AI Verify Foundation: In June 2023, the Singapore Governments Infocomm Media Development Authority (IMDA) took a significant step by launching the AI Verify Foundation. This initiative represents a global open-source community aimed at bringing together stakeholders from around the world to shape the future of international AI standards.
The core objective of this endeavor is to foster the development and utilization of AI Verify, a groundbreaking AI governance testing framework and software toolkit designed to assess AI models against ethical governance principles. AI Verify accomplishes this through a combination of technical tests, process evaluations, and automated reporting.
Crucial to the success of this initiative is the collective wisdom of the global open-source community. Dr. Ong Chen Hui, Assistant Chief Executive of Business and Technology Group at IMDA, emphasized the importance of this vibrant community in keeping up with the rapid growth of AI. With over 90 corporate members in the Foundation, the collaborative effort is poised to be instrumental in shaping the AI landscape.
To support this initiative and ensure responsible AI development, IMDA has teamed up with Red Hat, a leading provider of enterprise AI solutions, to tap into the open-source community’s expertise.
One of the primary motivations behind the development of AI Verify is the need to mitigate the risks associated with AI models. While many businesses may not be building foundation models themselves, they will increasingly rely on testing toolkits like AI Verify to ensure that the AI applications they deploy are safe, fair, and transparent. Protecting customers, business partners, and the business itself is a top priority.
The decision to embrace an open-source approach is strategic. AI Verity’s flexible architecture allows the open-source community to customize report templates and technical tests to meet the unique governance requirements of AI applications in various domains, from healthcare to finance. Transparency in the code is also a key consideration to encourage trust in AI toolkits.
Guna Chellappan from Red Hat shared that IMDA began collaborating with Red Hat last year to open source AI Verify. They are in the early stages of working with these communities, and Red Hat’s role is to help organize these open-source projects.
Also Read: Tech Titans Thrive with AI in Uncertain Ad Market
AI governance toolkits, like AI Verify, can be likened to seat belts for different vehicles moving at different speeds. While the concept applies universally, different AI applications will require customized tests to address their specific challenges. Just as race cars use racing harnesses for added security, AI tools will need tailored tests that can account for differences in application. Once these tests are in place to safeguard end-users, businesses will feel more confident in accelerating and scaling up AI innovation.
Currently, businesses are primarily focused on AI tools that enhance internal productivity, especially in highly regulated sectors like banking. The willingness to roll out public-facing applications will depend on the availability of such tests. Businesses must ensure that their applications can adapt to the diversity and local contexts of the countries in which they operate, and regulators will want to ensure compliance with local AI regulations.
However, it’s worth noting that regulators have not yet determined the necessary “seat belts” for AI. The specific regulations required to safeguard AI while avoiding unnecessary restrictions on innovation are still evolving. IMDA is actively working to accelerate testing sciences and standards, which will enable AI Verify to be used in different jurisdictions with varying AI regulations.
Ensuring the performance of foundation models, which may be concentrated in the hands of a few large companies, presents another challenge. Dr. Ong stressed the importance of making the expertise required to build foundation models more widely available. Some countries, such as the UAE, have taken the approach of developing their own national foundation models. These models must be responsive to local needs, reflecting the multiculturalism, for example, in Singapore.
Open-source foundation models, like Meta’s LLaMA-2, may offer a solution to checking against governance principles. However, big tech companies may still dominate the foundation models market due to the high barriers to entry for other players.
The AI Verify Foundation and its collaborative, open-source approach represent a pivotal step in shaping the ethical governance of AI on a global scale. By nurturing responsible AI development and involving a diverse range of stakeholders, this initiative holds great promise for the future of AI.
Our Reader’s Queries
What is AI verification?
AI Verification Mechanisms are essential tools that guarantee adherence to regulations by preventing or identifying any unauthorized use of AI or AI control over a system. These mechanisms play a crucial role in ensuring that AI is used ethically and responsibly, without any negative impact on society. By detecting any illicit use of AI, these mechanisms help maintain the integrity and trustworthiness of AI systems. It is imperative to implement these mechanisms to ensure that AI is used for the betterment of humanity and not for any malicious purposes.
What is the foundation of the AI?
Foundation models fall under the category of generative artificial intelligence (generative AI). These models are designed to produce human language instructions as output, based on one or more inputs (prompts). They rely on intricate neural networks such as generative adversarial networks (GANs), transformers, and variational encoders. By utilizing these advanced technologies, foundation models are able to generate unique and creative content that can be used for a variety of purposes.
What is governance in AI?
AI governance refers to the management, direction, and monitoring of an organization’s AI activities. This involves implementing processes that track and document the source of data, models, and associated metadata, as well as establishing pipelines for audits. By ensuring proper AI governance, organizations can maintain transparency and accountability in their use of AI technology.
What are the foundational areas of AI?
Logic, computation, and probability are the three key areas that form the foundation of many fields. Within these areas, there are several important concepts to understand. For example, algorithms are a crucial tool for solving problems in computation, while the incompleteness theorem highlights the limitations of logical systems. Other important terms include computable, tractability, NP completeness, non-deterministic polynomial, and probability. By mastering these concepts, you can gain a deeper understanding of the principles that underlie many areas of study.