Can Businesses Trust Artificial Intelligence?
The emergence of 5G and the IoT is expected to increase data traffic by ten-fold in the next decade. This massive increase in data traffic from the exploding number of connected devices will fundamentally change the dynamics of the mobile network and make it even more unpredictable. The increased data speeds will undoubtedly make it difficult for ISPs to respond to congestion, equipment failures, and random traffic spikes quickly by leveraging existing tools without impacting network performance. The industry is exploring how machine learning (ML) could dynamically respond to network-related issues and resolve them autonomously.
It is not clear whether the 5G machine learning (ML) algorithms of the future will have equal access to information from which to base their decisions, or whether some will become smarter than others (presumably, some will). This could create scenarios where the smarter (or dumber) ML algorithms will expect other algorithms to act in a certain way. If they do not, it could create unanticipated responses that cascade through a network. Algorithms appear to have contributed to the 2010 flash crash in the US stock market. Will we see similar volatility, due to ML algorithms gone wild, in the mobile technology networks of the future? It is hard to imagine that we would not.
The increased speed of the 5G network will greatly accelerate the number of potential applications for AI because large amounts of real-time data will be transported all over the world exponentially faster. Importantly, we cannot get to the next level of 5G mobile technology without ML, and the next level of 5G mobile technology will deliver more accurate and powerful applications of ML, so they are dependent upon one another. In other words, as AI and ML manage the complexity of the high-speed networks of the future, there will be even more opportunities for other AI and ML applications to emerge.
As more data will be moved around seamlessly in the future we will become less aware that it is happening. The trend of consumers providing data to confirm their identity is gradually shifting to consumers providing their identity in order to confirm their data. Today, a customer may enter a coffee shop and identify all his past transactions by providing a bar code on his smartphone, an email address, or a phone number. AI is shifting the focus from data the customer knows or possesses to who the customer is based on their physical body by leveraging biometric identification.
The way consumers engage with businesses will also change based on the relentless march of AI into every nook and cranny of consumer interaction. In 2018, Google showcased a video of its AI, Duplex, interacting with a hair salon employee and scheduling an appointment on behalf of a real person. Duplex made an appointment at a hair salon on behalf of its human boss and spoke with a restaurant employee who appeared to be a non-native speaker of English. Duplex was somehow able to navigate through the ambiguity of the conversation to determine that no reservations were required. Duplex and AI applications like it will clearly get smarter over time. Not all businesses will be pleased to deal with AI, but some could see the rise of AI as an opportunity to ride with the presiding tide of the times. Will businesses begin displaying AI-friendly stickers or flags on their front doors to symbolize their openness to engaging with AI? This seems to be just a matter of time.
We are on the brink of having AI transact with other human beings on our behalf, and the purveyors of AI are busily crafting ways in which this will occur more easily and naturally. Perhaps one day, in the not too distant future, two or more AI-powered virtual assistants will deal directly with one another to negotiate on behalf of their owners. This reinforces the notion that ML algorithms ought to be developed based on common standards, so as to be able to interact with one another in predictable ways.
In 2018, IBM showcased its AI system‒Project Debater‒to engage in the first ever live debate with humans. The 2016 Israeli national debate champion, Noa Ovadia, and Project Debater both prepared arguments for and against the statement “we should subsidize space exploration.” A short poll after the closing summaries showed that the audience felt the AI enriched their understanding of the issue more than their human counterpart.
Project Debater does not have a body. It cannot sweat or look sick. So, is AI its own medium of communication? Could AI potentially come to represent the embodiment of purely rational thought? It may be beneficial to businesses, in certain circumstances, to keep AI in a box and make it look like a computer, as opposed to anthropomorphizing it into a human-like body. On the surface, it may appear that AI cannot have biases, but this is simply not the case. The quality of an AI system’s arguments and predictions could be linked to many factors, such as the reliability of the data it was trained with and whether that data contained any hidden or historical biases.
Our interaction with companies will continue to evolve as we interact with AI in the form of chatbots and virtual assistants. Our voice patterns, tone, and word usage are already being tracked and consumed by machines. Companies will continue to use AI to personally craft messages, products, and services to cater to our needs based on our digital profiles in cyberspace, and based on our physical attributes and quantifiable demographics. AI-to-AI communication is starting to emerge, enabling two or more algorithms to interact in a predictable way. Whether this is achievable in a consistent manner, and under what circumstances, remains to be seen. As is the case with so many other aspects of AI, AI2AI interactions must ultimately be governed by universal standards, which simply do not yet exist.