more dangerous Orwellian state (a political system in which the government tries to monitor and control every part of people's lives).
Surveillance is one such domain where AI can be a double-edged sword—a domain seeing major investment and involvement by governments. AI is used for sifting through large amounts of data, maintaining profiles of individuals, and predicting potential antisocial behavior. AI techniques are also giving us better crowd-counting results and can detect faces in a crowd. Interestingly—and worryingly—tools have been developed that can identify faces that are partially covered by masks. While these tools could help ensure a criminal is brought to justice, they could also be used to suppress communities demanding their rights.
Currently, surveillance is mainly done using CCTV cameras. But that could change. Autonomic surveillance drones are a hot area in modern AI technologies. These can be used to monitor an area, such as schools, to detect unauthorized access. They can also track criminals, especially in places where aerial coverage or motor vehicle access would be difficult. They can quickly move to a location where people start gathering to monitor crowds and alert the authorities to brawls. China is a major player in AI-based surveillance. In 2019, Chinese companies, with Huawei leading the pack, were supplying this technology to sixty-three countries, thirty-six of which were a part of China’s Belt and Road Initiative.
Surveillance also extends to the cyber world, and censorship provides the perfect complement to it. There are governments that control Internet traffic that can enter (and exit) their countries. Sensitive material, such as explicit imagery and violent and racist content, is removed. Content filtering is not limited to governments. Social media companies like Facebook use AI to ensure the content on their websites is in line with the company’s policies. Sensitive content is then either removed or tagged so that access is limited. They use the publicly available data to better train their image recognition software. While surveillance and censorship have many benefits, such as child protection, it does open a debate on privacy. People with no criminal intent could suffer as a result of these controlling measures.
Another aspect of security is cybersecurity. Using AI for cybersecurity is not new. It is used by tools like Norton and Avira to identify and remove malicious code such as viruses and worms. Many off-the-shelf AI-based tools, both hardware, and software, exist that are used to prevent, detect, and then mitigate unauthorized access and potential denial of service attacks. For example, IBM provided cybersecurity services to protect Wimbledon’s online presence during the tennis Grand Slam in 2021.
Fixing our transportation woes
A popular 80s TV show, Knight Rider, featured K.I.T.T, a self-driving car that assisted the protagonist in his missions. It didn’t take much time to turn this idea into reality, and soon self-driving cars began to be built. But AI isn't limited to autonomous vehicles. There are several applications that have already gained much popularity in this domain. Let’s imagine up a scenario to see the scope of AI in transport.
Imagine Bob waking up late and missing the school bus. His parents have gone for a walk, so no one can drop him off at school. No worries. Bob remembers that this is not the 90s. He sits in his self-driving car and asks it to take him to school. The car checks the time and realizes that Bob is getting late. So, first of all, it puts on a soothing track to calm Bob. Next, it checks the navigation software to find the shortest path with the least congestion on it and takes it. Now imagine today is also “Murphy’s-Law” day, that is, anything that can go wrong will go wrong. Lightning strikes, and a tree falls, blocking the road. Navigation software recalculates the new shortest path, and the car takes it. Meanwhile, Bob is listening to music with his eyes closed. He reaches school just in time and happily makes for his class. But then he remembers that the teacher had asked him to bring Isaac Asimov’s novel I, Robot for the book club, and he totally forgot. Thinking quickly, he places an order for the book on a popular bookshop’s website. The book is dispatched, and the bookshop uses an autonomous delivery drone to drop the book at Bob’s location. Bob can now relax—thanks to AI.
The scenario isn't far-fetched. There are several off-the-shelf navigation products like Google Maps and TomTom that find the best path. Work is progressing rapidly on fully autonomous cars. Delivery drones are already in use, and companies like UPS already use them to deliver packages. But there are several problems that still need to be addressed. AI raises questions about ethics and legality. Suppose a self-driven car is involved in an accident. Who will be held responsible? The owner, the company who manufactured the car, the scientist who designed the algorithms, or will the car be accused and sent to the gallows? If the car can swerve and crash into a different target, what factors will guide its decision? These are valid questions that need debate.
But AI in transport is not limited to just these science fiction scenarios of autonomous cars and drones, together with the various ethical and legal issues that come with them. AI can also be used to put mechanisms in place for safe driving. Products based on image recognition can be used to detect or predict if a driver is going to doze off and prevent an accident. Truck platooning is another brilliant use of AI, in which one manually driven truck becomes the platoon leader, and trailing vehicles follow it using geolocation and spatial awareness. AI is used to process this data to ensure the trucks maintain their positions. AI is also used by railways, airlines, and shipping companies to determine optimal routes and schedules.
Helping our markets
AI is revolutionizing commerce, finance, manufacturing, and agriculture. Let’s have a look at some ways in which AI is being used.
For one, targeted advertisements and product recommendations have improved a lot over the years. For example, Amazon’s AI-based search engine (using the proprietary algorithm A10) has improved results' relevance considerably, leading to an increase in click-through and conversion rates. Amazon is also able to boast single-day shipping. It achieves this using AI to predict consumer demand, ensure product availability, and optimize delivery routes. More recently, Amazon has eliminated the time customers spend waiting in lines in cashier-less Amazon Go stores. A customer just walks in, specifies credit card information, and then whatever they put in their shopping basket (and do not return to the shelf) is charged to their card.
Financial markets have also benefited. Algorithmic trading relies on AI-based, fully automated tools. Financial institutions use AI to detect fraud, such as credit card and insurance fraud. Banks rely on AI to create profiles of their customers to determine their credit worthiness. Customer experience has also improved, and service times have reduced as a result of automating tasks that, previously, customer representatives would do.
Factories also benefit in multiple ways. For example, by accurately forecasting demands and by reducing wastage and raw material usage, factories are able to decrease their inventory requirements significantly. They also use image recognition tools to detect defects. They have also been able to reduce the need for human intervention at various stages by introducing robots.
Farming has also seen its fair share of progress due to AI. Plant health and the presence of pests and weeds is monitored via drones equipped with regular and infrared cameras. This enables targeted spraying of pesticides and fertilizers, preventing wastage, and reducing overall material costs. Robots are used in all stages between sowing and harvesting and even in post-harvesting packaging and classification into different quality grades. AI is not limited to plants. Animal health and ovulation times are also monitored using wearable devices that ensure actions are taken in a timely manner. In many places, the farmer-to-consumer supply chain also now relies on AI-based technologies to minimize the amount of produce that gets spoiled.
Looking beyond the horizon
The above discussion has discussed some of the ways in which AI is helping build the future. The solutions that have been mentioned are far from perfect. As with anything, there is a lot of room for improvement, and there are application ideas waiting to be found.
Pause and reflect. Has AI impacted your daily life? Do you use products and services relying on AI under the hood? Can you think of the next unicorn to benefit humanity?
It is a marathon, and we are still close to the starting line. In the coming years, AI will improve our productivity—manifold. By handing many of their mundane tasks to AI, humans would have more time at their disposal to focus on their creativity (and their selves, families, and friends). With more personalized solutions, an improvement in the quality of people’s life can be expected. Elusive solutions to complex problems would be found. Entertainment would be more entertaining. There’d be no more time wasted browsing lists to select the right movie for movie night. No game would feel algorithmic. Every experience would be different from the previous one. Access to knowledge would be easier and conveyed in a manner most suited to the individual. It can be expected that resource utilization would be optimal. AI would determine the exact amounts needed and, in doing so, would save precious resources like energy, water, and raw materials. And hopefully, AI would be used to protect people and not create a dystopian future— which, too, is never a possibility to be discounted easily. It was because of this hope I chose not to mention the use of AI in military and weapons research in this blog article.
The age of men is over—the time of the AI has come
Not really, I just like rephrasing Lord of The Rings quotes. AI has indeed come a long way—but it is still far from replacing humans.
It will still be a long time before they can replicate creativity and conscience—if they ever can.
The last decade has seen tremendous progress in generative AI. OpenAI is leading the effort with the launch of tools like ChatGPT for text, DALL-E for images, and Jukebox for music. These tools use prompts from humans to generate the desired content, and subsequent prompts can be used to fine-tune it. But the question remains: can AI imitate human creativity?
To learn more about generative AI, head over to Make Your First GAN Using PyTorch for an introduction.
Let’s take the example of paintings. Western art is traditionally divided into six main eras: Renaissance, baroque, classical, romantic, modern, and Contemporary. Each era consists of multiple styles. For instance, contemporary art includes psychedelics, minimalism, realism, graffiti, photorealism, and so on. The list doesn’t end here. There are non-Western art movements as well. There’s Islamic art with a focus on calligraphy, abstract art, and mathematical patterns. African art has a different flavor. Far East Asian countries have their own style. All these art movements usually start with an artist (or a group) who was able to imagine a style that did not exist yet.
Will AI be able to do that?
Music, too, has several genres. Chuck Berry is considered the father of Rock 'n Roll. Will we have AI pioneering a musical genre?
Will companion robots and virtual assistants be able to tell jokes that are not hackneyed? Will companion robots be able to be romantic and not sound like a cliche from a teen flick? Can AI pick the nuances of human language—and intonation?
Let's not forget the flip side. Will humans be able to cross the "uncanny valley" (The unsettling feeling people experience when AI closely resembles humans in many respects but not convincingly enough)?
Replicating creativity is relatively easier, but can AI replicate conscience? Will their definition of right and wrong be a set of conditional statements? Will they be able to apply ethics to a scenario they have not encountered previously?
Self-awareness of human life and the "algorithmic-ness" of AI are summed up beautifully in an episode of the TV show Star Trek: The Next Generation, in which Captain Picard says to Commander Data, his android first officer, "It is possible to commit no mistakes and still lose. That is not a weakness. That is life."
Technology is evolving at the “speed of thought”— with AI at the forefront. Companies need to be prepared and to move quickly; otherwise, its whiplash could lead them to lose their business. This article has just scratched the surface. The submerged part of the proverbial iceberg is much bigger. So just dive in and enjoy the journey!
If you are a manager, AI Project Management: Deploying and Maintaining AI for Business and Grokking AI for Engineering & Product Managers can help you learn how to integrate AI into your organization.
As always, happy learning!
Top comments (0)