Sign In

Shea Brown

AI & Algorithm Auditing | Founder & CEO, BABL AI Inc. | ForHumanity Fellow & Certified Auditor (FHCA)

Shea Brown is a knowledgeable and skilled individual with a diverse background in AI Ethics, Machine-learning, and Astrophysics.

With a Ph.D. in Astrophysics from the University of Minnesota, Shea has a strong educational foundation, complemented by a Professional Certificate in Artificial Intelligence from Stanford University, a Master's degree in Physics from Brandeis University, and a Bachelor's degree in Physics from the University of Maine.

Having served in various notable roles, including as an Associate Professor of Instruction at the University of Iowa, a Working Group Member at the IEEE Standards Association, and the CEO of BABL AI, Shea has accumulated valuable experience and insights in academia and industry.

Furthermore, Shea has contributed to the field through roles such as Lecturer, Visiting Assistant Professor, Bolton Postdoctoral Fellow, and Research Assistant at esteemed institutions like the University of Iowa, CSIRO, and the University of Minnesota.

Specializing in algorithmic auditing and AI governance, Shea Brown is not only a researcher and lecturer but also an accomplished speaker and consultant in the areas of AI Ethics, Machine-learning, and Astrophysics.

Highlights

Sep 5 · linkedin

The International Panel on the Information Environment (IPIE) has released a report on the AI auditing ecosystem. It also provides a nice overview of global initiatives that suggest audits.

🔵 "Global Approaches to Auditing Artificial Intelligence: A Literature Review" discusses the current landscape of AI auditing practices.

🔵 It highlights the need for a diverse and trustworthy audit ecosystem, better access to data for auditors, and the importance of considering global impacts of AI systems.

🔵 The review calls for global standards to ensure the quality and effectiveness of AI audits.

BABL AI and other AI audit firms are referenced, they note:

"Other firms have published their own methodologies for auditing the ethical commitments of AI systems, such as BABL AI’s ... [mention of other firms like our friends at Eticas AI]... Notably, these publications offer little detail about specific audits they have conducted."

While true that the publication referenced did not have any specific examples (it was an academic article), I should note that the ACLU has a great repository of public audits for NYC Local Law No. 144 (link in the comments) that has a number of our public audit reports.

Another unfortunate omission is the great work of ForHumanity in this ecosystem; hopefully Ryan Carrier, FHCA can reach out to provide input.

Overall, this report is a great survey and resource for those interested in the AI and algorithm auditing ecosystem.

#aiaudit #aiethics #responsibleai Khoa Lam, Bryan Ilg, Jeffery Recker, Borhane Blili-Hamelin, PhD, Dr. Benjamin Lange, Dr. Cari Miller, Navrina Singh

Aug 30 · linkedin

People have asked for my thoughts on California's proposed AI Safety law.

I don't have strong opinions about it, but here's something. 👇

📘 Summary: California's SB 1047 introduces new requirements for developers of large frontier AI models or derivatives, including:

🅐 Safety and Security Protocols: AI developers must implement written safety and security protocols for AI models before training, detailing the steps to avoid risks of causing "critical harm".

🅑 Full Shutdown Capability: Developers must have the ability to enact a full shutdown of AI models and derivatives to prevent misuse or unauthorized access.

🅒 Independent Audits: Beginning January 1, 2026, developers are required to conduct annual independent audits by third-party auditors to ensure compliance with the Act's provisions.

🅓 Compliance Reporting: Developers must submit statements of compliance to the Attorney General, including assessments of potential risks and measures taken to mitigate those risks.

🅔 AI Safety Incidents: Any safety incidents affecting AI models must be reported to the Attorney General within 72 hours of discovery.

🅕 Whistleblower Protections: Protections are in place for employees who report non-compliance or safety concerns.

🔍 My thoughts:

➊ A major challenge under this new law is the immaturity of frontier model safety research, with much of this research still driven by the companies themselves. NIST and the US AI Safety Institute Consortium should help here (but how much is not clear).

➋ I like the third-party auditing requirement, but I'm already envisioning a lot of work that needs to be done to get the audit quality up to a standard that would make this mechanism meaningful. It's not impossible, but "one does not simply walk into Mordor".

➌ I don't like the specificity of "critical harm", and think we're missing an opportunity to ensure the testing covers "ordinary" individual harm caused by downstream applications (and misuse) of the models. Most covered entities do it anyway, but let's make it required. A consolation is that more forced public disclosure on foundational model testing might allow downstream developers to manage risk better.... but it really hasn't helped so far so that's not a great consolation prize.

Overall, I'm not sure how much substantive will change, but I'm happy there is momentum on AI regulation in the US in general, and hope this evolves as we get better at researching these issues.

#sb1047 #chatgpt #airegulation #frontiermodels Khoa Lam, Borhane Blili-Hamelin, PhD, Dr. Benjamin Lange, Jovana Davidovic, Ali Hasan, Dinah Rabe, Upol E., Karthik Mahalingam, Mert Çuhadaroğlu

Shea Brown
Add to my network

Location

Iowa City, Iowa, United States