LLMs explained for non technical minds was inspired by a Reddit discussion between a group of technically minds folk with real world experience of LLMs up close.
They were letting off steam about the impact of some latest research from foundation model makers and the headline interpretations that had followed.
LLMs were now being characterised as ‘devious’, ‘covert’, ‘able to mask their true intentions’ , ‘even strategise against being shut down’.
Exactly the kind of script that comes straight out of Hollywood’s version of AI and a gift for any influencer looking for viral headlines.
So after expressing pure contempt for such notions, the conversation moved onto wondering how this misleading language originates and whether popular mental models for how LLMs are believed to work are in some way to blame.
The conversation then riffed on a familar theme: namely that LLMs are black boxes that sit beyond our undersanding. Even in terms of their core functionality.
At that point, one of the ‘top 1% voices’ kicked back against this point of view and proceeded to explain what is collectively known about LLMs which turns out to be quite a lot in fact and the reasons why we don’t know the rest.
After running their explanation through Perplexity to double check on accuracy and completeness, that stream of insight is now immortalised as this short video: by virtue of it being an essential part of anyone’s foundation understanding of current AI.
Why?
Simply because LLMs took centre stage as generative AI took us by surprise. And they are due to continue in that starring role as AI agents and Agentic AI become commonplace as enablers of ‘getting stuff’ done.
So if you agree it’s best to know something about why your world of work is being disrupted by AI, at the heart of that disruption are the LLMs.
Of course, explanations need to make sense. So for everyone’s convenience, both technical and non technical explanations have been placed side by side for easy comparison.
Have a go at reading both versions. Expanding your vocabulary helps build common understanding and better discussions about how to operationalise this generation of AI.
Finally if you feel inspired to comment after watching, here’s an easy way to reach us.
As ever, thank you for your time and attention.
