Today, your life is being affected by decisions made by AI and machine learning systems. These technologies influence everything from hiring and lending decisions to the content you see online.
When those decisions produce harmful outcomes, many people worry about the technology itself. But the greater risk beyond AI becoming conscious is that people build, deploy and govern AI systems poorly.
For decades, science fiction has imagined futures where conscious computers and AI would either help humanity or work against it. While those stories capture our imagination, they can distract us from the risks we face today. The challenges surrounding AI are rarely the result of self-aware machines. More often, they stem from human decisions, human biases and human oversight failures.
The real risk isn’t AI consciousness
Those concerns are not hypothetical. People are already worried about how AI is being used and governed. The Data and AI Impact Report found that 62% of respondents worried about data privacy, 57% transparency and explainability and 56% ethical use. These concerns point to a common theme: trust. Most people are not worried about conscious machines making independent decisions. They are worried about the people designing, deploying and overseeing AI systems.
When AI causes harm, the root cause is often not the technology itself. It’s the thoughtlessness, inexperience and even malicious intent behind how the technology is being developed and used.
When human bias becomes machine bias
In a now-famous example, Amazon built a recruiting tool in 2014 that discriminated against women. The company trained a machine learning model using resumes submitted by past applicants to determine which candidates would be the best fit for hire. But the majority of the resumes they used to train their model came from male applicants.
As a result, the model learned patterns that favored male candidates and began discriminating against the female applicants. The historic trend of men dominating the technology industry became a bias within their model. Female coders have always existed, often going unnoticed and unrecognized. Despite this context, this model bias wasn’t caught until a year later, when it was noticed that too few women were being recommended for hire.
The Amazon example illustrates an important reality: AI systems often reflect the strengths and weaknesses of the people who build them.
AI adoption requires accountability
Humans make mistakes. Humans discriminate. Likewise, few individuals operate without strong checks and balances, which is common for AI systems today.
As builders, users and maintainers of AI systems, we need to take on the responsibility for developing AI systems that are better than humans. We need to understand the risks of the AI systems we create and build strategies for mitigating those risks. We need to educate and train users on how to leverage AI effectively and how to report issues.
Trust must be earned
Generative AI being able to produce human-like text has brought AI to the forefront of the public’s imagination. AI has become accessible to more people than ever before. New ideas for the potential of AI and new innovations are being churned out at a rapid pace. With this rise in popularity, there is also a greater scrutiny of AI systems and emerging regulations worldwide.
Trust is key to AI adoption and trust is earned. Trust requires transparency not just in AI, but in the systems that support it.
The Data and AI Impact Report found that 46% of organizations face a trust dilemma, meaning there is a gap between perceived trust and actual trustworthiness. Actual trustworthiness is demonstrated in investments in governance, oversight and practices that make AI systems reliable, ethical and transparent.
As AI becomes more deeply embedded in business and society, organizations will need to show not only that AI works, but that it works responsibly.
