Building AI We Can Trust

In Support of the White House Executive Order to Safeguard AI

Today, President Biden issued a landmark Executive Order to ensure the rapid innovation in AI is executed responsibly, safely, and in a human-centric way. It is vital that AI is developed and deployed in a trustworthy and responsible way for all users, which we believe will be supported by the frameworks and core principles of the Executive Order.

At Appen, we believe in the power of AI to bring equity to every corner of the world. We also uphold the incredible responsibility of all governments and corporations to build AI that makes a positive impact and reduces harm at all costs.

One critical component of the responsible development and deployment of AI is humans. Humans must be involved in every facet of the creation and use of AI to guarantee it performs as intended. We believe that for AI to serve humanity – its intended purpose – it has to be built by humanity.

 

On the White House’s Four Principles of the Ethical Development of AI:

Safety – we agree that we must create rigorous standards for testing and monitoring AI models to ensure they are behaving as intended. Proper implementation of red-teaming and RLHF is key to developing safe, secure and trustworthy AI. These processes aid in identifying vulnerabilities and mitigating risks by simulating both real-world challenges and potential threats. It also emphasizes learning from human feedback to refine AI behavior over time, making AI systems safer, more secure and ultimately, more trustworthy. More trust in AI and the processes governing it fosters greater public confidence in their use.

Transparency – The entire process of AI development, testing and use should be transparent as it is relevant to matters of the public and AI users. Content provenance and watermarking is critical to tracing and auditing content origins, and verifying authenticity. The use of data and model sheets can help users understand the underlying data and model quality and relevance, including potential biases that may exist. Furthermore, the results of testing and continuous monitoring should be revealed, and the inclusion of humans who represent end users is a critical step in the process.

Individual Privacy – Privacy is a fundamental human right. Without proper safeguards, AI can exacerbate the risks associated with privacy breaches, making it easier for malicious actors to extract, identify, and exploit personal data. As companies increasingly rely on data to train their AI systems, the opportunity and incentive to act maliciously is greater than ever. Privacy by design should be embedded from the outset of AI creation, rather than an afterthought. This consideration should be provided throughout the conceptualization, design and deployment of AI technology. Additionally, being open about how data is collected, used and protected builds trust and demonstrates a commitment to privacy.

Freedom from Discrimination – If not handled responsibly, AI can perpetuate and deepen discrimination, bias and other forms of inequality and abuse. Understanding the source of data used to train AI is vital to identifying and mitigating bias. Provenance information can help researchers and developers assess the potential biases in the training data and address them in the AI system. If handled responsibly, AI has the power to make technology more inclusive. For example, AI-powered technologies can bridge gaps and increase accessibility while equalizing access.

“AI is not just about algorithms, it’s about people,” responds Appen CEO, Armughan Ahmad, to the White House AI Executive Order today. “At Appen, we understand that human involvement is not a stage in AI development — it’s the core. We incorporate human input at every stage of the AI lifecycle to ensure unbiased, helpful and harmless models are built with our data. Humans are at the helm of this promising journey, steering the course of AI towards a future where technology augments human potential.”

Ultimately it is the responsibility of those developing these technologies to act with the best intentions, eliminating the potential risk and harm of AI through careful and transparent processes. We have the incredible opportunity to develop a technology that can transform the world for the better, if done properly. We are eager to review the full details of all eight categories of the White House Executive Order and stand by these core fundamentals of safe AI development and implementation to ensure safety, progress and equitable opportunity for all of humanity:

  1. New Standards for AI Safety and Security
  2. Protecting Americans’ Privacy
  3. Advancing Equity and Civil Rights
  4. Standing Up for Consumers, Patients, and Students
  5. Supporting Workers
  6. Promoting Innovation and Competition
  7. Advancing American Leadership Abroad
  8. Ensuring Responsible and Effective Government Use of AI
Website for deploying AI with world class training data
Language