In the fifth episode of the AI Experience series, we look closer at the importance of contextual understanding in the advancement of AI. For more on this topic, visit www.AIXexchange.com. Developers and potential partners interested in LG’s ThinQ Platform should check out thinq.developer.lge.com.
From AI speaker that recommend movie and TV shows matched to our personal tastes, AI is helping to make daily life more convenient in numerous ways. But we’ve all had that experience when Siri or Alexa responds with the wrong information or doesn’t understand the question. For AI to improve, it needs to better understand context.
“These limitations have inspired the call for a new phase of AI, which will create a more collaborative partnership between humans and machines,” said David Foster, head of Lyft Transit, Bikes and Scooters. “Contextual AI is technology embedded, understands human context and is capable of interacting with humans.”
When the “Age of Contextual AI” arrives, human-AI relationship will take a giant leap forward. Machines and services that can understand context in terms of physical space, the end user’s personality and style of communication – as well as myriad other factors that we take for granted, but are central to our ability to comprehend how society and personal interactions work – will become more like partners and more integrated than ever into the way we live.
Let’s take a look at five different areas that, if successfully navigated, will go a long way toward ensuring that the concept of contextual AI comes to its full fruition.
Different kinds of spaces come with different sets of rules – some written and some unwritten. Based on those rules, we might dress and act in a certain way and also interpret what someone is saying, or what is going on around us, through a space-specific lens. One could argue that the pandemic has made environmental context even more difficult to discern for AI, with our homes now doubling as offices, classrooms, gyms, movie theaters, etc. This shift raises the importance of creating AI that can not only partially determine meaning or motive through reference to the specific type of space, but also from the nature of the interactions taking place within that space.
“We have to assume that AI is going to operate in a heterogeneous world with a lot of non AI-friendly consumers and devices or vehicles,” said Foster. “So the ability to be adaptive, predictive and context-aware is going to be key.”
When AI systems designed for different spaces and areas of our lives eventually converge, there will be serious implications around the exchange of personal data, including the way it is shared, when it is shared and for what purpose. To prevent or reduce risk, it is crucial to include end users in the development process, and to have both developers and policymakers work collaboratively to consider and uncover as many of the consequences as possible of a more integrated and contextually-aware AI network.
With any new technology comes new moral considerations. Given the speed at which AI is developing, it is challenging for humans to evaluate whether each action is being performed in a responsible or ethical manner. Therefore, the task of imbuing AI with a set of values that informs its behavior has become particularly pressing.
While it is ultimately the developers who will ensure that core human values are built into AI systems, the task of choosing what those values are should be not be the domain of any one group. A broad collective that encompasses the full diversity of the modern human experience should be enlisted for this task, as this will help to establish a value set that is more representative of society as a whole and not geared to the interests or beliefs of any single segment.
To engineer an AI that understands context, including distinctly human principles and patterns of behavior, demands the collection and analysis of significantly more user data. Increasing concerns around privacy in the digital age will undoubtedly result in users opting out of providing potentially sensitive personal information, which will unfortunately hinder AI from reaching its full potential, and individual AI solutions from rendering full value.
By transparently declaring and detailing the purpose and need for the collection of each data point, companies developing AI can help foster trust in their technologies and in their ethics as a business. If users can see the need for granting access to certain data and a clear benefit from doing so – such as the collection of home appliance usage data to help them achieve a greener, more cost-effective household – they will be more likely to share that information and to perceive the company as an ally in helping them reach their goals.
AI is already being used in the creative sphere in a variety of interesting ways, from helping to write pop ballads to suggesting creative ideas in filmmaking when IBM’s AI platform, Watson, created the first-ever AI-made movie trailer for 20th Century Fox’s horror film, Morgan.* While such usages represent considerable advancements in AI’s capabilities, experts question the extent to which the technology can develop its own sense of creativity.
“You can give AI a bunch of training data that says, ‘I consider this beautiful. I don’t consider this beautiful,’” says Arvind Krishna, senior vice president of hybrid cloud and director of IBM Research. “Now, if you ask it to create something beautiful from scratch, I think that’s certainly a more distant and challenging frontier.”
However, the idea that AI will one day be able to develop its own sense of creativity is essential to the unlocking of its full potential – specifically, the ability to go beyond the input data to develop new ideas and come to new contextual understandings that will enrich our lives and allow us to accelerate towards a brighter future.
Non-verbal cues such as a speaker’s facial expressions, tone of voice and body language play an important role in interpersonal communication and are factors that AI will need to be able to recognize if it is to provide better, more advanced services. However, because non-verbal cues can differ from person to person, the challenge of designing AI systems that can perceive such differences and respond accordingly, is considerable.
With the development of contextual understanding, AI will become more natural to interact with and more reliable in its ability to provide the information – or perform the tasks – we require of it. As it learns to infer meaning from the situation and surroundings, and the communication style of the end user, artificial intelligence will come closer than ever to being a genuine partner to humankind.