“Intelligence”, Memory, Context & Happiness . . .

Ralph Aboujaoude Diaz makes me laugh a couple of times week, sometimes deliberately! 😉

He wrote recently, that according to Charles Bukowski, “Sadness is caused by intelligence, the more you understand certain things, the more you wish you didn’t understand them.”

That is quite philosophical and plausible!

He leads on to the call to action “Please, let’s drastically accelerate the development of Artificial Intelligence so that machines can finally become super intelligent and take our inescapable sadness away. And let us, humans, be dumb and happy forever.”

That works for me. Thanks Ralph!

At an even deeper and more practical level, on exploiting intelligence, Mark Stouse wrote a great piece that is worth a read AND which may well help your own effectiveness and even happiness, without resorting to Ralph’s recipe!

Why Some AI Systems Remember — and Others Don’t

Mark highlights something in GenAI / LLMs which may not be obvious.

Some have memory and some do not. 

• Claude is stateless and doesn’t remember anything you’ve taught it from one conversation to the next. By choice. By design. No persistence, maximum privacy, minimal long-term risk.
• By contrast, ChatGPT and Perplexity allow you to instruct them to remember things across conversations. They can retain context.

This is not mere “feature wars”. It has deeper implications.

Memory is relevant to both facts and context. And as we know, “context is king”.

LLMs are not neutral. They embed trade-offs between privacy, continuity, accuracy, and governance.

Mark offers us a choice of approaches;

• Stateless AI → safer for privacy, but limited in compounding value.
• Memory AI → essential for continuity, but requires strong governance of what gets remembered.
• Peer Review AI → best practice for critical decision-making, because it exposes errors no single system can catch.

The latter is a fascinating, and potentially groundbreaking, approach to making AI really work effectively.

He calls it the “The Peer Review Advantage”.

Why wait for the mythical perfect LLM when you can have “peer review” across multiple models?

“Each AI has its blind spots. By running the same problem through different systems — some with memory, some without — you surface divergence. Divergence is signal. It tells you where to dig deeper, where assumptions are shaky, and where governance needs to step in.”

Mark has developed an AI-moderated “peer review” system to minimize or even eliminate hallucinations (“screwups”).

He describes running three GenAI tools simultaneously on the same prompts and THEN gets them reacting to what each other had written . . . . . creating an “AI roundtable conversation”. All three tools got to respond to questions, and then each one got to respond to the other tools’ answers.

We are all challenged with the need for verification of generated content. This is a big step forward . . . .

On the funny side, it turns out that the tools can often recognize each other’s content and can get a bit “spiky” with each other . . .  😉

If you want to use GenAI to be more than just chatbots and be instrumental in real business decisions, Mark’s approach might  might just create more success and even happiness, rather than frustration!

But of course, we should remember Ralph’s wise words too. “Intelligence is overrated, sadness is optional, and happiness is so much easier when you don’t think”.

You can read Ralph’s take on intelligence here . . .  

Mark’s thought-provoking article “Why Some AI Systems Remember — and Others Don’t” can be read here . . .     and the related story of the development his peer review strategy here . . . 

Thanks for reading . . . 

“Intelligence”, Memory, Context & Happiness . . .