·
Christopher S Penn
“Consumer spending isn’t driving company profits as much as reducing expenses and improving efficiency. That could be a problem.” writes the Wall Street Journal about this season’s earnings reports. Gee, you think? Anyone who’s spent more than a week in an Econ 101 class learned that you cannot cut your way to growth, and “lean
·
Christopher S Penn
At a private company workshop yesterday (I do those btw), one of the attendees asked a very salient question: “given how much AI is taking over individual tasks, what are the meta skills that people need to have to succeed in an AI first world?” The answer for this is simple but not easy. I
·
Christopher S Penn
Here’s a very blatant example of gender bias baked into an AI model. Each week, Katie Robbert reads aloud the Trust Insights newsletter. In preparation, I take the raw video, transcribe it word for word with NVIDIA Parakeet, and then give to Gemini 2.5 Pro to turn into a YouTube description. Now, to be clear,
·
Christopher S Penn
This excellent short video by Cassandra J. Klepac on TikTok is another great use case of what not to do with AI: use AI-generated models in an article about beauty standards. The prompts and the image generation are not inherently harmful themselves in the abstract – but the use of AI in this particular use
·
Christopher S Penn
What if AI is the new radio? This may seem like an absurd analogy, but let’s take a step back 100 years. How did we do marketing a hundred years ago? We had the radio, we had the newspaper, we had word of mouth, we had some direct mail and catalogs. But the big star
·
Christopher S Penn
Deep research isn’t just research. It’s one of THE most powerful tools you have access to, because it’s really just an agent. Here’s a simple example. This morning, I was troubleshooting the marketing automation software we use at Trust Insights. I was getting system errors (it runs on a self-hosted server) and I had a
·
Christopher S Penn
If you blocked OpenAI’s web crawler from training on your data, your site is not listed as a source in ChatGPT Search. That’s the new Faustian bargain. If ChatGPT Search catches on, then you’re not included. Exhibit A: asking ChatGPT Search about the 1924 NY Times Op-Ed about immigration (which does exist). No sources listed.
·
Christopher S Penn
Imagine you’ve just walked into a vast public library. You walk up to the librarian, and you say, “I’d like a word, please.” “With the manager?” “No,” you say. “A word from your books. In fact, a word and all the words near it in your books.” “What word?” “How about the word Trust?” you
·
Christopher S Penn
There are three kinds of data. Data that is correct. This is a rare beast, but it does happen. Data that is predictably wrong. There’s a measurement error, but you know what it is and you can compensate for it. Data that is unpredictably wrong. There’s a measurement error, and you don’t know what it
·
Christopher S Penn
People need to see use cases to understand a new technology. People who understand the technology often don’t have the domain expertise, so it’s difficult to create the use case. That’s why so many of the use cases you see publicly for generative AI are so… mundane. Write some blog posts. Write some emails. Write
