AI as a Problem Solver
Paving A Way For The Kind Of AI We Need
A practical step for AI going forward
“We use computer languages to provide commands, but not to have the computer be knowledgeable and solve problems. When in fact, Artificial intelligence does require the computer to solve problems”
A practical step AI could solve are problems of duplicated information
Taking the marine domain as an example, after decades of software development and technology advances, systems’ use in the marine industry has sophisticated integration needs and persistent problems of duplicated information.
AI can help, provided our approach is to make computers knowledgeable rather than command crunchers. But why do databases have duplicated information?
The data model is not rich enough
One reason is that the data model of the software vendor is not rich enough. So that developers must provide customized data fields to fill in the gaps between systems that need to share information. For example, gaps in substantiated information between vessel maintenance systems and purchasing systems for vessels.
If, in the data model, there is no understanding of the many connections between repairs on a vessel and purchases for a vessel, then customization needs arise. But customizations needs, arising from an insufficiently rich data model, come with new problems, such as rigidity in software. “For computers to help people, it is significant that they too have models of the world like people do.”
Preparing the way for the kind of AI we need
Software today must prepare the way for the kind of AI we need in the future. We need AI to be more like people and to understand more about the world. People understand the world because they make inferences. People make inferences very fast, thanks to extremely efficient retrieval of information relevant to a task and its goal. But one may ask how does AI help to avoid duplicated information?
By understanding a domain like a domain expert
We need AI to come to understand a particular domain, like a domain expert does. On this basis, AI, like a domain expert, can suggest where additional attributes can clarify identification problems and process variation. Places where the domain expert experiences these.
Additionally, AI which understands a particular domain must present users with interfaces that are helpful enough for dealing with inferences and process variation. We refer here to the quality of human-computer interactions providing for the sophistication of human reasoning and information retrieval. It is reasonable to expect that AI, which understands the world of a particular domain, will naturally provide good user experiences. Finally, the benefits from teaching computers to be problem solvers, will require collaborative work from multiple suppliers and sources, using methods of abstraction and integration.