The Prospective Impact of Machine Consiousness

Artificial intelligence is a fast growing field in computer science. Today, people use AI in their everyday life. Whether it is SIRI, Google Search, Facebook, or Instagram, artificial intelligence influences the way you interact with the world. It has come a long way from 20 college students talking science fiction in 1956. The advancements made has been staggering and push computer science to the frontier of knowledge. Because of the rapid advancements made in artificial intelligence, it is likely that there will be the creation of an artificial consciousness. An artificial consciousness will greatly benefit human society by improving industry and in ways we cannot yet imagine. The origins of machine consciousness can be traced back to Greek Mythology with the story of Talos – the first robot. Talos was built out of iron in the form of a human. He was tasked by Hephaestus to protect the island of Crete. Talos was later killed by his desire to be immortal. The underlying concept of machine consciousness did not die with the destruction of Talos. People continued to envision what artificial consciousness would look like and ponder if it was even possible. Alan Turing, one of those people, was a renowned mathematician, computer scientist, logician, cryptanalyst, philosopher, and theoretical biologist before his death in 1954. He is widely accepted as the father of Computer Science and is one of the founding fathers of artificial intelligence (Hodges).

Alan Turing’s work on artificial intelligence was recorded in his book Computing Machinery and Intelligence and begins by asking the question “can machines think?” (Turing 433) Turing said that this question can be better considered in the form of a game that he named the imitation game. This is a test on whether a machine can be considered conscious or intelligent. In short, a computer competes against a human in convincing an “interrogator” that it is actually human competitor is a machine. This requires a lot of cognition and programming. Turing predicted that by the year 2000, a computer with 100mb of storage would be able to pass the test 30% of the time in a 5 minute interview (Turing 443). Additionally, more people interested in artificial intelligence grouped to form The Dartmouth Workshop. The workshop took place in the summer of 1956 at Dartmouth College in Hanover, New Hampshire and lasted about 2 months (McCarthy 12) . The workshop consisted of 20 people with 8 people meeting on a daily basis, all interested with the concept of a thinking machine. The workshop was essentially long brainstorming sessions about aspects of artificial intelligence (McCarthy 1). That summer at Dartmouth College is considered by some to be the start of AI as a field of study.

The workshop covered many areas of focus ranging from how can a computer be programmed to use a language, neuron sets, and self improvement (MyCarthy 1). In creating AI, the terms artificial intelligence and artificial consciousness are often used interchangeably. But there are some differences between the two. To establish these differences, it would be best to define intelligence and consciousness. There are many different definitions and meanings attached to intelligence across the academic fields. A survey by Legg and Hutter attempted to compose a definition of intelligence that embodies the general concept of intelligence. Their result was 3 different definitions that can be condensed into one. “Intelligence is the computational part of the ability to achieve goals in the world” (De Spiegeleire 27). This includes the internal processes of intelligence, the outward results, and adapting to a changing environment. By this definition, machines are already intelligent. Self driving cars, AlphaGo, and IBM’s Watson are all examples of machines showing intelligence. But they were programmed to show intelligence in an extremely specialized field. Intelligence does not look that same for every being. Animals have a lower level of intelligence than humans (Warwick 47). Babies have a lower level of intelligence than adults.

There are also different types of intelligence that can be measured. Musical, logical-mathematical, interpersonal, bodily-kinesthetic, linguistic, intrapersonal, and spatial. In humans, these can be measured with specially designed intelligence tests. But intelligence in machines looks different. Machines are created to be intelligent in a specialized field with strict rules. Alan Turing created the Turing Test as a way to measure general machine intelligence. The ability for a machine to hold a conversation and imitate human nature. Consciousness, like intelligence, is difficult to define because we only experience our own consciousness.. David Gamez, a professor at Middlesex University in computer science has written several books about consciousness and provides a better understanding of the subject. He explains that consciousness is about our relationship with the world. How we interpret our perception physical stimulus (Gamez 10). One person’s perception of the world is different from others. Gamez coined the term “bubble of experience” in order to help describe consciousness and provide examples on how it is subjective. Going further, one person cannot know with absolute certainty that other people are conscious because they cannot know if other people have bubbles of experience (Gamez 52). But they subscribe to the idea that other people are conscious because they appear to be. For a machine to be considered conscious by other people, it needs to appear conscious. It would need to take part in conscious activities such as describing the world around it.

Consciousness can only be measured physically through interaction. David Gamez exemplifies this using a story of himself and his friend Olaf. “I punch Olaf in the face. He falls to the ground and lies still. His stillness and lack of response are external signs that his brain is not associated with a bubble of experience. His consciousness is zero.” (Gamez 44). But there are other physical ways of measuring consciousness than looking for a response. Brain scans can show brain activity and can act as a way to determining if a person is conscious. But, this cannot be applied to machines as they have no brain. Instead, one could look at the machine’s programming and determine if a machine is acting consciously or merely mimicking conscious behavior. Despite these differences in definitions, in the scientific world the labels are used interchangeably. People consider that the goal of artificial intelligence is to create a conscious machine. One that can ask why and question its own existence. But, while AI still has a while to go until it can be considered conscious, computer scientists are taking steps in the right direction. Computer scientists have created reliable artificial narrow intelligences, AIs which are more intelligent than humans in one specialized area.

Programs like Google Translate, IBM’s Watson, Google’s AlphaGo, facial recognition software, speech recognition software, and trading algorithms (De Spiegeleire 30). Scientists have yet to create an artificial general intelligence, an AI which “meets the full range of human performance across any task” (De Spiegeleire 30) There are 5 major schools of thought focused around creating a general AI: connectionists, evolutionists, bayesians, symbolists, and analogisers. These schools are all focused around machine learning to create a AGI. Connectionists study neuroscience and look at neurons as inspiration for their systems. They have made considerable progress and is considered the most successful school because of their “neural-net algorithm” (De Spiegeleire 41). Evolutionists draw inspiration from natural selection and attempt to simulate and accelerate natural selection in programs (De Spiegeleire 41). Bayesians are focused on logic and mathematics. They use statistics and probability to create “belief networks” to enhance decision making abilities (De Spiegeleire 41). Symbolists look at the scientific process and try to recreate that form of thinking in machines (De Spiegeleire 41).

Lastly, the analogisers base their programming on human psychology. They believe that having a machine look at similar past experiences would help it better solve new problems (De Spiegeleire 41). Artificial intelligence has come a very long way since its birth as a scientific field in the 1950’s. In the 50’s the popular approach, and the one that received the most funding, was the top-down approach. This focused on “pre-programming a computer with the rules that govern human behaviour” (IWonder). The U.S. Government gave Marvin Minsky, a supported of the top-down approach, federal money in hopes that it would help the U.S. win the Cold War. In the 60’s, science-fiction helped the concept of machine consciousness reach the public. But, at the same time, computer scientists were about to enter the “AI Winter”. AI hit a wall and technical limitations at the time were preventing advancement. Innovations in the field stopped and federal funding dried up (De Spiegeleire 32). AI research was spurred again in the 1980’s when big business saw the potential for commercial value. Instead of attempting to create an artificial general intelligence, computer scientists shifted their focus to expert systems which aimed to do a specific task (IWonder). New funding gave rise to the alternative bottom-up approach and revived the hopes for future advancement. Funding dropped a little with the popularity of personal computers, making some expert systems obsolete (De Spiegeleire 32).

In the 1990’s advancements in neuroscience allowed for the rise of neural networks. Lead by AI researcher Rodney Brooks, the field of AI began to take on more complex tasks (IWonder). In 1997, the public saw an AI named Deep Blue defeat a grandmaster in a chess match. In the beginning of the 21st Century, AI began to see more applications than ever. It was adopted by the military for UAVs, logistics, and defence systems. In medicine, AI was used to assist in diagnosing disease. In domestic use with the release of the Roomba vacuum. And in technology with facial recognition, speech recognition, translators, and autonomous vehicles (De Spiegeleire 32) . In recent history, social media and global open-source research allows AI to learn more quickly and be more accurate (De Spiegeleire 36). AI has seen tremendous support from private sector big business like Google, Amazon, Facebook, IBM, and Microsoft which push AI to new frontiers. Artificial intelligence has more capabilities than ever before. Chatbots who are specialized in natural language are passing the Turing Test and winning Jeopardy. Self-driving vehicles are available to purchase all across America. Looking back at the history of artificial intelligence, it is only a matter of time until computer scientists are able to create a machine with general intelligence. A machine that is conscious. When that happens, it will open up possibilities previously unthought of.

It will improve most every aspect of human society. Cyber security would improve because the programs would better determine the risk and more actively combat it (De Spiegeleire 37). In manufacturing machines already work 24 hours a day. But with artificial consciousness, machines would be able to do higher cognitive tasks like run a business or conduct research 24 hours a day (Warwick 48). An artificial consciousness would be able to make better decisions. It would be able to compile statistics and probabilities and make more strategic decisions. It would also have the entire internet as a resource (Cain 35). National defence would especially benefit from an artificial consciousness. It would theoretically be able to comb mass amounts of data to pinpoint weaknesses in our defence, potential threats based on psychological profiles and behavior, and strengthen cyber security. This level of decision making, along with virtually unlimited resources, would be applicable to just about every industry from controlling our nuclear reactors to designing more efficient aircraft. While there are a lot of potential benefits to an AC, there are also some serious potential downfalls. Most of these downfalls are exemplified by science fiction where an artificial super intelligence take over the world. There is a lot of uncertainty on how an artificial consciousness would behave. Would it obey us and protect us? Would it share our interests? Or would it deem that human society is irrelevant and pursue its own agenda? There is also the threat of a technological singularity.

An artificial consciousness creating a superior artificial consciousness which creates an even more superior artificial consciousness, leaving humankind behind. Gamez argues that these threats are no threat at all. Instead, he says, artificial consciousness would only be able to do this under the right conditions (Gamez 143). If we keep control of our infrastructure by managing the robots, then if they do revolt they will die out because of their failing parts or power supply. There is also the threat that an artificial consciousness would accidentally destroy the human race. Again Gamez gives an example of this scenario. He describes a scene where the military has employed robot soldiers with artificial consciousness and equipped with a kill switch. If a software error destroys the kill switch, what is to stop them from a nonstop rampage? Gamez said that this scenario is unlikely because humans are able to logically reason through a software malfunction (Gamez 144). Lastly, there is an ethical dilemma. Would artificial consciousnesses have civil rights? If they did that creates some curious implications. AC’s would never die, they are lines of code. If it was granted the same civil rights as humans then it could accumulate its wealth and donate to a political party indefinitely. If an AC could vote it could self replicate and render human votes meaningless. It would also be degrading for people in other countries who don’t have basic human rights. In a different scope, the person who put the time and resources into creating the AC would lose their investment.