Friday, December 20, 2024 — Meeting Recap

Casual Conversation and Job Opportunity

Shawn and Chris have a casual conversation. Chris mentions he is still looking for work, and Shawn suggests he reconnect with Fred at Whiz, as Fred needs someone in Singapore where Chris is based. Chris agrees to pursue this opportunity. Shawn hints at an unusual interaction Chris had that they should discuss privately later. Chris has an upcoming food delivery so goes on mute briefly. Shawn alludes to purposefully challenging circumstances he is facing currently.

Networking and Certifications Discussion

In the meeting, Shawn initiated a discussion about networking and encouraged everyone to connect on LinkedIn. Matt Alvarez then took the lead, inviting everyone to introduce themselves and share any topics they wanted to discuss. Alhaji shared his recent achievement of passing the Terraform Associate Exam and offered advice for others considering the certification. He emphasized the importance of hands-on experience and learning from failures. Kyle also shared his plan to take three beginner Google certifications offered for free by his school. The conversation ended with Matt Alvarez thanking Alhaji for his insights and encouraging others to share their thoughts and experiences.

Addressing AI Security Challenges

Alhaji initiated a discussion on AI security, specifically focusing on the safety and interaction aspects. Matt Alvarez suggested that securing AI infrastructure and generative AI models is a never-ending race due to the constant evolution of social engineering tactics. Neil disagreed, stating that it's a common challenge in security and that there are approaches to mitigate these issues. He mentioned the OWASP top 10 for large language model applications as a starting point for defenders. Neil also highlighted the work of Prompt Security, a startup focusing on AI security, and the importance of not training AI models on confidential data. Jay emphasized the importance of intermediate steps in the process, such as anonymizing data and tying AI models to business systems with user authorizations. The team agreed that while there are challenges, there are also effective ways to address them.

AI Tools and Security Concerns

In the meeting, Kimberly shared her experience with an AI tool, Loopio, which significantly improved their security questionnaires. However, she also mentioned concerns about the tool's ability to crawl internal data without proper controls. Shawn expressed his dislike for Loopio, stating it never provided him with a good answer. Neil discussed his experience with an AI tool, highlighting its limitations and the need for more training data. He also mentioned a conversation with a vendor who has fully embraced generative AI. Matt Currie shared his approach of using multiple AI agents to fact-check each other, which has improved the accuracy of their responses. Paul raised a question about the use of Bayesian statistics in AI models, which Neil responded to, explaining that Bayesian models are useful for tasks like image generation from a prompt. The team also discussed the potential of Microsoft Purview SDK to protect custom AI apps.

AI Applications and Governance Discussion

The team discussed the potential of AI and its applications, with a focus on its use in INFOSEC and network fault detection. They agreed that AI should be used judiciously and not as a replacement for human cognition. The team also discussed the importance of governance and assurance programs for AI, with a particular emphasis on ensuring that AI tools deliver what they are intended to. The conversation also touched on the potential for AI to be used for malicious purposes, but the team did not find substantial evidence to support this. The team also discussed the need for documenting the components of AI, such as its version, training data, and license structure, to ensure its reliability and security.

AI Hype and Misuse Concerns

The team discussed the topic of AI, with various perspectives shared. Chris and Matt Alvarez expressed concerns about the overuse and misuse of AI, while Jay suggested that AI might become a standard part of infrastructure in the future. Matt Currie shared his experience of using AI to improve efficiency in his team's operations. Kimberly compared the current AI hype to the 5G hype in the telecom industry, warning about the potential for overinvestment and disappointment. The team also engaged in light-hearted banter, with Chris joking about the need for more inclusive lighting in Zoom calls.

Exploring AI Benefits and Concerns

The team discussed the potential benefits and concerns of AI. Paul and Chris emphasized the importance of not being overly critical of AI, but also recognizing the need for guardrails. Shawn and Neil highlighted the usefulness of AI in note-taking and language assistance, respectively. Kimberly shared her positive experience with AI in creating a battle card, while Matt Alvarez offered to help address concerns. Mario, who has implemented generative AI in healthcare, offered to share his knowledge and experience. The team agreed to gather questions and concerns for a future discussion on AI implementation and risk mitigation.

Speed of Light and AI Discussion

In the meeting, Don initiated a discussion about the speed of light and its implications for internet speeds. Kyle clarified that the speed of light is a constant and cannot be surpassed, even with the use of undersea cables. The team also discussed the potential of quantum entanglement for faster communication. Matt Alvarez humorously suggested that he would figure out how to fold space and increase internet speeds. The team also touched on the topic of artificial intelligence, with Matt Currie emphasizing the need to use AI as a supplement to human teams, not a replacement. The conversation ended with David sharing his experience using OpenAI for his son's English exams and wishing everyone a happy holiday.