Down the Security Rabbithole Podcast (DtSR) DtSR Episode 656 - Deeper Down the AI Rabbithole Part 2
8 snips
Jun 3, 2025 Join Erik Bloch, a technical expert on AI, and John Dickson, who offers insights into its unique characteristics, as they delve deeper into artificial intelligence. They discuss the intricate mechanics and challenges of AI, touching on its energy demands and ethical implications. The conversation shifts to AI's transformative impact on fields like healthcare and law, emphasizing human oversight. They explore the risks of blind trust in AI and hint at the exciting technological evolution ahead, leaving listeners eager for the next installment.
AI Snips
Chapters
Transcript
Episode notes
AI as Prediction Engines
- AI models primarily act as prediction engines generating outputs based on learned data patterns.
- They infer likely answers without true independent reasoning or domain understanding.
War Story: Automation Risks
- John Dickson shared a 1991 war story about the HARM missile shooting unintended targets due to internal databases.
- This highlights early automation risks predating AI concerns.
AI in Security: Potential and Cost
- AI security applications can help reduce false positives and automate triage, saving analyst time.
- However, current compute costs limit feasibility to very large or dedicated service providers.