Korva is a scroll of the internet’s best ideas.
Sell AI by what it does
Before AI was real and part of everyday life, the technology industry and the labs used fear-based marketing to raise the money to build their models. By the middle of last year, that approach looked headed for trouble, and public opinion bears it out.
The alternative is to talk about what the product can do for the person using it. A model release usually arrives with a blog post, statistics, and benchmarks. Picture an announcement that you can now ask for a custom piece of furniture, get the exact image of what you want, and have it made for less than you would pay elsewhere. Every release brings use cases like that, each one a particular problem a person can now go solve.
How do fear appeals change behavior?
Psychologists have studied fear appeals since a 1953 Yale study by Irving Janis and Seymour Feshbach, which used messages about dental hygiene. They found that a mild appeal changed behavior more than a strong one. Later, Kim Witte's extended parallel process model of 1992 proposed that fear leads to action only when people also believe they are able to respond.
Why do AI model releases lead with benchmarks?
Benchmarks are standardized tests that let labs compare models on identical questions, which makes progress easy to chart. MMLU, a widely used example introduced in 2020, tests knowledge across 57 subjects, from law to mathematics. Because scores are single numbers, they fit neatly into announcements and charts.
Why do use cases persuade more than specifications?
Marketing professor Theodore Levitt is credited with the line that people do not want a quarter-inch drill, they want a quarter-inch hole. Clayton Christensen later developed this into jobs-to-be-done theory, illustrated by a study of commuters who bought milkshakes as a way to pass a long drive. Christensen and his co-authors described the approach in the 2016 book Competing Against Luck.
More from The Prof G Pod with Scott Galloway
