The prospect of artificial intelligence displacing the human brain raises many existential questions.
Already, AI’s abilities seamlessly intuit our needs and optimize our decisions, suggesting everything from purchase recommendations to the ideal route to take on the daily commute. Additionally, countless people look to chatbots, robocallers, and virtual assistants for guidance every day — not to mention the roboadvisors that give daily stock tips. In short, the current and future applications of AI are astonishing.
But the fact is that AI, for all its amazing achievements, can’t yet carry out one crucial task the human brain does naturally: reason. Common sense is simply amiss in AI programming. Its current uses, which do exceed human capacities in many ways, exist persistently outside the realm of replacing the human brain. Real-life applications of AI are finding their limits.
Keeping Aware of the Dangers of AI
The point upon which humans’ abilities diverge from those of AI’s is when intuiting beyond a set of guidelines: AI is bound by sets of rules and objectives, whereas humans can reason within and without rules.
When presented with identical situations defined by imperfect boundaries, an AI will follow through to the logical solution, whereas a human might clearly see the objective beyond the rules to reach a very different conclusion. Unforeseen problems (and even dangerous outcomes) can be the result — and the world is already full of poorly defined boundary conditions. No one wants to be the first murder victim of a self-driving car calculating odds of survival between two unavoidable paths, no matter how well the autopilot had operated for the previous thousands of miles.
Further, when discussing AI’s limits, we would be remiss not to address AI’s potential for abuse. Many examples in recent memory prove that without proper safeguards, AI bots can easily be taught malicious, antisocial, or deceptive behaviors.
Back in 2016, Microsoft released a chatbot — Tay — on Twitter. It took a matter of hours for it to learn from others and, unfortunately, post incredibly racist tweets. In another case, a State Farm commercial aired footage from what was supposedly a 1998 ESPN report where an analyst made accurate predictions of what was to happen in 2020 — but it turned out to be AI-generated. Similarly, an alleged clip of former President Barack Obama using an expletive to describe President Donald Trump made its way through the social media sphere, but that was also doctored with the help of AI. Examples of AI’s abuse go on and on.
To be fair, major tech companies have been fairly quick to curb these abuses. Amazon enacted a one-year pause on its facial recognition product for police use, and IBM took this one step further by completely abandoning its facial recognition project. Part of this is likely due to the current political environment, but the reality is far more worrisome: According to a study of three commercial facial recognition systems conducted by computer scientist Joy Buolamwini, the difference in error rates when recognizing darker-skinned women’s gender (more than 34% in two cases) versus lighter-skinned men’s gender (never above 0.8%) was astounding.
The Potential of AI
Despite the potential for AI-related abuse and problems, there are many widespread and promising future applications for AI. One such application that’s already widespread is the automation of repetitive tasks. From assembly lines to bookkeeping and legal discovery, current AI bots can now quickly learn to repeat myriad tasks with little instruction or supervision.
Another AI application that registers above human capacity is uninterrupted spans of attention — a critical function when watching for anomalies. Vigilant AI bots currently monitor networks for aberrant behavior to raise the alarm for potential hacking. They can also be found poring over mountains of medical scans for infinitesimal signs of cancerous cells. Professional sports teams even use them to analyze innumerable hours of plays for overlooked signs of a promising superstar. What’s astounding is that these bots often learn the signals and scans of normal behavior on their own.
What might be most promising, especially for anyone looking to invest in AI technology, is the advanced computer thought process of companionship. Emoshape, an Ascend portfolio company, currently offers an emotional intelligence layer that can equip chatbots, virtual assistants, and digital avatars with their own personality. Empathy, humor, sarcasm, jealousy, and happiness are just a few of the basic human emotions that might no longer be limited to human-to-human interactions. Soon, talking to Siri in the car could resemble a scene right out of “Knight Rider” (an ’80s crime drama about a police detective and his talking car, for those not versed in Hasselhoff).
Considering Investments in AI
These tech advancements are exciting and often mind-bending. Before you pull the trigger and invest in AI technology, however, remember that AI and ML are now widely available as cloud services at little cost; adding artificial learning to a business is as simple as downloading an app from the app store. This means the investment space is busy and crowded with teams upon teams cultivating a menagerie of applications.
This isn’t to say opportunities to invest in AI technology are vanishing; they can still be found in unique applications or companies with exclusive access to rare or elusive data. Augmented intelligence, for example, is one of the areas of real-life AI applications that appears promising; rather than replacing a person, the technology is there to improve his or her job performance. Self-driving technology is another application that isn’t going anywhere.
If you’d like to learn more about future applications of AI or AI startups to watch, feel free to download our whitepaper, sign up for our newsletter, or schedule an appointment to talk with our general partner and expert in data-centric tech, Dan Conner.
Image by Amanda Dalbjörn