Generative AI

AI-generated content: Navigating AI's impact on truth and integrity

Artificial Intelligence (AI), and especially generative AI, has played a massive role in redefining the relationship between humans and technology. It has reshaped the way we communicate, innovate, work, and consume information. Although this has brought AI to the forefront of the global discourse, it has also presented a question of whether we can truly trust it.

A recent study has revealed that 73% of consumers worldwide trust AI-generated content, whether concerning financial planning, medical diagnoses, healthcare advice, personal relationships, or career planning. However, with the growing threats of cyberattacks and deepfakes, it has become necessary to establish a robust framework to enforce generative AI ethics and establish AI integrity.

AI revolution in content creation

AI-generated content is an umbrella term referring to any content or media (text, images, audio, video, or multimodal) that a generative AI or machine learning model has created. As these models advance, AI-generated content is becoming indistinguishable from human-produced work. This has led to a paradigm shift in the creation and dissemination of information, which has significant social, financial, and regulatory implications.

Discerning truth in AI-generated content

In the new landscape of AI-generated content revolution, it is becoming increasingly challenging to find genuine and valuable information as the principles that have historically guided the determination of truth often fall short. Although AI-generated content utilises some of these principles, like rule-governed or decentralisation of information, it often fails to adhere to others, like commitment to reality, social learning, or accountability and transparency. As the disparity between the cost of generating new information and the cost of finding the proverbial "needle" in the haystack of AI-generated content keeps increasing, discerning the truth becomes even more challenging.

Things like inappropriate or under-representative data, data evasion, or data poisoning (manipulated or misinformation in training data) can result in "hallucinations" or deepfakes, making it extremely difficult to separate misinformation or disinformation from the truth. This can lead to rumours, misleading information, hoaxes, propaganda, and much more, which can bring generative AI ethics and AI integrity into question.

Authentication techniques for preserving AI integrity

AI authentication is an emerging field that is helping tackle the issue of AI integrity by verifying data, models, and outputs generative AI produces. Some of the common authentication techniques that can help enforce generative AI ethics include:

Provenance generation and tracking

Provenance generation and tracking facilitate tracking of the history and quality of the datasets behind the AI-generated content. This includes tracking the origins of the dataset, record of any changes or alterations to the dataset, or any other relevant factors that can affect data integrity. Embedding a signature on the metadata is a great way to create credentials that can help ensure the integrity of the training data and the authenticity of the output.

Watermarking

Similar to the watermarks of "Draft" or "Not for Release" on documents, watermarking in AI-generated content involves embedding a signal into the content that highlights the fact that a generative AI model created the particular piece of content. Collaborative efforts in the tech industry have resulted in the development of various watermarking techniques, including dataset watermarking, model watermarking, or differential watermarking, to help distinguish between AI and human-generated content.

Human authentication

As the name suggests, human authentication requires human involvement to discern or flag AI-generated content within the AI value chain. This involves collaborative efforts from a diverse group and selective intervention to leverage people's collective wisdom and knowledge while maximising resource utilisation. However, it has the potential drawback of creating bottlenecks with delayed authentication, human error, or the introduction of implicit bias in the authentication process.

AI-based filtering

AI's immense capabilities can also help detect AI-generated content. Tools like AI plagiarism detectors or AI content detectors are already helping academicians and researchers enforce integrity and ethical conduct. We can expand these capabilities to provide the first layer of filtering to discern truth in the era of the AI content revolution.
Some other authentication techniques include:

• Blockchain technology
• Statistical detection
• Behavioural analysis
• Turing testing

These can help establish AI integrity in today's information landscape rich with AI-generated content.

Framing a balanced future

As we move forward, we have to diversify our focus from simply leveraging the prowess of AI to preserving the value of human creativity and the integrity of available information. Some best practices that can help navigate this digital future are:

• Enabling AI authentication innovations to grow and thrive.
• Promoting transparency and awareness around AI-generated content.
• Leveraging public-private partnerships to understand opportunities and limitations of various authentication techniques.
• Recognising AI authentication as a shared responsibility.
• Fostering interoperability and collaboration between different authentication techniques.
• Augmenting human intervention at data, model and output levels.
• Investing in the development of robust AI authentication frameworks.

How can Infosys BPM help?

Reinforcing generative AI ethics is imperative when preserving truth and AI integrity in the era of AI-generated content. Whether you want to optimise business processes or access insights for well-informed decision-making, the Infosys Generative AI Business Operations Platform can help you lead the generative revolution while accelerating value creation and upholding AI integrity. Discover our services to reimagine operations and drive AI-first digital transformation across your business.