Evening Reading - March 13, 2025

It's nighttime in America, which means it is time for Shacknews Evening Reading.

2

Hey Shacknews, it's time for Evening Reading. Let's officially close out our day of posting. Please take a look.


In case you missed it at Shacknews:

the man with the briefcase on Cortex

And now... Other Stuff From The Internet!!!

Steve from Gamers Nexus asks the AI question many of us have been wondering about

the man with the briefcase on Cortex

Be sure to watch the whole Gamers Nexus video here on their YouTube channel.

Their video on 50 series PhysX support is also excellent. 


Kevin Hart returns to Hot Ones this week

Congratulations to the show on its 10th anniversary.


Rock music played on untraditional instruments

Freebird!

You're in the jungle, baby.

Bar codes being turned into music kinda fits here too. 


British people infuriate Italian chef again

:)

I had to dig up my all-time favorite Gino rant.


Some people just don't get Mario...

That paper is a solid B+ in my book.


Couple who met in League of Legends get buried by this comedian

That punchline was ruthlessly hilarious. 10/10, no notes.


Memester Hunter Wilds

Who doesn't want to recreate their special someone in their favorite video game?


Two veteran game devs weigh in on AI

Tim Sweeney tweeted a pretty extensive breakdown of his experiences with different chatbots.

Here's the whole thread combined:

The reason for the disconnect is pretty apparent when you dig in. AI is a godsend for most of the programmers in the world, but it doesn’t help those doing work which lacks numerous similar independent implementations for AI to train on sufficiently to help.

So most researchers and developers on the leading edge of a field of technology find it useless for their work, and most programmers doing rote implementation find it incredibly useful, leaving each group thinking the other group is crazy.

Personally I’ve tried throwing about a dozen problems at AI to see what advice it can offer, and it ranges from Wikipedia levels of understanding to the confidently yet ruinously erroneous. Graph theory, linear algebra, functional logic, iterated function systems: no real help.

(person asks which models he has tried)

The three top OpenAI models, Grok 2 and 3, and several Llama variants. They all had surprisingly similar competencies at summarizing information and similar failings in reasoning one or more steps beyond what’s in the literature.

Ultimately with each problem I ended up caught in a loop of it confidently giving incorrect information, me pointing out the error precisely, and it apologizing and confidently giving a new answer, still wrong but in a new way.

Still, Grok3 is a better search engine than Google or Duck Duck Go, a better encyclopedia than Wikipedia, and a better health advisor than any web site.

My hypothesis is that we won’t have AGI from just this real-time autocomplete model of generative AI. Deep thought and rumination are going to be needed to explore numerous possible avenues for solving not-previously-seen problems.

One more bit to add. I’ve read the most relevant papers and publications on these topics as well as online discussions, and it’s clear that all of these AI models are trained on them specifically. The Reddit connection is the most annoying, in that AI treats confused Reddit posts from users speculating about or just plain wrong about topics as similarly authoritative to academic papers on the topic. And it’s very easy to spot, from prose characteristics, when it’s Reddit regurgitation rather than facts. What AI clearly lacks here is the sort of training time contemplation that would lead a person to draw conclusions about the reliability of sources in order to derive a consistent world model. And it’s clear that text regurgitation isn’t doing anything like what mathematicians do before stating a theorem (such as proving it either formally or informally).

Here's what George Broussard had to say:

And a follow-up tweet:


Kendrick's Korner

Shout out to Eat Your Broccoli for posting this sweet gameplay of Not Like Us in Beat Saber!

Super Mario Sunshine x Squabble Up = High Quality Entertainment.


Other quality videos from across the World Wide Web

Awesome!

Bollywood is usually good for some laughs.

That's pretty funny, for a drive-thru video that didn't involve anyone being hurt or attacked.

This guy is the Don Rickles of making fun of people's cars.

Why?

We are all living in Doechii's world right now.

That's a very sweet detail in The Sims 3.


Much Madonna Music

the man with the briefcase on Cortex

Check out my Shacknews Chatty thread breaking down the Top 10 Madonna songs of all time. Reply with your favorites for maximum engagement.


There you have it, Shacknews. Your Evening Reading for March 13, 2025. Please consider playing our first game Bubbletron for free today to support our broader Shacknews efforts.

the man with the briefcase on Cortex

What are you up to tonight? Let us know in the Shacknews Chatty comment thread below.

CEO/EIC/EIEIO

Asif Khan is the CEO, EIC, and majority shareholder of Shacknews. He began his career in video game journalism as a freelancer in 2001 for Tendobox.com. Asif is a CPA and was formerly an investment adviser representative. After much success in his own personal investments, he retired from his day job in financial services and is currently focused on new private investments. His favorite PC game of all time is Duke Nukem 3D, and he is an unapologetic fan of most things Nintendo. Asif first frequented the Shack when it was sCary's Shugashack to find all things Quake. When he is not immersed in investments or gaming he is a purveyor of fine electronic music. Asif also has an irrational love of Cleveland sports.

From The Chatty
Hello, Meet Lola