A Brief History of AI
Artificial intelligence is the hottest and most discussed topic in computer science today and with good reason. Artificial intelligence is the part of computer science that deals with creating systems that show “intelligence.” Artificial intelligence started in the 1950’s and has gradually progressed to the booming field it is today. Its’ impact touches on every field imaginable and will have a huge, positive impact on humanity and the global economy. There is a lot of mystery surrounding artificial intelligence and how it began, so let’s dive into the history.
History of AI
Artificial intelligence started off on a much smaller scale than what we see talked about today. In 1950, Alan Turing published a paper introducing us to a simple aspect of what AI might look like. In “Computing Machinery and Intelligence” Alan Turing wrote on the notion of “machines being able to simulate human beings and the ability to do intelligent things, such as play Chess.” In 1956, John McCarthy coined the term of “Artificial Intelligence” at the Dartmouth Conference devoted to the topic of machines being capable of making their own decisions.
The first running AI program was also revealed in 1956 and gave hope to this new concept. It was called the Logic Theorist written by Allen Newell, J.C. Shaw, and Herbert Simon of Carnegie Mellon University. This was the first program written to mimic the problem-solving ability of a human. It went on to prove 38 out of the first 52 theorems in Whitehead and Russell’s Principia Mathematica. The program even found new and more elegant proofs for a few of the theorems.
Other initial programs of AI were in interactive games. When you play video games today, there is almost always a ‘play against the computer option.’ This was first introduced in checkers, and later, more notably in chess. Claude Shannon wrote the first paper, “Programming a Computer to Play Chess” and since then it has been a refinement process on how a computer learns to play chess. There were two approaches, one, Type-A was brute-force solutions. This is where the computer checks every possible move from each scenario and decides the best plan of action. The other approach, Type-B is when the computer uses special strategies and rules of chess to examine only a few moves.
Initially Type-B approaches were favored — which modeled the goal of artificial intelligence and limits the need for computing power (which was important in the 50’s and 60’s). However, in 1973 developers started playing around with Type-A programs and saw much better results. There are several reasons why Type-A programs have proven to be better than Type-B. For one, it is simpler to program. It takes more computing power, but it is a straightforward problem to check every possible move and figure out what is the best move to make. Whereas a Type-B program needs to be taught all of the different rules and strategies that it needs to follow to be successful. As the amount of computing power we had kept increasing, Type-A programs began to make more and more sense. To this day Type-A programs are still used and this is what IBM’s Deep Blue used to play against Grandmaster Gary Kasparov.
In 1979, the Stanford Cart navigated across a room filled with chairs in five hours and became the first autonomous (no human intervention) vehicle. “The cart moved in one meter spurts punctuated by ten to fifteen minute pauses for image processing and route planning.” This was a major stepping stone in image processing and again showed the ability for a computer to take in input and react intelligently.
In 1986, Mercedes-Benz built the first driverless car that could drive up to 55 mph on empty streets. It was equipped with many cameras and sensors, but it worked! This same process is now being applied to self-driving cars today, but it is on a more grandiose scale. Autonomous vehicles today analyze input in milliseconds and are able to map objects in great detail through advanced lidar technology.
In 1997, the first Robocup soccer match happened having over 40 teams participate and 5000 spectators attended. The eventual goal of RoboCup is to have a team of robotic soccer players beat a human World Cup champion team. (https://www.youtube.com/watch?v=NFNEOooEQX4).
Another notable development in AI happened at Northwestern University when they developed Stats Monkey. This is a program that writes sport news stories without human intervention. It is able to take a broadcast and stats of a sports game and then compose a story summarizing the game. Throughout the past 10–15 years there have been numerous advancements in speech and image recognition for computers as well.
What it is and How it Works
As stated above, artificial intelligence is the process of making a computer think intelligently and being able to perform tasks that normally require a human, such as decision making, speech recognition, and image recognition. However, there are two types of artificial intelligence when we think of it today. One is Artificial General Intelligence, which is a computer capable of doing everything a human can. Second, is Artificial Narrow/Applied Intelligence, where a computer can do what a human can do but only in a narrow bound (i.e. just play chess efficiently).
A concept within artificial intelligence that has been gaining a lot of popularity is Machine Learning. Machine learning is the ability for a computer to take in data and discover patterns, as well as generate insights to make predictions. When computers are given certain amounts of data we are able to “teach” models for how to interpret it. Off of this baseline, computers can look for future patterns and trends to make predictions. This is paramount for computers looking at economic trends, advertising, and quantitative finance.
At the core of artificial intelligence is problem solving; the computer first gathers facts and data about a situation through sensors or other inputs. Then, the computer will compare this information with what it “knows” and decides what the information means. This allows it to run predictions on the best course of action or figure out the most likely outcome based on similar data.
Artificial intelligence’s biggest challenge is understanding how human intelligence works and applying that to computers. How do you have a computer learn naturally and create a neural network? Right now, computers can learn in a limited scope, but nowhere near the scale it will be when we have considered artificial intelligence to have reached the “Artificial General Intelligence” stage described above.
Some modern robots have the ability to learn, but it is in a limited capacity. Learning robots are able to recognize if an action produced a desired action and will try that same action if a similar problem is encountered. For example, when a robot encounters an obstacle and the way this problem is solved is by backing up and taking two steps to the right. The next time an obstacle is encountered this would be the first solution the robot would attempt.
The latest development has been in neural networks which is attempting to mimic the structure of the human brain as a way for computers to classify information. “Neural networks give computers the ability to think and understand the world in the way we do, but allows computers to retain the innate advantages they hold over us such as speed, accuracy, and lack of bias.” The neural network system works based on probability. When data is fed in, the computer will make a prediction with a degree of certainty and is then told whether it is right or wrong and modifies the approach it takes in the future.
Real World Applications
One application of artificial intelligence is with almost everyone wherever they go. Almost all smartphones have some type of voice assistant, and in the case of the iPhone it is Siri. Siri is able to translate verbal commands into the information that we are requesting. Whether it is asking for directions, sending a message to a friend, or adding an event to a calendar this can all be done through Siri. Siri uses natural language processing (act of translating our language into understandable commands) and machine learning to predict and understand questions.
Another major application from artificial intelligence that is revolutionizing a whole industry is driverless cars. This has the ability to take every human driver off the road. By removing the human component driverless cars have the potential to be safer and quicker than the transportation we experience now.
“If it were true that the algorithms are demonstrably, measurably, statistically better than a human driver, then we should not let human drivers on the roads. If you wanted to drive, go to Lego Land….driving can be a fun recreational activity, we just don’t need you on the road.” — Frank Chen, Andreessen Horowitz.
The Obama White House projected that the jobs of ~1.4 million (taxi, bus, truck drivers) drivers are threatened by autonomous vehicles. Researchers also estimate that self-driving cars will reduce traffic fatalities by ~90%.
Tesla, Google, and Uber are leaders in creating driverless cars and all have prototypes out/working. Tesla’s cars have an autopilot mode that recommends you stay aware of what is going on, but it is fully functional to take you from Point A to Point B. Uber has also rolled out self-driving cars in Pittsburgh (there are still driver’s behind the wheel just in case something was to go awry).
Artificial intelligence is also being used to move objects. It sounds simple, but it has had incredibly large business implications for Amazon. Amazon has been using small robots to move goods across the large fulfillment centers so humans do not need to spend time walking in search of certain products. This has allowed Amazon to cut their operating expenses by 20%, which for a company of their magnitude, is billions of dollars.
This idea will find its way into a lot of different industries and even “non-tech” companies. As robots and artificial intelligence become more of a normal, everyday thing, other industries will start adopting these cost saving mechanisms. Agriculture is a huge and time intensive industry. Over the past couple of centuries there have been numerous improvements that changed our society from being overwhelmingly farmers to a society with less than 2% being farmers now. This trend will continue as robotics will be able to further automate machinery in the agriculture industry that is currently operated by humans.
Lastly, machine learning is being used for pattern recognition. This has a wide range of applications, but is one of the most useful as well. Illnesses, stock market trends, and advertising can all be broken down into patterns recognizable by a computer.
Many illnesses are patterns and as machine learning is being applied to the healthcare industry some illnesses can be detected before a human would recognize it. There is a chance that a computer will catch an illness that a doctor may completely miss. Right now, breast cancer checks aren’t sensitive enough to detect about 17% of cases. On March 3rd, Google announced that it was using artificial intelligence to detect breast cancer. “They trained several computers with data Google used a flavor of artificial intelligence called deep learning to analyze thousands of slides of cancer cells provided by a Dutch university. With 230,000 new cases of breast cancer every year in the United States, Google (GOOGL, Tech30) hopes its technology will help pathologists better treat patients. The technology isn’t designed to, or capable of, replacing human doctors.” If this technology is able to help detect some of the cases that doctors miss, as well as save the amount of time doctors spend looking at charts — it is saving doctors a lot of time and money. This is time and money that can be spent treating patients and potentially saving lives of people who would not have been diagnosed with breast cancer.
Artificial intelligence will also be affecting everyday life in office buildings. We all spend a lot of time searching and gathering information. As AI works to automate the more rudimentary roles workers are freed up to spend more of their time on impactful areas, such as helping a customer have a better experience with their product. This will also help to increase productivity in the workforce. “We now have a methodology to automate people in these [white collar] roles. What this means is that if I have a company, I may not fire people — companies tend to try to minimize firings. But I may dramatically slow the rate at which I hire new people and instead invest in automation. Ultimately, this leads to fewer job opportunities in the long term in these areas.” — Jack Clark, OpenAI.
In advertising, there are several major ways that pattern recognition and artificial intelligence is being used to advance the field. One way is to target similar audiences. For example, Facebook is able to track all of their users’ interests that they put on their profile — what TV shows you watch, favorite sports teams, books, music, etc. This helps Facebook to build a profile of who you are and they can group you with people that have similar interests. Facebook can then track how people in this group react to different advertisements. Say one person that fits your exact interests clicked on an ad for the Cardinals, you might see that same ad because you fit that person’s demographic.
Machine learning can also be applied to predict other ads you might like. The example above compares you to a group with similar interests. But everywhere you go online leaves data — and accessing what websites you have been to also creates a profile of who you are as a consumer. This allows ads that would be specifically targeted to what you want. There is an incredible amount of data and pattern matching that is used to make sure that each ad that you see is delivered for a reason.
Snapchat is also using machine learning to impact the advertisement slots they offer to companies on their platform. They use a system called Goal Based Bidding to analyze what users are more likely to swipe up on a certain type of ad. “With goal-based bidding, advertisers can inform Snapchat of when their main goal is increasing swipe-ups — perhaps for app installs, web views or movie trailers — instead of focusing solely on impressions. They can then provide a value for how much they think a swipe is worth, allowing Snap to auto-optimize bidding and delivery to a target audience that’s likely to engage with the ad.” This type of advertising has led to a major increase for the companies using Goal Based Bidding. Around 20% of advertisers on Snapchat’s platform use it, and those that do have reported 40% improved efficiency in interactivity with their consumer.
And lastly, machine learning can optimize the advertisements that are delivered to consumers. By looking at the results of hundreds of thousands of different advertisements you can find patterns that lead to higher click through and sign up rates. Taking advantage of these patterns allows you to enhance current advertisements and tailor a specific ad to what a specific audience wants to see.
Conclusion
Artificial Intelligence is one of the most discussed topics in computer science today and with good reason. Artificial intelligence started out very simply with teaching a computer to think logically and compete in a game that requires strategy, such as chess. There is a long way to go to “Artificial General Intelligence” where a machine is as smart as a human and has the ability to learn on its own. But, the field of artificial intelligence has greatly expanded from its small beginning to the large impact it is having today on every industry.
Most notably, it is revolutionizing the automotive industry leading to a loss of jobs (for taxi, truck, bus drivers), but an increase in safety and ultimately creating jobs in other realms. And lastly, artificial intelligence is having a huge impact on the broad field of advertising. Leading to more personalized ads that you are more likely to interact with and enjoy seeing. Artificial intelligence is still in its infancy, but it will continue to have an incredible impact on our lives.
