Skip to main content

We built standard
Responsible AI Programs for over six years...

Then We Took a Step Back To See
Where We’ve Been and Where Things are Going

Where AI is now: LLMs + Tools

LLM Video Generating AI The Internet Enterprise Database LLM LLM Video Generating AI The Internet Enterprise Database LLM

What tomorrow's AI looks like: LLMs + Lots of Tools + Autonomy = Agentic AI

Database 1, 2, 3, 4, 5, 6, 7, 8, 9, ... Narrow AI 1, 2, 3, 4, 5, 6, ... Generative AI 1, 2, 3, 4, 5, ... LLM + Autonomy = AI Agent The Internet Accurate, objective, sane part of the internet Inaccurate, biased, insane part of the internet Database 1, 2, 3, 4, 5, 6, 7, 8, 9, ... Narrow AI 1, 2, 3, 4, 5, 6, ... Generative AI 1, 2, 3, 4, 5, ... LLM + Autonomy = AI Agent The Internet Accurate, objective, sane part of the internet Inaccurate, biased, insane part of the internet

AI's ultimate destination: Clusterf&!k AI

Database 1, 2, 3, 4, 5, 6, 7, 8, 9, ... Narrow AI 1, 2, 3, 4, 5, 6, ... Generative AI 1, 2, 3, 4, 5, ... Agent #1 The Internet Accurate, objective, sane part of the internet Inaccurate, biased, insane part of the internet Orchestrating AI Database 11, 12, 13, 14, 15 Narrow AI 1, 2, 7, 6, 9 Agent #2 Generative AI 4, 5, 7 The Internet 3rd-Party Orchestrating AI Agent ?? ? ?? ? ?? ? Database 1, 2, 3, 4, 5, 6, 7, 8, 9, ... Narrow AI 1, 2, 3, 4, 5, 6, ... Generative AI 1, 2, 3, 4, 5, ... Agent #1 The Internet Orchestrating AI Database 11, 12, 13, 14, 15 Narrow AI 1, 2, 7, 6, 9 3rd-Party Orchestrating AI Agent Agent #2 Generative AI 4, 5, 7 The Internet ?? ? ?? ? ?? ?

The Evolving AI Risk Landscape

Narrow AI Generative AI (Multi-Agentic) AI Predictable contexts = easy testing Unpredictable = difficult testing Oh no Data scientists responsible End-users more responsible Utterly unclear Human in the loop, straightforward HIL needs more training Crumbling Monitoring/intervention straightforward More difficult Worse Clear when to perform risk assessments Difficult to determine Much worse Communication is easy Complicated communication Much much worse
  • Narrow AI Predictable contexts = easy testing Data scientists responsible Human in the loop, straightforward Monitoring/intervention straightforward Clear when to perform risk assessments Communication is easy
  • Generative AI Unpredictable = difficult testing End-users more responsible HIL needs more training More difficult Difficult to determine Complicated communication
  • (Multi-Agentic) AI Oh no Utterly unclear Crumbling Worse Much worse Much much worse

So We Created the Ethical Nightmare Challenge

A rapidly implementable and scalable solution to avoiding the worst of AI

Learn More About Our Solution