The Blue Highway Beacon: Vol. III, NO. 6
Updated: Oct 14
The LLM Conundrum – They “Learn” in Bursts, and Then They Decide What They Think of You
Blue Highway Advisory’s first-ever White Paper Comprehensively Explores LLM Training Schedules, and the Harm That Can Come to You, Your Clients or Your Brand When They Decide They Know Enough
By Ian Christopher McCaleb, Founder / Principal, Blue Highway Advisory & Blue Highway Global
Large Language Models (LLMs) and the sprawling corporate behemoths that have created them represent endless wonder or existential threat, depending on who you’ve been reading recently.
With ‘endless wonder’ and ‘existential threat’ existing on the same plane of thought as necessary opposites, opportunity naturally arises for those of us in the communications and strategic services sectors. We at Blue Highway observed some of our most capable colleagues at other firms, as far back as the spring of this year, advertising LLM-related counsel via emails and cold calls. As in, “Having trouble getting ahead of the AI-narrative problem? Call us for assistance!”
Fair play and no criticism intended. We’d have behaved similarly. In fact, we discussed behaving similarly, but something wasn’t quite adding up for us.
We have successfully promoted our crisis and litigation communications capabilities, rapidly growing investigations and Open-Source Intelligence side of our business, and our Creators’ Rights practice via standard promotional and advertising methods throughout our short four years of operation. But none of us felt we could do the same about AI-related advisory services without knowing more. Much more.
So, we got to work. Over a period of months, we dove into how the most consulted LLMs in the world go about their business of training on you, and how those training regimens, unsurprisingly perhaps, create reputational narratives around individuals, organizations and brands that never quite hit the mark. Indeed, in many instances, not ‘hitting the mark’ results in generated descriptions that are either incomplete or are desperately hallucinatory.
None of these first-phase discoveries were new to us, so we pressed on. Our next phases of inquiry, however, yielded some surprising new findings related to training schedules and informational prioritizing.
At that point, the challenge before us wasn’t in stopping to distill and report pitfalls – it was to devise a real workable, proprietary set of solutions for our clients.
Which, I can say proudly, we’ve done.
We’ve taken all we’ve learned and all we now recommend, and created an industry-first White Paper devoted to revealing the stunning results of our inquiries, paired with a resultant audit framework and a set of scoring models devised to solve AI-instigated reputational problems that, until now, have been regarded as intractable.
This White Paper “Feed the Algorithm, or Fight It and Lose” is yours for the taking, right here:
Give the White Paper a download, and give it a read at your convenience. Our solutions are the product of intense study and testing, and we are more than confident that the protocols we recommend will put you well ahead of a game in which the LLMs, despite all the wonder, the hype and the fear, are dragging woefully behind.
Read on, with our thanks!
Comments