E-commerce and the cloud share real estate with enterprise information systems (IS)—artificial intelligence (AI) and socials do the same. My reporting aims to clarify some distinctions and classifications within an organisation’s digital footprint and the interaction with the worldwide cyberverse. Operating a business can be a mammoth task, but when it is tamed, respect for these generic concepts may help with decision-making, strategy, and securing competitive advantage.
Intelligent systems
By definition, intelligent systems can make decisions on their own, while AI receives precepts from the environment which determine its responses. People who really understand computing generally say that ‘artificial intelligence’ is not a truly accurate title for this technology. Knowing how AI tools are made clarifies that stimulus and responses at the application level are still bound to the mandates of their creators.
The latest iterations of AI can pass the Turing test with the casual user, but this author believes that it will be some time before a system is developed that passes this test universally. To be classified as AI, a machine must be able to: carry out tasks by displaying intelligent, human-like behaviour; behave rationally by perceiving the environment and taking actions to achieve some goals. By focussing on outputs and capabilities, without delving too far into the backend, the term ‘Artificial Intelligence’ becomes more accurate.
Machine learning happens when systems and tools, derived from algorithms, learn and improve through experience and the use of data. Learning is classified as supervised, unsupervised and reinforcement.
Supervised learning: based on a model that is built using historical data, inputs can be classified or measured. Reflective capabilities ensure that the processing of new inputs, assists with classification and measurement into the future.
Unsupervised learning: a model in which algorithms are applied to data to identify the patterns within. The outputs highlight clusters and densities and produce visual representations and broad groupings, otherwise unrecognisable, that can give insight.
Reinforcement learning: in uncertain environments, feedback loops within the learning systems use experiences to identify the best pathway from input to solution.
The foundations upon which AI rely come from transferring expertise from an expert to a computer, and then to the user. This involves:
Knowledge acquisition: domain experts and documented sources share their knowledge, understanding, and the processes involved in maintaining quality.
Knowledge representation: acquired knowledge is organised as rules or frames and stored electronically.
Knowledge inferring: AI is programmed to support and make suggestions for its specialist domain.
Knowledge transfer: AI responds to a user in the form of information and recommendations.
Expert systems (ES)
An expert system is the result of this four-step process and will attempt to mimic human experts in a specific domain. They can offer advice and recommendations and are penetrating society at and increasing rate. Credit card issuers were early adopters of this technology, and IBM Watson is an ES that entered the market while this innovation was still young. The benefits of using this type of AI can increase productivity and quality, but the transfer of knowledge from an expert to an effective ES can present difficulties.
Neural networks
Another classification of AI/ES are neural networks whose design draws on the anatomy of the biological brain. They differentiate from other systems by the way they identify patterns within complex data and make predictions based on this recognition. Detection of credit card fraud, predicting monthly rainfall, and making stock predictions have all been successful by using this model.
Education, agriculture, and energy conservation are industries that were pioneers and early beneficiaries of the emergence of AI. However, adoption of this technology does not come without challenges. The main bottleneck faced by AI is tech limitations, and there are issues attached to data use, ethics, social, economic, and organisation/management of the system. The development of AI can minimise these concerns with design-thinking and a user-centric approach. In this way, solutions can remain socially, ethically, and legally responsible.
IBM have recently launched their z17 Mainframe core infrastructure to facilitate an AI world. Not surprisingly, they have anticipated and respect the primary difficulty facing the global rollout of artificial intelligence, that is, tech limitations. This innovation incorporates a collaborative approach to nurture its support of existing IS and operational platforms and software.
IBM's Z17 mainframe brings a new level of agility to the transaction processing layer of enterprise IS. With this technology advance, an intelligent system can now be more effective at monitoring transactions and making decisions in real time. Learning as they operate, TPS AI platforms are able to identify and deliver the most relevant information to supervisors and managers with a previously unattainable level of efficiency. The responsiveness this can provide should be a game-changer, if wielded effectively in the worldwide marketplace. There also comes an element of trepidation as I write these words, with Moore’s Law still mostly accurate, ethics and morality will hopefully continue to underpin data use and decision-making.
Conclusion
For the most part, these articles have looked at IS within the enterprise. The external forces from the global digital marketplace can also be categorised and organised. It is ultimately the interactions between these spheres that are worth monitoring and managing.