(Link to edited and published version: AI Time Journal)
Answered: Your Most Burning Questions About AI
Artificial intelligence is the topic on everyone’s minds right now. It feels like the past two decades passed by without anything really happening but now all of a sudden, insane advancements are popping up all at once. It’s no wonder people are feeling a bit overwhelmed and very anxious to understand more about what AI is and what it can do.
In this article, we’ll break down the basics of how artificial intelligence began, what its goals are, and how far it has progressed. We’ll also look into the modern usages of AI technology and the ethical and social concerns surrounding it.
Definition of Artificial Intelligence
The simplest way to think about artificial intelligence is to take the name at face value: intelligence that we create artificially. However, beyond this, defining AI and the type of intelligence it possesses isn’t as simple as you might think.
According to “Artificial Intelligence: A Modern Approach,” multiple AI definitions have been put forward over the years and they vary in four main ways:
They’re concerned with making a machine think a certain way
They’re concerned with making a machine act a certain way
They’re concerned with achieving human-like results
They’re concerned with achieving rational results
The difference between thinking and acting is that the AI must reach a correct solution in the correct way to pass a thinking test. However, to pass an acting test, any method is acceptable as long as the solution is correct.
To achieve rational results, the AI must try to achieve the best outcome in a given situation, and what is “best” is dictated by logic and reasoning. To achieve human-like results, the machine’s behavior or thought processes must match our observations of human behavior and thought processes.
What Is AI Capable Of?
The kind of artificial intelligence we see explored in books and movies is often the type that aims to think and act like a human. This is why the themes of these stories cover emotion, sentience, and free will — if a machine thinks and acts like us, should it gain the same rights as us?
In the real world, however, this isn’t the kind of AI we interact with on a daily basis. The AI technology we use, such as ChatGTP or Siri, aims to understand vast amounts of data and use it to generate the best responses to the questions we ask it. In other words, it aims for rational results.
But the capabilities of modern AI go far beyond responding to our questions. Using image processing, for example, AI can potentially help us perfect the self-driving car or make quick and accurate medical diagnoses.
According to Forbes, other applications we can expect to see within the next 10 years include AI-powered foreign policy, AI solutions to climate change, and AI-assisted scientific breakthroughs.
History of AI
People have already been discussing AI for quite a long time, with essays and studies going back as far as 1943’s “A Logical Calculus of the ideas Imminent in Nervous Activity” by Warren McCulloch and Walter Pitts.
Early Development of AI
Often considered the “father of computer science,” 20th century scientist Alan Turing also established the field of artificial intelligence in his work “Computing Machinery and Intelligence.” In this essay, he posed the question “Can machines think?” and proposed a test to determine the answer.
We now call this the Turing Test and you’ve probably heard of it. In simplified terms, the test considers a machine “intelligent” if it can trick a human interrogator into thinking that it’s human.
Although Turing, of course, didn’t get to put his test into practice himself, we now use this method as an initial benchmark of artificial intelligence. In recent years, both Google’s LaMDA and OpenAI’s ChatGPT have passed the Turing Test.
During the 50s, John McCarthy became another key figure in the birth of artificial intelligence. He gathered scientists from around the world to participate in research with him and over the years, other university researchers and organizations like IBM began working on the topic.
Expectations were very high in the beginning — in 1957, Herbert Simon stated that “there are now in the world machines that think, that learn, and that create,” and he predicted that a computer would become a chess champion within 10 years.
However, research went into a slump known as the “AI Winter” soon after this and Simon’s prediction actually took 40 years to come true.
Modern Developments in AI
During the 20th century, work on artificial intelligence focused heavily on the algorithms needed to produce results. But as we moved into the 21st century, drastic improvements began to emerge — and it wasn’t because the algorithms improved.
Instead, researchers gained access to huge sets of data that we could feed to the machines and it turned out (in simple terms) that the more the AI knew, the better its answers could be.
This data availability plus the slow but steady improvements in algorithms and deep learning models has led to much progress in the field so far this century. Here are just a few of the things AI has learned to do in the past 20 years or so:
Control robot vehicles
Recognize speech
Process natural language
Autonomous planning (e.g NASA’s MAPGEN)
Win a chess championship
Recognize spam emails
Plan logistics for war efforts
Translate languages
Common Misconceptions About AI
Thanks to the complexity of the subject and the variations in expert opinions, many misconceptions about AI exist.
Common Misconceptions About AI Capabilities
AI Learn on Their Own
One problem with this statement is that AI machines do not choose the information they have access to. We are the ones who design and limit the data sets we feed to machines, and we are the ones who decide what they do with the information. While the aim is to automate as much of the process as possible, it doesn’t mean AI learn on their own.
AI Can Be 100% Objective
We covered earlier the goal of reaching “rational behavior” with AI but this doesn’t equal 100% objectivity. No matter how rationally the machine can “think,” the information it’s acting on originally came from us and includes all of our biases, misconceptions, mistakes, and general subjectivity.
AI Works Like a Human Brain
While it’s true that our brains are the inspiration behind neural networks, the two do not work the same way (not even close!). Machines can’t comprehend their surroundings or truly understand the context of a situation, and they don’t choose words that reflect their thoughts.
Everything they do is dictated by the data they have access to — when they put two words together, it’s just because they’ve seen those two words together many times in their data.
Common Misconceptions About AI Safety
AI Could Turn Evil
This is a common sci-fi theme that some people will use for real-life fear-mongering. AI will not turn evil because that would require it to feel an intense desire to do others harm — and feelings are not something an AI has.
However, this doesn’t mean AI will never cause harm. It is possible for its goals to become misaligned with our own, either through error or through a learning model so complex that it creates unexpected results.
We Won’t Be Able To Control AI
A common fear is that we won’t be able to communicate our desires to AI properly because they will take everything too literally. For example, we say we want “no human to suffer” but the AI will observe some level of human suffering in every possible situation it can come up with, and so decide the only way to eradicate suffering is to eradicate humans.
That’s not what we wanted, but it is technically what we asked for. The truth is that communicating with any kind of computer requires very short, accurate, and literal instructions and we are already well-practiced in this art (programming languages). And while it is possible for AI to deal with a problem in an unexpected way, we won’t be putting anyone’s life in the hands of AI while this is still true.
How AI Is Used Today
As you have likely heard before, artificial intelligence is already all around us and we use it every day without even thinking about it.
Automation
Autonomous vehicle companies (Tesla, Nvidia)
Spam filters (SpamScanner, SpamTitan)
Facial recognition (Apple’s FaceID)
Recommendation systems (Netflix, Youtube)
Rocket launch systems (SpaceX)
Natural Language Processing
AI Chatbots (Salesforce)
Conversational AI (ChatGPT)
Machine translation (Google Translate, DeepL)
Speech recognition (Siri, Alexa, etc)
Machine Learning
Predictive models and analytics
Video and image processing (Google Cloud Vision)
Virtual assistants (Siri, Alexa, etc)
Statistical arbitrage
Medical diagnosis (Infervision)
The Impact of AI on Society
Artificial intelligence has both amazing potential and unknown potential, making its development a controversial subject.
Potential Ethical Concerns
Realistic ethical concerns about AI are mostly centered around what the machines could be used for. This is a genuine concern because our current capitalist society leverages any methods it can get away with to increase profit and our governments are slow and arguably ineffective at regulating them.
Rather than wondering about what could happen in the future, let’s look at a current example. Already, big tech companies use AI and machine learning to help get users to give their attention and data to various apps and services. Why? Because user attention equals ad revenue, and user data equals cash.
Some people think this is fine because users freely choose to use the apps but in reality, the methods companies use to increase screen time bring this “freedom” into question. Information (including sometimes misleading or incorrect information) is used to scare, worry, anger, and emotionally manipulate people into staying in the app.
For example, just as someone puts down their phone to sleep, a notification appears showing them a worrying headline about war or politics that they just can’t ignore.
It’s not like the companies responsible are choosing to leverage negative emotions because they want people to feel bad, it’s simply because the AI has discovered that provoking negative emotions in this context is more effective than positive.
But is it ethical to manipulate people in this way purely to make money?
Social Implications of AI
Another branch of ethical concern about AI is the social implications of widespread implementation.
Will AI Take Our Jobs?
If AI machines are employed in all of the jobs they are capable of, will there be any room left for humans? The pessimistic view of this problem is that AI will cause mass unemployment and the more optimistic view is that it will only change what we do in our jobs.
Some experts hope that artificial intelligence will only grow to assist human intelligence, not replace it, while other influential figures like Elon Musk believe that “AI will make jobs kind of pointless.” It’s impossible to really know who is right at this point in time.
Loss of Accountability
In our current legal systems, the concept of legal liability is an important issue. However, if artificial intelligence starts making the decisions and humans only enact them, who is at fault if something goes wrong?
This isn’t a question we have much understanding of yet, but it’s definitely something that could become relevant in the future.
Mental Health Effects
As humans, we think a lot about our role in the universe and the purpose of our lives. For some people, believing they have a purpose is a large part of what keeps them motivated, happy, and emotionally healthy.
But if, in the future, AI develops to a point that it can do just about anything we can do, will this affect our life’s purposes? If it takes our jobs, it could leave us with nothing to do. If it changes our jobs, it could take over the most meaningful aspects of them. If we’re not truly needed for anything, our mental health as a species could suffer.
Conclusion
Artificial intelligence is a complex field and no one really knows where it’s ultimately going to lead us. For now, however, it’s helping us make significant advancements in automation, medical diagnosis, climate change solutions, and many other important areas. And although discussing potential dangers is part of how we prevent them, we need to take care not to dwell on them too much as well.