Uncle Mike's 1985 AI Presentation

The DEC Years

Uncle Mike already had 12 years at DEC in 1985. Digital Equipment Corporation (DEC) was the NVIDIA+Microsoft of the A.I. world back then. They made the “cloud servers” that AI researchers from MIT, Stanford, & Carnegie Mellon developed expert systems and Lisp environments on.

Expert Systems & Overselling

The first commercial expert systems (like XCON, used internally by DEC to configure computers) became legendary — it saved DEC tens of millions annually and was one of the first real-world AI success stories. Uncle Mike still had the foresight to warn about overselling AI.

Not About Replacing Jobs

A big part of the selling process, even back then, was making clear that AI was not set up to take people’s jobs, and should not be adopted just for AI’s sake. “Deciding to get into A.I. should not be based on ‘interest’ in A.I., but should be based on business goals & needs…”

Change Management

This set of slides is my favorite:

• “Change is the real goal: commitment to change is therefore a real prerequisite"
• "Success in technology transfer depends more on managing the expectations of people than on managing the technology itself"
• "Give enough time for the significance of A.I. technology to ‘sink in’…”

Banger One-Liners

Throws out some banger one-liners:

• “Understanding and capturing expertise is now the most important part of the job"
• "If you automate a ‘mess’, all you get is an ‘automated mess’"
• "Set goals to do ‘more with the same’, rather than ‘the same with less‘“

Skeuomorphism, Before the Word Existed

Then warns about skeuomorphic design patterns a full decade before Mark Cuban put ROI (Radio on Internet), and 28 years before Jony Ive launched iOS 7 to explicitly abandon skeuomorphism.

Getting into A.I.: Some of the Risks

”A.I. is often viewed as the probable solution to whatever is the current largest unsolved problem in an organization.”

Understanding AI

”In other words, people who are neither familiar with traditional software technology nor with the ‘expert’ knowledge in an area will not understand or appreciate what A.I. technology is doing for them.”

Model Performance

Build with improvement in mind. “Performance issues are often temporary: with new software and systems being released every year, and with the cost per ‘MIP’ coming down every year, performance issues can often be solved without rewriting the AI application.”

Don’t Oversell It

”If you oversell A.I., over-emphasizing the popular ‘sensational’ and imaginative claims for A.I. technology, the users in your organization will almost certainly become disappointed.”

Technology Transfer

”Technology transfer, whether it is office automation or artificial intelligence, is always more successful if it is pulled into an organization rather than being pushed by the software engineers.”

The Closing Slide

Chat, is it too late for us?

”You can buy the tools for change, but you can’t buy the environment for change.”