3 minutes; 1 minute for Overview and bold text.
Overview
My first AI/LLM blog post explains how AI/LLMs work, shows that they will replace many managers while enabling competent adopters to prosper, and outlines strategies. The hypersonic pace of change–literally by week–buttresses that post’s thesis. This second post further articulates strategies and outlines specific tasks that DISC has started, and you should too.
Future posts here will articulate AI/LLM trends in web marketing. One crucial point to remember is that the ~3-year Entity–MUM revolution in search marketing remains unaffected by the AI/LLM revolution. Both AI/LLM and Entity-MUM SEO entail collective human thought and responses. As with ~3 decades of SEO, the underlying rules by which ranking operates are vastly more sophisticated than the rules dictating actions we take. Search marketing–and all marketing is about people searching for solutions–now returns to timeless marketing precepts developed over several decades by leading ad agencies and business schools.
AI/LLMs encompass the sum of all knowledge. Maybe not the entire Library of Congress and all professions’ publications yet, but close enough, and soon. AI/LLMs can learn to innovate, and much faster when guided by industry professionals with AI/LLM literacy. Merging core databases with live web content, AI/LLMs will encompass all relevant to a task.
Footnote: LLMs don’t process submitted web pages or files with the intelligence of their pretrained database wherein something like wisdom renders mere information, but this weakness is improving steadily. Core Pretraining runs happen in multi-month intervals. When a prompt (question) to an LLM requires it to process a live URL or file you uploaded, care must be taken to ensure the LLM uses that “retrieval augmented generation” (RAG) optimally, and even then errors and ommissions sometimes happen.
Making Custom GPTs to Achieve Goals
On 11/6/23 OpenAI introduced GPTs as purpose-focused zones any subscriber can make. This entails inputting your firm’s documents, writing executive guidance just like for a human manager, and tuning knobs in ChatGPT’s output rules. It’s like hiring an executive to head up an initiative, such as optimizing PPC or SEO, or hiring and onboarding, or vetting grant applications–anything. Echoing my first post, Can AI Replace Upper Management? Lines in the Sand, this process indeed replaces upper management labor. But not all of it, as expertise is needed to choose, set up, tune, deploy, and error-check GPT’s professional output.
The amount of your firm’s data and documents that AI/LLMs can digest is expanding rapidly, but we’re far from complete inclusion of firms’ cloud content. Methods exist for managing this lack, like “Sparse Priming” and methods for mimicking human memory. Even without those methods, ChatGPT can now digest over 100,000 words along with images and data.
DISC is now setting up GPTs to improve all we do for clients. Client can adapt these GPT schemas to serve various business processes and objectives.
An Example Based on SAW
“SAW” is DISC’s SEO Writing & Architecture guidance we’re wrapping up for clients this year. In past years, DISC delivered such guidance in docs that enabled clients to write high-ranking content. (A lookback: around two decades ago, when SEO was new-ish, DISC sold many “CMS-SEO” guides for several thousand dollars each–like buying a $7000 article!) Now clients will be able to deposit text and image content, and soon videos too, into a clone of DISC’s upcoming SAW GPT in order to produce content according to the SAW rules. In other words, you won’t have to load all the rules into your brain’s RAM, nor frequently refer back to the guide, in order to make content that ranks and converts highly.
Sources
Henceforth DISC’s documents will have this Sources appendix, to point to further learning and indicate how much was written by an AI/LLM vs. DISC people. AI/LLMs don’t know what they “forgot” to consider, so professionals need to provide proper priming and parameters. Therefore documents like this should show that professionals stayed closely involved. Also, professionals must keep using their minds. Imagine a college grad who never wrote a paper: he would lack the muscular mental agility needed to manage real-life business.
- I wrote this entirely myself, but proofread it via three top AI/LLMs; they made many good suggestions but also missed a few weaknesses I resolved.
- Vast research informs this post. Here’s advice on sources, updated 8/22/24:
- In general, if a Youtube video sports a young face with a big, exaggerated expression, pass on it. A few of them are good, but most are over-hyped. Instead, follow established leaders. IBM has good ones, and recently Google started a great “Deepmind” series. (Google has been forced to damage its retail-level Gemini LLM in response to big gov pressure to lie, though this seems to be improving as of late 2024).
- Months ago I recommended https://www.youtube.com/@DaveShap, and several of his 2023 videos remain worthwhile, but by mid-2024 his interests became more tangential. Occaisionally he still posts great stuff.
- Periodically check https://openai.com/blog.
- DISC is preparing a more complete resource list to reveal in a future blog post.
See the 3rd post in this series: Deploying GPTs for Web Marketing. –>
