Many executives who are funding AI initiatives are unaware of the unknown risks associated with the use of AI, the extent of which could lead to significant losses and negative public image. yeah.
Most business leaders and AI teams want to gain first-mover advantage and quickly capitalize on investor and market trends.they are fail quickly and move fast and break things. But with AI technology now able to code itself through new generative AI capabilities, and with less human oversight, we all need to slow down and take steps to enable more trustworthy technology. there is.
Consider the harm untrustworthy AI is already causing – False cancer treatment advice from an AI-based Oncology Expert Advisor program that cost developers $62 million, Amazon’s female-biased recruitment of AI Tools, facial recognition software that victimizes innocent people. Arrested.
These incidents have shattered investor, customer and public confidence. These have resulted in bad press and class action lawsuits, damaging the company’s reputation and lowering market valuations. These incidents have also prompted calls from policy groups for regulation and activists to retaliate against violating companies.
Fortunately, this kind of problem can be minimized. First, we need to know what trustworthy AI is. Then you need to know how to serve it.
What is trustworthy AI?
In a field as rapidly advancing as artificial intelligence, it is difficult to identify the characteristics of trustworthy AI systems. Attributes such as safety, accuracy, and fairness can be tested mathematically with certainty in some AI applications, but these same attributes can be tested in other AI applications, as the AI use case examples cited above demonstrate. It can be difficult, if not impossible, to guarantee in your application.
Still, we can’t build trustworthy AI systems if we don’t know what trustworthy AI means to us. What are the basic principles that help build trust in AI systems?
Below is a list of the 12 principles of trust outlined in my book. What You Don’t Know: The invisible impact of AI on your life and how to regain control. This philosophy is the result of my experience working with Fortune 100 AI teams to successfully scale and operationalize AI initiatives. Building trust has always been the most important factor in successfully implementing an AI project.
The 12-step program is aimed at data scientists and developers who understand that bad things can happen when AI fails, and who have taken leadership roles in rapid and rapid response reform. A culture in which AI has grown.
Each doctrine below is accompanied by a pledge that details the steps AI leaders must take to adhere to the doctrine.
12-step pledge
- Humane. We evaluated whether the use of AI would benefit humanity. We asked whether the AI we developed could do more harm than good to society, the environment, and the pursuit of individual life, freedom, and happiness. We evaluated the potential for bad actors to use AI in unintended and harmful ways. We have safeguards in place to prevent such malicious activity (where applicable). We conducted risk and impact assessments to understand whether AI use cases pose a high, medium, or low risk of causing significant harm.
- consensual. We have asked individuals, business partners, or third parties for permission to use their data for specific AI purposes developed.
- Transparent. For the AI that we develop that influences decisions about an individual’s life, livelihood, and well-being, we want to make sure that those affected are aware of the existence of these algorithms and what impact they may have. I actively communicated in a language they could understand.
- accessible. In addition to notifying the individual whose data the AI used, we documented the decisions the AI made about that individual and made that information available online or in an app. Anyone affected by our AI decision-making system is free to see the results for themselves at any time.
- Infiltrate agency. We have set up and notified an objection program for individuals who feel that the algorithm’s recommendations or source data about them may be incorrect.
- Explainable. I explained the judgment and source of AI in an easy-to-understand manner.
- Private and secure. All information used to develop, train, deploy, manage, and govern the ongoing operation of AI systems is kept private and secure by design, including when third-party vendors and business partners are involved.
- Fair, high-quality data. The data used to train and develop AI systems is based on sound data standards, including data thoroughly analyzed and calibrated for bias, bad data, missing data, etc. A sane and appropriate proxy was used for unobtainable or missing data.
- Responsible. When part of the AI system malfunctions, the company has declared, trained and communicated who is responsible for fixing it. Corporate policies and guidelines related to new AI systems have been updated to ensure that there is always an incident response plan for emergencies that may occur with AI systems.
- Traceable. We set up monitoring tools, processes, and people to understand which parts of our AI system had problems and when they did.
- incorporating feedback. We have provided a way for users, affected people and experts to input into the continuous learning of AI systems. Where possible, we ensured diversity of feedback in this process to mitigate any biases that could have introduced into the system at launch and in the future. This can be as simple as giving each piece of data and each decision a thumbs up or thumbs down rating.
- It’s under control and fixable. When AI systems fail, become skewed, or become corrupted, the company has set up the ability to detect it immediately using model drift and data drift monitoring tools, processes, and designated personnel. We have an incident response plan in place to fix the emergency so that our safety and freedom are guaranteed while the high-stakes AI is down.
Why You Need Trustworthy AI
At a macro level, AI is the foundation of systems that impact the world, and through the use of AI in financial credit scoring and employment systems, it can impact areas such as income-based inequality. Environment due to the huge carbon footprint and water cooling requirements of generative AI. And there are diplomatic tensions between nations, such as China’s alleged surveillance of US citizens via TikTok.
At the micro level, AI can be applied to job and retirement planning, securing a mortgage, job and driver safety, health screening and treatment coverage, arrest and police treatment, political propaganda and fake news, conspiracy theories, and even us. It affects individuals in every way, from society to society. Children’s mental health and online safety.
You need AI you can trust to ensure:
- physical safety. Examples include self-driving cars, robotic manufacturing equipment, and virtually any human-machine interaction where AI is introduced into instant decision-making.
- health. Examples include AI-assisted robotic surgery, AI-assisted imaging tools, clinical decision support, generative AI-based chatbots, health insurance claims systems, and more.
- Ability to secure necessities life such as food and housing. Examples of AI systems that affect these include employment systems, financial credit scoring systems, and mortgage systems.
- rights and freedoms. Potentially infringing AI systems include predictive police systems, face matching programs, and judicial recidivism prediction systems.
- Democracy. AI systems that can violate democratic principles include over-the-top AI surveillance programs, divisive fake news created with generative AI, propaganda, conspiracy theories, and terrorist campaigns spread through AI endorsement engines. Including recruitment.
Key Benefits of Trustworthy AI
AI is a huge investment for companies, a key competitive differentiator, and in some cases a major cost efficiency improvement. Either way, you need to make sure it’s trustworthy so that you can enjoy the following AI benefits:
- Improve your brand reputation.
- Increase market competitiveness.
- Improve your AI return on investment.
- Increase AI adoption rates for your company and customers.
- Reduce the risk of class action lawsuits and material damages.
- Improve regulatory compliance.
- Reduce the risk of incidents that cause serious harm.
- Increase public confidence.
Courtney Abercrombie is the CEO and founder of AI Truth, a non-profit AI ethics organization, and the author of What You Don’t Know: AI’s Unseen Influence on Your Life and How to Take Back Control.please contact her [email protected].
