Garoetpos.com – One of the challenges for those who are tracking the artificial intelligence industry is very surprising. That is, there is no accepted standard definition of what artificial intelligence really is.
AI figures have a slightly different definition of what AI is. Rodney Brooks said that “artificial intelligence does not mean one thing … it is a collection of practices and works that are collected by people”.
Of course, that is not very suitable for companies that need to understand the breadth of what AI technology is and how to apply it to their specific needs.
In general, most people will agree that the basic purpose of AI is to enable machines to have cognitive, perceptual, and decision-making abilities that were previously only possessed by humans or other intelligent beings.
Max Tegmark defines AI as “non-biological intelligence”. It’s quite simple but doesn’t fully understand what biological intelligence means itself, and trying to build it artificially is a challenge.
At the most abstract level, AI is machine behavior and function that mimics human intelligence and behavior. Specifically, this usually refers to what we consider to be learning, problem solving, understanding and interaction with the real world environment, and linguistic communication and conversation.
But specific problems, especially when we try to apply that intelligence to solve very specific problems that businesses, organizations and individuals have.
Of course there are some of them who pursue AI technology with the aim of solving the ultimate problem: creating artificial general intelligence (AGI) that can handle problems, situations, and thought processes that humans can do.
The purpose of AGI itself is to translate AI in academic and laboratory settings because it answers the basic question of whether intelligence is something that only biological entities possess. But the majority of those who talk about AI on the market today are not talking about AGI or solving basic questions about this intelligence.
Instead, they see the application of a very specific subset of AI to narrow the problem area. This is a classic AI / Broad / Narrow (Strong / Weak) discussion.
Because no one has succeeded in building an AGI solution, all AI solutions are currently narrow. Although of course there are a number of narrow AI solutions aimed at resolving broader intelligence questions, most narrow AI solutions do not try to achieve anything greater than the specific problems of the technology being applied.
What should be known here is that we do not do narrow AI to solve the general AI problem, but narrow AI for narrow AI. It will not be broader for certain organizations.
In fact, it must be said that many companies don’t really care about AGI, and AI’s goals for those organizations are not AGI.
If that is the case, then it seems that the industry’s perception of what AI is and where it goes is different from what many researchers or academics think.
What interests the company most about AI is not solving the question of general intelligence, but rather that there are specific things that humans have done in the organization that they now want from the machine.
The number and magnitude of the tasks differ depending on the organization and the type of problem they are trying to solve. If this is the case, then why bother with vague terms where the original definition and purpose diverge quickly from what is actually practiced?
What is cognitive technology?
Perhaps the better term for narrow AI that is applied for that narrow application is cognitive technology. Instead of trying to build artificial intelligence, companies use cognitive technology more to automate and enable various problem areas that require multiple aspects of cognition.
In general, you can group these aspects of cognition into three “P” categories, borrowed from the autonomous vehicle industry:
Perceive – Understanding the environment around you and the input that comes from the sensor. Cognitive technology related to perception includes the recognition and classification of images and objects (including face recognition), natural language processing and formation, unstructured text and information processing, robotic sensors and IoT signal processing, and other forms of perception computing.
Ability that focuses on perception is an area of AI research that gets the biggest boost from the development of sophisticated neural network approaches, and Deep Learning in particular.
Predict – Understand patterns to predict what will happen next and learn from different iterations to improve overall system performance. Cognitive technology that focuses on prediction using a series of machines
Learning, reinforcement learning, big data, and statistical approaches to process information in large volumes, identify patterns or anomalies, and suggest next steps and results.
Neural networks are very helpful here, but so are other ways to do machine learning and even simpler approaches such as knowledge charts and Bayesian statistical models.
Cognitive technology that focuses on predictions covers the range from big data analysis to complex and human-like decision modes.
Plan – Use what is learned and felt to make decisions and plan the next steps. Cognitive technology that focuses on planning includes models and methods of decision making that try to imitate the way humans make decisions.
Initial efforts included an expert system. Newer methods use various approaches used in situations such as cognitive-enabled cybersecurity or loan decisions.
Cognitive technology that focuses on planning is an area that can use larger AI-general research to improve because current machines lack intuition, common sense, emotional IQ, and other factors that make humans much better in planning and decision making.
From this perspective, it is clear that while cognitive technology is indeed part of Artificial Intelligence technology, the main difference is that AI can be applied both for AGI purposes and for narrowly focused AI applications.
On the other hand, to use the term cognitive technology instead of AI is the acceptance of the fact that the technology applied borrows from AI capabilities but has no ambition to be anything other than technology that is applied to narrow and specific tasks.
The atmosphere in the AI industry has changed. Marketing hype, venture capital dollars, and government interest all help drive the demand for AI skills and technology to the limit.
We are still very far from AGI’s final vision. The company was quickly aware of the limits of AI technology and we risk industry reaction when the company pushes back to what is being overpromised and under-delivered, as we experienced in the first AI Winter.
The big concern is that interest will be too cold and investment and AI research will again slow down, leading to other Winter AI. However, perhaps the problem has never been with the term Artificial Intelligence.
AI has always been a noble destination for organizing research and academic interests, such as building settlements on Mars or interstellar travel. However, as the Space Race has produced technology with wide adoption today, so does AI Quest will produce cognitive technology with broad adoption, even if we never reach the AGI goal. (*)