Alarm over AI has been growing, with leaders from various sectors voicing concerns about both the increasing power of AI and its role in society.
Thousands of CEOs, technologists, researchers, academics and others signed an open letter in early 2023 calling for a pause in AI deployments, even as millions of people start using ChatGPT and other generative AI systems.
More specifically, they called on "all AI labs to immediately pause for at least 6 months the training of AI systems more powerful than GPT-4."
The March letter starts by calling out AI's "profound risks to society and humanity" and chastising AI labs for engaging in "an out-of-control race to develop and deploy ever more powerful digital minds that no one -- not even their creators -- can understand, predict, or reliably control."
They're not the only ones worried about AI: Forty-six percent of respondents to a February 2023 Monmouth University poll said AI would do equal amounts of harm and good, while 41% believe the technology would do more harm. Only 9% of respondents believe computer scientists can develop AI that would benefit society.
The reality is that AI has many potential advantages and disadvantages. In fact, the open letter signatories acknowledged both the technology's negatives and positives, stating that "Humanity can enjoy a flourishing future with AI."
Top 5 advantages of AI
The benefits of AI include the following:
1. 24/7 availability
One of AI's biggest, and most cited, advantages is its 24/7 availability. Other computer technologies operate around the clock, and companies have benefited from the high availability of such systems -- but only insomuch as humans have been available to work with them. AI's ability to make decisions and take actions independent of human involvement in many business circumstances means the technology can work independently, ensuring continuous operations, said Jordan Rae Kelly, senior managing director and head of cybersecurity for the Americas at FTI Consulting.
AI not only works continuously, it scales almost infinitely, said Sreekar Krishna, principal and national leader for AI at KPMG US. He cited the personalized recommendations companies such as Amazon and Netflix offer to their customers. While a salesclerk who works often enough with a customer might be able to extend such services to that same individual after enough interactions, AI can do so for hundreds of thousands of customers at the same time. AI similarly demonstrates this scalability in the financial industry, where institutions use the technology to instantly verify and validate millions of transactions and monitor for potential fraud every day. "You can't scale to that degree with humans alone. You need automation, and AI is integral to that automation," Krishna said.
3. Improved accuracy and reduced rate of error
Unlike humans, AI systems don't get tired or become distracted. They're able to process infinitely more information, and consistently follow the rules to analyze data and make decisions -- all of which make them far more likely to deliver accurate results nearly all the time. "That's not to say these platforms are perfect," Kelly warned. To deliver such accuracy, AI models must be built on good algorithms that are free from unintended bias, trained on enough high-quality data and monitored to prevent drift.
4. Enhanced safety
An example of AI's ability to improve safety is General Motors' Super Cruise feature that ensures a driver pays attention to the road. According to information on the company's website, "When it is engaged, Super Cruise uses a Driver Attention System that monitors the system status and works to detect your head and eye positioning, reminding you to pay attention to the road and steer manually when needed." Other automotive makers also offer AI-powered capabilities, such as lane-departure warnings, aimed at boosting driver safety.
But AI's safety contributions extend beyond the roadways. It's also used in manufacturing production lines to keep workers safe by, for example, stopping machinery when they get too close to certain areas. And it's used within robots to handle dangerous tasks ranging from defusing bombs to accessing burning buildings, sparing humans from performing those life-threatening jobs.
5. Performs mundane and repetitive tasks
Experts also credit AI for handling repetitive tasks for humans -- both in their jobs and in their personal lives. "It unburdens the grunt work that humans have had to do," Krishna said. As more and more computer systems incorporate AI into their operations, they can perform an increasing amount of lower-level and often boring jobs that can bite into individual's time. Everyday examples of AI's handling of mundane work includes robotic vacuums in the home and data collection in the office. That, in turn, leaves humans with more time for higher-value tasks.
Top 5 disadvantages of AI
Frequently cited drawbacks of AI include the following:
1. A lack of creativity
Although AI has been tasked with creating everything from computer code to visual art, it lacks original thought. "It can only know what it knows. It can't think outside the box, no pun intended," Kelly said. "It's limited by what it can ingest."
AI essentially makes predictions based on algorithms and the training data it has been fed; and although machine learning algorithms help the machine learn over time, it doesn't have the capacity humans have for creativity, inspiration and new ways of thinking. "It's not going to replace critical thinking; it's just going to be another arrow in our quiver," said Chaim Mazal, chief security officer at Gigamon, a maker of cybersecurity technology. A Feb. 2023 report from the World Economic Forum states that even though AI can support and enable human creativity, the "leading opinion is that AI cannot generate fundamentally new ideas on its own." Whether AI could develop those capabilities is a subject of discussion and debate, even as Microsoft researchers in an April 2023 paper assert that AI has evolved to reason like humans.
2. The absence of empathy
AI can be taught to recognize human emotions such as frustration, but a machine cannot empathize and has no ability to feel. Humans can, giving them a huge advantage over unfeeling AI systems in many areas, including the workplace. Consider the fact that the service sector dominates the U.S. economy, with the Brookings Institute calculating that in 2023 "four out of five American workers in the private sector are employed in the service economy, doing everything from delivering care in hospitals and nursing homes to making and serving food to ensuring products make it from ports to store shelves and into consumers' hands." Although some positions -- and aspects of others -- can be automated using AI, Krishna said many of those roles "require empathy and touchpoints."
3. Skill loss in humans
Although experts typically list AI's ability to free people from repetitive and mundane tasks as a positive, some believe this particular benefit comes with a downside: a loss of skills in people. People often advance their knowledge, as well as their personal and professional crafts, by first learning and mastering easy repetitive tasks, allowing them to understand how those tasks fit into the bigger chunks of work they must accomplish to complete an objective. But as AI takes over those entry-level jobs, some have voiced concerns that people could lose their ability to know and understand how to perform those tasks. That could stymie their ability to truly master a profession or trade; it could also leave them without the necessary capabilities to step in and perform the work should the AI fail.
4. Possible overreliance on the technology and increased laziness in humans
Similarly, a contingent of thought leaders have said they fear AI could enable laziness in humans. "We're all concerned about whether it will make us less mindful or thoughtful or thinking," said Bill Wong, principal research director at Info-Tech Research Group. He and others noted that some users seem to use AI without double-checking the results, assuming the technology works flawlessly when it does not. AI is far from perfect, with the online site AI Incident Database "indexing the collective history of harms or near harms realized in the real world by the deployment of artificial intelligence systems."
5. Job loss and displacement
The equivalent of 300 million full-time jobs could be lost to automation, according to an April 2023 report from Goldman Sachs Research. The authors also estimated "that roughly two-thirds of U.S. occupations are exposed to some degree of automation by AI." The story is complicated, though. Economists and researchers have said many jobs will be eliminated by AI, but they've also predicted that AI will shift some workers to higher-value tasks and generate new types of work. Existing and upcoming workers will need to prepare by learning new skills, including the ability to use AI to complement their human capabilities, experts said. "Although the impact of AI on the labor market is likely to be significant, most jobs and industries are only partially exposed to automation and are thus more likely to be complemented rather than substituted by AI," wrote the Goldman Sachs Research authors.
Examples of good AI use cases
AI has enabled significant advances in many areas of society, according to experts. The following use cases illustrate the positive side of this technology:
- Newer model cars use AI-enabled systems to ensure safety on the road by monitoring blind spots, alerting drivers when their attention wanes and taking preventative measures such as braking automatically to avoid crashes.
- Swiss researchers announced in spring 2023 that they used AI as part of a medical treatment plan to help a paralyzed man walk for the first time in 12 years.
- Scientists have always used the latest tools to help them advance their research, and that was certainly the case during the COVID-19 pandemic. Case in point: Researchers created an AI model to help them predict which COVID variants would become dominant, as well as when and where surges in cases would occur.
Examples of bad AI use cases
Although AI itself is a neutral entity, its use in some circumstances demonstrates its limits and potential to harm others. These real-world examples demonstrate how AI can be inappropriately used:
- In late spring 2023, a New York lawyer faced judicial scrutiny for submitting court filings citing fictious cases that had been made up by ChatGPT. The lawyer acknowledged using ChatGPT to draft the document and told a federal judge that he didn't realize the tool could make such an error.
- In 2019, a video of an apparently drunk Nancy Pelosi, a California Democrat who was then U.S. House Speaker, circulated online. The video was a deepfake, a form of media that has been altered using AI. The believability of that Pelosi video, and others that followed, set off alarms about how AI-generated content could be used to distort truth and spread misinformation.
- Another well-publicized example of a poorly executed AI use case happened in 2016, when Microsoft released a chatbot on Twitter. Microsoft engineers designed the bot to act like a female teenager and expected Tay -- the name given to the bot -- to learn to be more like other teens as she engaged online. However, Tay apparently did not have guardrails to block racist, misogynistic and antisemitic language and it became an offensive and hostile bot that Microsoft had to shut down.