For the last few years the term “Artificial Intelligence” or “AI” has been bandied around legal conferences with the presumption that everyone knows exactly what is meant. As a long time software engineer turned lawyer, it has always bugged me that the term has been overhyped and used nonchalantly. This brief article hopes to educate the reader in better understanding this exciting new technology.
What is “AI”?
My first experience of AI was in science fiction movies such as “Star Wars” with characters like C3PO and R2D2, Hal in “2001: A Space Odessey” and many others. In all these movies, the machines have become sentient, self aware acting with expertise, volition and even surprise.
The recent explosion of supposed AI in the legal sector and other industries would make one think that the use of the term AI is warranted. But on a closer examination, the systems making such claims aren’t sentient, self-aware or volitional. They are just clever software.
Examples of supposed “AI” are everywhere from autonomous cars to world chess champion beating software like Alphazero. But is the AI term warranted? Autonomous cars deploy various sensors, data and computation to perform complex work of driving. Alphazero is a superfast and efficient software program that builds on a baseline of rules in order to beat other chess programs. These are hardly AI like the movies. At best they are superfast, efficient, pattern matching, massive data crunching software. They are hardly going to take over the world like Skynet in Terminator. The human species will be safe for another year.
In general terms, the AI that we currently have are reactive machine learning systems that are able to respond to external stimuli using large amounts of data at its disposal. The ultimate type of AI are like what we see in the movies, that can understand the people, creatures and objects in the world and the thoughts and emotions that affect its own behaviour. It is conscious of its own existence, feelings and can predict the feelings of others.
We can try and label these types of AI in the following ways:
Artificial Narrow Intelligence (ANI)
This is the AI that exists today and are programs that perform a single task such as autonomous driving cars, playing chess or analysing raw data. ANI can perform real time tasks that takes information from a specific database operating in a pre- determined and pre- defined way. ANI may be very sophisticated and appears to be able to interact with us and process human language but ANI is nowhere close to having human- like intelligence. They can’t think for themselves.
Google Assistant, Google Translate and Siri are examples of ANI. They can only perform within a range of operations that they are designed for. Machine learning is a method of data analysis that is a subset of ANI where systems can learn from data, identify patterns and make decisions with minimal human oversight.
ANI is sometimes referred to as “weak” AI, it is still a great achievement in human innovation and intelligence. ANI systems are able to process large amount of data and complete tasks faster than any human can. IBM Watson, an ANI system is used by Korean Air to digest large amount of historical maintenance records in order to help crews to diagnose and solve problems much more quickly and to reveal hidden connections and provide real time support for issues in the air.
ANI systems act as a building block for more advanced AI systems that will surely be developed in the future.
Artificial General Intelligence (AGI)
AGI or “strong” AI are systems that exhibit human intelligence performing any intellectual tasks that a human can. These are the sorts of AI we mentioned before that are seen in movies which are conscious, sentient, with emotion and are self aware. These types of systems will be able to independently build a broad range of functions and form connections and generalizations across different domains, significantly reducing the time required for training. AGI systems will be just as capable as humans by copying our multi- functional competencies.
Artificial Superintelligence (ASI)
ASI would be the apex of AI research and development. In addition to copying the multi-faceted intelligence of human beings, it will go a step beyond where its cognitive ability will be superior to a human’s. It will be exceedingly better at everything they do because of overwhelmingly greater memory, faster data processing and analysis and decision making capabilities. This is the stuff of science fiction for now but the ultimate goal in AI research.
Benefits of AI Technology Undoubtedly the world is reaping the rewards from this technology and it can only get better. From medical, business, financial to the legal sectors, we have seen a greater uptake of the technology for problem solving and analysis. We are doing things that were just science fiction only 30 years ago.
On the other hand, there are dangers inherent in all technologies and we must guard against abuse such as invasion of data privacy, social manipulations and discrimination.
For now, AGI and ASI are still decades away (and may not even be achievable) and the benefits of ANI far outweighs the risks involved.