In the not too distant future, machines will be capable of doing some amazingly human things, like showing compassion, lying, creating music, multi-modality communication, exhibiting the capability for self-reflection and recursive thinking. Or at least they will be able to simulate these human capabilities.
Much of the advanced research into AI is focused on technology singularity or General Machine Intelligence, while more practical applications such as speech and vision recognition research is focusing on something known as Pragmatics. Pragmatics is the ability to understand meaning based on context. As an example, words or visual cues alone cannot define proper meaning, and the ability to have context awareness offers great advancement to current recognition capabilities.
Over the past 60 years, there have been many “AI Winters”, where AI optimism has been proceeded by long periods of plateaued progress, research dead-ends and commercial and governmental budget cuts. An example of this may be upon us with AI driver-less cars and end-to-end deep learning. As noted most recently in the press, autonomous driving vehicles are still not accurate enough to safely drive across country. Autonomous driving technology currently requires frequent human intervention as the vehicles actively disengage from driver-less mode. In addition, many researchers believe that the current technology may ultimately not be the right approach.
As with most other radical technological advances, AI hype-cycles are to be expected. While there may be additional plateauing periods in AI’s future, digital transformation at the Enterprise-level (with existing AI capabilities) will steadily continue for the foreseeable future. The real story here is that AI has reached a critical inflection point where it is now one of the primary drivers of digital transformation.
AI First Strategy is currently empowering major business and social movements across the world and is at the heart of societies next wave of digital optimization. While exuberant hype-cycles are very common with new technologies, we are just at the beginning of an AI digital revolution. Currently, over a billion people benefit from some aspects of AI technology in their everyday lives and the depth and breadth of the impact is still in its infancy.
While fear of futuristic AI has existed for a long time, it is helpful to understand AI‘s progression over the past 60 years to understand where it is headed. A brief AI timeline would look something like this:
- 1950s – Ability for Computers to play simple games
- 1970s – Algorithm tweaks
- 1980s – Rule-based Expert Systems
- 1990s – Parallel Computing
- 2000s – Emergence of Neural Networks
- 2010s – Machine Learning Platforms
- 2020s – Widespread Cognitive Computing Services
- 2030s – General Machine Intelligence
Over the next decade, thousands of useful AI technologies and applications will increasingly be imbedded into our daily lives through computer, mobile and third-party infrastructures worldwide. These services will provide additional exponential benefits as AI technologies are made generally available to increasingly wide-spread populations.
AI Corporate Landscape
At the Corporate level, there is a lot of enthusiasm around AI technology, but not much ability to execute. Many corporate behemoths in the Financial Services, Logistics, Transportation and Healthcare industries are leveraging AI technology at the department-level with the primary goal of shining light on corporate dark data that is currently not understood or being utilized. The current state of AI is all about mining the potential data gold mine at the enterprise level using special purpose cognitive computing appliances. Amazingly, only a few concentrated companies own the majority of the data in the world. (i.e. – Google, Amazon, Facebook, Alibaba, etc.)
In reality, there are very few AI stand-alone companies and the ones that do exist will most likely be acquired by existing companies like Google or Amazon. Successful standalone AI companies will ultimately need to solve organizational, societal or individual pain-points in order to be successful. Eventually, AI will be a lot like electricity, it will provide the means to both digitize and optimize everything in our world, however specialty analytical appliance development will be necessary. For corporations, the most important question remains as to what problem are trying to solve with AI?
There are many ways to classify Artificial Intelligence capabilities. A simple ranking from basic to more complex looks something like this:
Basic Reactive Machines – Inability to form memories or improve based on past experience.
Simple Limited Memory – Takes into account relevant data regarding the environment, however does not truly learn or make new representations that improve over time.
Environmentally Aware – Machines that dynamically create increasingly accurate representations about the environment as well as other environmental agents to improve performance over time.
Self-awareness – The ability to make inferences based on intentions in similar situations. This type of advanced AIs will someday be able to create increasingly accurate representations, fully leverage past memories and base decisions on past experiences.
While each of these evolutionary steps represents incremental improve, General Machine Intelligence requires significant architectural breakthroughs to become a reality.
The problem that AI researcher are currently grappling with is truly understanding what human understanding means. Human intelligence can be characterized as logical, contextual, emotional, spiritual aware and dynamically cognitive. Today’s AI enabled computers can understand verbal commands, distinguish pictures, identify diseased cells, drive cars and play games better than most humans can. The trend is for machines to exceed human performance on more and more tasks over the next decade, however making machines truly understand is quite challenging and perhaps several decades of R&D will be necessary to achieve this lofty goal. It is likely that if there is going to be a next AI winter, it will be because AI General Intelligence is such a difficult problem to solve.
The ability to learn context from structured and unstructured data, understand complex relationships and automatically translate language means that computers can evolve with specialized architectures to increase their human-like performance. AI advancements with concepts such as discrete infinity, understanding contextual features and defining semantics on the fly is fundamental to deep learning and general-purpose intelligence. Current state-of-the-art AI cognitive platforms are beginning to understand multiple communication modalities, supports multiple cognitive architectures and provide fluid interpretation and understanding of language.
While we all want our lives to be improved by AI, we don’t want to suffer the potential consequences of human task obsolescence or even worse, accelerated knowledge worker obsolescence. After all, no one fears change that will improve their lives, it’s the possibility of negative change that is unpalatable. Some of societies’ greatest fears emanate from AI’s potential unintended consequences. Allowing machines (which cannot be computationally understood by humans) to make decisions, replace knowledge workers and control widespread social outcomes has its risks. Researchers will continue to endeavored to dissect human intelligence architecture and replicate it in software, however broadly-applicable intelligence will remain illusive for some time.
Bright Future for AI
Barring the few mad-scientists out there using AI for nefarious purposes, AI has a very bright future to fuse human and machine intelligence for the benefit of society. A massive opportunity exists to target inefficient industries with maturing AI technologies, to solve pain-points and provide dramatically increased value. Exponential technologies such as AI will ultimately disrupt existing businesses like never before at an unprecedented pace. If your organization is interested in gaining strategic insight or augmenting business processes with AI technology, HBSC Strategic Services can help. Please contact us at firstname.lastname@example.org or call us at 800-970-7995.