Hiruy Amanuel: How Can We Ensure Trustworthy AI?

Unfortunately, Hollywood has given artificial intelligence (AI) a bad name. When most people think of AI, they think of shiny, gun-toting cyborgs or faceless servers trying to bring down institutions and take over the world. But there is so much more to know about AI than what you see on TV.

That said, trust and trustworthiness are extremely important when creating and using AI. To achieve this, Hiruy Amanuel outlines 5 ways companies can ensure trustworthy AI.

Explainability

One of the many issues people have with artificial intelligence is explainability. Understanding why an AI system made a decision can be anywhere from difficult to impossible. To ensure trustworthy AI, companies that use AI must be able to dive deep into the processes of their AI system to understand and explain the system’s decision-making. Having full transparency will create greater trust in AI.

Integrity

With artificial intelligence, there exists a need for machine learning integrity to ensure trust. Machine learning integrity ensures that an AI generates any output according to a developer’s predetermined operating and technical parameters. In doing so, companies can be confident that their AI system or systems work as intended. They can also set up limitations to regulate the use of AI. Having predetermined parameters will keep AI from doing anything other than what it was designed for.

Regulations

All artificial intelligence systems must work within a series of well-crafted and lawful, and ethical guidelines. AI developers and builders in the European Union must follow a series of regulations when creating AI systems.

The regulations for AI in the EU include:

• AI empowers humans and is controlled by humans.
• AI is secure and does not violate user privacy.
• Data and algorithms used in and by AI are transparent.
• AI cannot harm the environment or living beings.
• There should be no unintentional bias while developing AI.
• AI must be able to be held accountable.

These regulations do not affect AI systems that create little or no risk to European citizens, such as artificial intelligence used in video games.

Reproducibility

Another aspect of creating and maintaining trust in AI systems is to ensure the system’s reproducibility. AI systems make decisions and cause specific actions. Any decision made by an AI should be reproducible. If not, it will not be clear why the AI made the decision at all. Establishing reproducibility is another essential step to maintaining transparency and trust in your AI.

Conscious Development

The team of developers working on an AI system must ensure that the decisions made by the AI will benefit humans and their businesses. To accomplish this, AI systems must be aligned with human principles, values, and ethics. The developers must bear this in mind during every step of the process. Their mindset(s) will inevitably influence whatever AI they build, so they must be conscious of their intentions and the AI system.

About Hiruy Amanuel

Hiruy Amanuel is a committed philanthropist who has invested in multiple educational and technological efforts in East Africa. Through increased access to quality education and technical resources, Amanuel believes the Horn of Africa can rapidly develop revolutionary technologies.