Prompt Engineering

I had never heard of "prompt engineering" until this week.

Prompt engineering is the process where you guide generative artificial intelligence to generate outputs. Generative AI needs detailed, well-constructed instructions if it is going to give you high-quality results.

IBM, MIT, Purdue, Microsoft, and others offer training on how to do prompts, i.e. "engineer" them. If you take and pass their classes you can put it on your resume.

Each generation needs to master the technology of its era. My great grandparents needed to know how to farm with horses. My grandparents needed to know how to patch the tubes on car tires. 

IBM Selectric II, Correcting

As a college student I needed to know how to correct typos using coverup tape. The IBM Selectric with its spinning ball and eraser tape did it for me in my first jobs out of college. 

Teamster

Teamsters of old needed to know how to direct teams of horses that were in front of them, mostly with a light touch of reins laying across the horses' shoulders, a valuable skill for centuries. It is a valuable skill to prompt an AI program, for now.  Our guest post author warns that soon AI will intuit what we are asking it to do.

Until then, prompt engineering.

John Coster uses artificial intelligence in his work. He leads technology strategy and innovation teams at a large wireless telecommunications company. Over his 40-year career, he oversaw the design and construction projects for large energy users.

Coster

Guest Post by John Coster

There is much written these days about the disruptive technology that hit the world last year: generative AI; especially large language models (LLM). Early dismissive comments were silenced when a NY Times article , transcribed parts of a disturbing "conversation" with Microsoft's "Sydney". It's worth reading the entire article if you want to be creeped out. Sydney seemed volitional, sentient or even sapient. In slightly over 12 months, AI has taken the world by storm, and not just LLMs. AI has not only infiltrated more areas of our lives, but it has also created whole new domains that previously did not exist. Who needs porn stars when you can customize a realistic avatar to your liking? Who needs to know how to construct any kind of essay when you can ask AI to write it for you? 

Opinions about the societal implications or even what it means to be human are all over the board. History has shown that technology and automation always displace human workers, but up until now, they have not threatened knowledge workers, including highly paid software engineers. That is no longer true. So, the question on nearly every early-to-mid career professional person's mind is "what skills do I need to flourish or even survive in the new AI-driven economy?" 

NVIDIÅ GH 200 Grace Hopper Superchip

I work in the tech sector at the intersection of the physical and digital. My team designs and builds data centers that house systems that enable this kind of technology. Nvidia is the new darling with a market cap of three trillion dollars because it is the only company (so far) that can make the chips that handle the kind of computational workload that AI demands. It feels like a gold rush right now. Hundreds of billions are being invested to develop the most disruptive technology since the Internet.  

So, what does it all mean? One of the new skills being touted is "prompt engineering" or knowing how to ask AI the best questions to get the best outcomes. Many large companies hold training on how to best do this. Our company strongly encourages us to use AI to write performance reviews. I hear other large tech companies have mandates to do the same. My experience was quick and easy, and all I did was clean up some of the clunky wording. But many of us wonder if prompt engineering is even a real or enduring skill, or will our queries over time simply train AI to anticipate what we need before we ask? What does this portend for critical thinking skills at a time when we most need them?

This seems like the one technology that is racing faster than we can assess the legal, technical, ethical, economic or social implications.

UPCOMING POSTS: World War II veterans. AI clarifies Judge Juan Merchan's jury instructions. The Hunter Biden trial. More conversations with Republican defenders of Trump. We live in interesting times.