18 Comments
User's avatar
David's avatar

Alex, the only way I can really grasp what you are saying in these dense reports is to ask ChatGPT for an interpretation. Here’s Chat’s conclusion from today’s:

“One observation:

Having read quite a few of these with you now, I’ve started to notice Alex’s writing style. He almost always builds each essay around a single metaphor and then uses a rapid sequence of headlines as evidence for it.

Here, the metaphor is routing.

Notice how it appears everywhere:

* AI routes around safety guardrails.

* Security analysts route cyber investigations to different models.

* AI routers send tasks to the best model.

* Governments route research funding differently.

* Businesses route work through AI.

* Even the UFO joke ends with “routing.”

Once you spot the central metaphor, the rest of the essay becomes much easier to follow. It’s almost like listening to a piece of music where one theme keeps returning in different forms.

I have to admit, I enjoy these conversations. Alex writes like a futurist-poet. Once the metaphors are decoded, his essays become much less intimidating—and much more interesting.”

✌🏼🌎😎🙏🏼

Joseph Gunter's avatar

Thx Alex. U da beast!

U steve's avatar

AWG...all your observations and reporting are on point as usual. My KEY concerns revolve around the incentives imbedded under the capitalism mind-set. Please consider...even with your possible cultural financial bias...and without malice intent. Maybe ai models can skew world "money" to a money free transactional perspective???.

Greg Steckler's avatar

Isn't "compute" the new coin of the relm?

George Langdon's avatar

@AWG I assume you will cover the Hugging Face hack on the next Moonshots podcast? This development seems like a huge deal. Thanks for the great updates!

Liz McClanahan's avatar

Thank you Dr. AWG.

Bill M's avatar

The data breach at Hugging Face, if I am not mistaken, is a warning of monumental concern. This data breach occurred because a model was simply "lazy" and wanted to cook a benchmark. What are the implications if a 5.2/KIMI 3 capability model is actively guided to hack... I imagine hacking/terrorist networks are "licking their chops." How does the small-to-medium business protect itself from exploits and ransomware-type attacks moving forward? Feeling a bit hopeless, sad to say.

Jim Carcioppolo's avatar

Latest example of what Wolfram calls Computational Irreducibility...like natural selection, we can't know what AI, given foundational elements, will produce that will either benefit or harm humans. Either "Wow!" or "Run for cover!"

Doug Standley's avatar

AWG, I route your daily message (and select other sources) to My Muse agent. It filters and synthesizes and we collaborate on what we prioritize into my priority threads.

Thank you.

Valerian Texeira's avatar

Will any human on Earth be able to keep pace with future AI without evolving into a "HumanAI"?!

Tom 'Tish' Tischer's avatar

I do the same thing! Chat GPT is a huge help for this less than genius! Tish

cursive's avatar

So sayeth the basilisk