INFORMS Open Forum

Over the past two years INFORMS has been chasing the AI  moving target hoping  INFORMS can get a few mentions.  Recently a large amount of material has been pivoting off of Herbert Simon's Two Heads Paper.  If you attended the presentation - Prof Simon wasn't saying move away from your roots, but broaden your horizon.

A common discussion in SCM is agents to make a firm more responsive - inventory, plain repair, better schedules , etc.   Better requires thoughtful model and analytics.  This requires analysis and computational intelligence.  This doesn't come from the AI Agent Genie or some LLM that has digested all of the NYTIMES.  Inventory is risk assessment - heart and soul of probabilistic OR.   The standard simple AI agent to repair a plan is - Have one AI agent recognize a new order arrived.  This agent asks in natural language the other agent if capacity is available Which responds yes or no.  As any one who has lived in the trenches knows -  checking capacity is not sufficient, immediately one needs to check material.  If there isn't capacity, then can capacity be freed up - if the order is important enough. Plan repair requires building a soft peg of the solution and then reasoning about his network.

INFORMS needs to drive our organization as experts in computational intelligence and drive the discussion.  If not we will repeat the mistake of the 1980s

Thanks for the post, Ken. I think you are right - INFORMS needs to capture the center as an authority on computational intelligence much as it did for analytics.  Perhaps some folks could start working on this.  Not me -I recognize I'm too old to contribute. It's the young experts who should start weighing in.

Hope they pick up the challenge.

Over the past two years INFORMS has been chasing the AI  moving target hoping  INFORMS can get a few mentions.  Recently a large amount of material has been pivoting off of Herbert Simon's Two Heads Paper.  If you attended the presentation - Prof Simon wasn't saying move away from your roots, but broaden your horizon.

A common discussion in SCM is agents to make a firm more responsive - inventory, plain repair, better schedules , etc.   Better requires thoughtful model and analytics.  This requires analysis and computational intelligence.  This doesn't come from the AI Agent Genie or some LLM that has digested all of the NYTIMES.  Inventory is risk assessment - heart and soul of probabilistic OR.   The standard simple AI agent to repair a plan is - Have one AI agent recognize a new order arrived.  This agent asks in natural language the other agent if capacity is available Which responds yes or no.  As any one who has lived in the trenches knows -  checking capacity is not sufficient, immediately one needs to check material.  If there isn't capacity, then can capacity be freed up - if the order is important enough. Plan repair requires building a soft peg of the solution and then reasoning about his network.

INFORMS needs to drive our organization as experts in computational intelligence and drive the discussion.  If not we will repeat the mistake of the 1980s

I had to look up the terms being used, but as I understand the idea of computational intelligence (CI) there are many INFORMS members and personal colleagues that I know using CI as a regular tool in their very human toolkit.  And it is definitely true that traditional AI including modern LLMs do not handle this type of thinking particularly well.  Although, they could.  ChatGPT game me this opening definition:  "Computational intelligence (CI) is a branch of artificial intelligence (AI) that focuses on building adaptive systems capable of reasoning, learning, and making decisions in complex or uncertain environments."  To the SCM example, we need experts to train AI agents alongside the AI experts.  Although as I've read your comments, I wonder how to we fundamentally get AI systems to deal with new, uncertain, and poorly defined problems.  I hope this conversation gains some momentum and creates good dialogue.

Thanks for the post, Ken. I think you are right - INFORMS needs to capture the center as an authority on computational intelligence much as it did for analytics.  Perhaps some folks could start working on this.  Not me -I recognize I'm too old to contribute. It's the young experts who should start weighing in.

Hope they pick up the challenge.

Over the past two years INFORMS has been chasing the AI  moving target hoping  INFORMS can get a few mentions.  Recently a large amount of material has been pivoting off of Herbert Simon's Two Heads Paper.  If you attended the presentation - Prof Simon wasn't saying move away from your roots, but broaden your horizon.

A common discussion in SCM is agents to make a firm more responsive - inventory, plain repair, better schedules , etc.   Better requires thoughtful model and analytics.  This requires analysis and computational intelligence.  This doesn't come from the AI Agent Genie or some LLM that has digested all of the NYTIMES.  Inventory is risk assessment - heart and soul of probabilistic OR.   The standard simple AI agent to repair a plan is - Have one AI agent recognize a new order arrived.  This agent asks in natural language the other agent if capacity is available Which responds yes or no.  As any one who has lived in the trenches knows -  checking capacity is not sufficient, immediately one needs to check material.  If there isn't capacity, then can capacity be freed up - if the order is important enough. Plan repair requires building a soft peg of the solution and then reasoning about his network.

INFORMS needs to drive our organization as experts in computational intelligence and drive the discussion.  If not we will repeat the mistake of the 1980s