e428 — Is you is, or is you AIn’t my AI?

By | |
Comments Off on e428 — Is you is, or is you AIn’t my AI?
"Artificial Intelligence" printed on a piece of paper inserted in a typewriter.
Photo by Markus Winkler on Unsplash

Published 14 August 2023

Andy, Michael and Michael get together for a show full of generative AI: audio, bucket lists, books and agents.  Andy starts things rolling with a Nvidia follow on article from last week, and a throwback to e156 from 2016.  Then things start to get exciting with a number of generative AI examples, beginning with Meta’s AudioCraft, an open source example of generative audio.  Next up is Söka, a new app that allows you to create your own bucket list, with the help of generative AI to provide suggestions.  Each of the co-hosts gave this a whirl and share their interesting results.

Then, continuing with open source, Michael M brings forward an update from an April discussion during e412 where the generative agents that are essentially non-player characters (NPCs) that interact with one another powered by a large language model.  The Smallville environment is now open source, so you can give this experience a try yourself.

Next is a disturbing development where author Jane Friedman noticed that AI generated counterfeit books were being sold online and attributed to her.  In her blog post, Jane shares her experience in what it took to correct Amazon and Goodreads, and what other authors have experienced.

This leads to a discussion about Benji Smith’s prosecraft.io service which was taken down last week.  Check out the discussion in the podcast, and see what Benji wrote about the experience from the links in the show notes below.

Andy takes a moment to step up on the soapbox and share his thoughts on the meaning of AI.  And concludes the episode with a fun reference to a keycap that is designed to run Doom.  Check out the show notes for this below.

What are you thoughts about AI, generative AI and algorithms overall?  What would you generate for your bucket list(s)?  Have your bots 🤖 drop our bots 🤖 a line at @gamesatwork_biz (our home for now) and let us know! 

These show notes were lovingly hand crafted by a real human, and not by a bot.  All rights reserved.  That’s our story and we’re sticking to it.

Selected Article Links

Follow ons from last week

TechCrunch article: Nvidia CEO: We bet the farm on AI and no one knew it

Games at Work e156: AI for All

AI: Generative Audio, Lists, Books & Agents

Meta article: Open sourcing AudioCraft: Generative AI for audio made simple and available to all

Söka: The easy way to manage your bucket lists and discover new goals

ARXIV paper: Generative Agents: Interactive Simulacra of Human Behavior

Open Source Smallville https://github.com/joonspk-research/generative_agents 

Games at Work e412: 3D or not 3D

Ars Technica article: Author discovers AI-generated counterfeit books written in her name on Amazon

Jane Friedman blog post: I Would Rather See My Books Get Pirated Than This (Or: Why Goodreads and Amazon Are Becoming Dumpster Fires)

TechDirt article: The Fear Of AI Just Killed A Very Useful Tool

Gizmodo article: Fiction Analytics Site Prosecraft Shut Down After Author Backlash

prosecraft.io

Benji Smith’s The Shaxspir blog post: Taking Down prosecraft.io

Doom!

Hackaday article: Running Doom in a Keycap Takes Careful Work

Bonus links

Security & Privacy

Security Now! e934 Revisiting Global Privacy Control

Wired article: Leaked Yandex Code Breaks Open the Creepy Black Box of Online Advertising

Podcast: Play in new window | Download (Duration: 33:49 — 39.3MB) | Embed

Subscribe: Apple Podcasts | Spotify | Amazon Music | Android | Podcast Index | Youtube Music | RSS | More

e427 — DeepBarbie Fakery

By | |
Comments Off on e427 — DeepBarbie Fakery
35mm film negatives - the kind that required a great deal of skill to deepfake
Photo by Immo Wegmann on Unsplash

Published 6 August 2023

Michael and Michael get together for a show full of intrigue and deepfakes – pictorially and in text.  While Andy was not available to record this episode, he was a guest on the TechGrumps podcast recently, and you can check that out in the show notes below.

Staying with the Barbie theme, Michael and Michael begin this episode with a fantastic plastic representation of you called B-ai-rbie.  Through this site, you can create a Barbie or Ken doll representation from your uploaded photograph.  This example reminded the co hosts of a deepfake example that would create a portrait of an uploaded photo in the style of a famous artist.  Links to that episode and MIT Tech Review story are included below.

Hiding in plain sight is one of the oldest camouflage tricks in nature, and the next story for this episode leverages this concept.  To protect against deepfakes, tools like PhotoGuard and Glaze allow for an easy way for people to view visual content, but the images are rendered in a way to confuse AI models from using the images.  Michael and Michael talk about the cat & mouse ways that these technologies will continue to evolve, and how encryption services like PGP could also be brought to bear.

Another example of hiding in plain sight is designed specifically for machine learning algorithms to pick up on the hidden text.  The Schneier on Security article references a story from the Washington Post where job applicants are attempting to take advantage of the machine reading of their resumes by including additional keyword text in white font on a white background.  So, instead of the human seeing something that the machine cannot, the human disguises text to be specifically read by the machine and not by a human.

It is an out of the ordinary thing to talk about the Unreal Engine when it comes to IRL automotive companies.  For this episode, there is a story about how Ford is using the Unreal Engine to power their 2024 Mustang dashboard, allowing for the retro throwback dashboard of the Fox-body in the new vehicle along with further driver customizations.

Michael and Michael wrap up with stories about the upcoming second season of Loki and a link to 3D print your very own Futurama brain slug.

What do you think about protecting visual content via solutions like Glaze or PhotoGuard?  If you could customize your car’s dashboard, what would you add or subtract?  Have your bots 🤖 drop our bots 🤖 a line at @gamesatwork_biz (our home for now) and let us know! 

These show notes were lovingly hand crafted by a real human, and not by a bot.  All rights reserved.  That’s our story and we’re sticking to it.

Selected Article Links

Andy on TECHGRUMPS 3.06: CELLULAR, MODULAR, INTERACTIVODULAR!

Deepfakes

BaiRBIE.me

Games at Work e240: Game of Life

MIT Technology Review article: Turn selfies into classical portraits with the AI that fuels deepfakes

MIT Technology Review article: These new tools could help protect our pictures from AI

arXiv paper: Raising the Cost of Malicious AI-Powered Image Editing

Github: photoguard

Glaze

(more) AI

Wired article: A New Attack Impacts Major AI Chatbots — and No One Knows How to Stop It

Schneier on Security article: Hacking AI Resume Screening with Text in a White Font

Washington Post article: Job applicants are battling AI résumé filters with a hack

Nvidia has started asking large GPU buyers who their end users are (per The Information): theinformation.com/articles/in

Their focus seems to be on picking favorites in a crowded market, but this could actually be an interesting hook for "know your customer" diligence to prevent horrific abuses of AI, if they cared to do that.

— Yoel Roth (@yoyoel) 2023-07-31T20:42:38.083Z

The Verge article: Meta’s AI ‘personas’ might launch next month

Unreal Engine in 2024 Ford Mustang

USA Today article: 2024 Ford Mustang goes back to the ’80s in salute to a hero from Detroit’s darkest days

Entertainment

Mashable article: ‘Loki’ Season 2 trailer: Ke Huy Quan joins in on the time-travel fun

Printables post: Futurama Brain Slug

Podcast: Play in new window | Download (Duration: 30:24 — 35.4MB) | Embed

Subscribe: Apple Podcasts | Spotify | Amazon Music | Android | Podcast Index | Youtube Music | RSS | More

e426 — Barbenheimer Chic

By | |
Comments Off on e426 — Barbenheimer Chic
Imagined Barbie Dream Police Box set inspired by Natalie Stevens' work. Mentions that Ken-companion is available separately.
Inspired by natalie.3dblah’s Dream Vehicles, mashed up with Nick Fewings on Unsplash

Published 31 July 2023

Michael, Michael and Andy don their rose colored AR glasses for this this infused episode.  Topics this week include stories and toots about the Barbie and Oppenheimer movies released last week, AI deepfakes allowing dead artists to sing songs, Raspberry PI powered BlackBerry devices, Redditor pranks, repurposed Selectric typewriters and the developer kit for the Vision Pro.

The co hosts start things off this episode with all manner of imagined Barbie Dream vehicles shared by New Zealand based 3D artist Natalie Stevens from Battlestar Galactica to a triceratops.  Her models inspired the above Police Box image for the episode.  

Continuing on the theme, the co hosts talk about the stories for PalmPilot emulation for the IMAX showings of Oppenheimer.  This reminds Andy of a Beepy / BeepBerry – check out the image in the show notes below!  

Continuing on the Barbie theme while pivoting to AI, the co hosts marvel at Johnny Cash singing Barbie Girl.  While there is a discussion in the YouTube comments that Johnny Cash *could* have sung Barbie Girl while he was alive, it is highly likely that he did not.  This AI generated Johnny Cash provides a terrific performance, foreshadowing much more to come.

Following up on Ian Hughes’ listener link on Simulation led to the example of how their generative AI was used to craft a South Park episode that leverages deepfakes of actors in a story about using deepfakes of actors.  This Escheresque self referential treatment of generating a show using AI about shows using AI was just outstanding.  

Switching gears, though staying squarely in the AI space, the team discusses the Redditor prank about the World of Warcraft posts about Glorbo being amplified by AI bots scraping Reddit.  This reminded Michael M of the reason why Van Halen would include a requirement in their concert riders to have a bowl of M&Ms with “absolutely no brown ones.”  Could such a technique be used to ensure proper provenance in LLMs?  

Wrapping up this episode included a discussion on old and new hardware, specifically repurposing Selectric typewriters as a Linux terminal and the requirements for those that are interested in gaining access to Apple’s developer kit and hardware for the Vision Pro.

What would be your Barbie Dream vehicle?  Whose movie character would you want to replace with a deepfake Danny DeVito?  Have your bots 🤖 drop our bots 🤖 a line at @gamesatwork_biz (our home for now) and let us know! 

These show notes were lovingly hand crafted by a real human, and not by a bot.  All rights reserved.  That’s our story and we’re sticking to it.

Selected Article Links

Barbenheimer

Seen one of these posted without credit, so sharing some more here because they are amazing and I'd buy every single one of them if they existed.

Made by New Zealand model maker and 3D artist Natalie Stevens (natalie.3dblah). More of her work here:
instagram.com/natalie.3dblah/

— Annie Hsh :mastodon:🖖:cofe4: (@fringemagnet) 2023-07-25T19:23:06.434Z

Natalie Stevens’ Instagram

ToysREvil blog post: Barbie Dream Vehicles by Natalie Stevens

Ars Technica article: IMAX emulates PalmPilot software to power Oppenheimer’s 70 mm release

The Verge article: Here’s why the best IMAX movies still need a Palm Pilot to work

Beepy

AI

The Simulation

Fable Studio paper: To Infinity and Beyond: SHOW-1 and Showrunner Agents in Multi-Agent Simulations

Ars Technica article: Redditors prank AI-powered news mill with “Glorbo” in World of Warcraft

Smithsonian Magazine article: Why Did Van Halen Demand Concert Venues Remove Brown M&M’s From the Menu?

Games at Work e424: What’s AI got to do with it?

Board Game Geek Overview: Small World of Warcraft

Hardware

Hackaday post: SELECTRIC TYPEWRITER GOES FROM TRASH CAN TO LINUX TERMINAL

The Verge article: Apple is taking applications for Vision Pro developer kits

Ars Technica article: Devs aren’t allowed to let Apple’s Vision Pro dev kits out of their sight

Podcast: Play in new window | Download (Duration: 27:08 — 31.7MB) | Embed

Subscribe: Apple Podcasts | Spotify | Amazon Music | Android | Podcast Index | Youtube Music | RSS | More

e425 — Cone Heads

By | |
Comments Off on e425 — Cone Heads
traffic cones on top of a fire hydrant and posts on the side of a street
Photo by Jon Tyson on Unsplash

Published 24 July 2023

Michael, Michael and Andy are all back together for this opening weekend episode.  Topics for this show include stories about autonomous vehicles, new & old games including , , , and as well as a discussion about the laptop.  

Starting things off this episode, is an article from Schneier on Security detailing how some people are causing autonomous vehicles to stop driving by placing a code on the hood of the car.  The discussion on this topic reminds Michael M of the Harvey Mudd College prank policy (see show notes below).  

Next on the list is a game that blends photography of real objects with the in-game / in-world experience called Viewfinder.  Check out the YouTube video below to see how it operates.  Could there be a View-Master + Viewfinder integration sometime in the future?  Andy brings up Dave the Diver (#DTD) as a game he’s spending time on.  He’s having so much fun with it that it is preventing him from getting started with Baldur’s Gate 3.

Sticking with the games theme, the cohosts reminisce about Dungeons & Dragons.  They react to an article from PC Gamer that takes a strong position that all the newfangled electronics that replace imagination are a detriment to the game.

Wrapping up this episode with a discussion on the Framework repairable & upgradeable computer, Andy shares a couple of additional cool hardware links.

Are you a purist when it comes to playing D&D or do you take full advantage of technology?  Are you a fan of repairing your computer yourself?  Did you have a weekend?  Have your bots 🤖 drop our bots 🤖 a line at @gamesatwork_biz (our home for now) and let us know! 

These show notes were lovingly hand crafted by a real human, and not by a bot.  All rights reserved.  That’s our story and we’re sticking to it.

Selected Article Links

Cones

Schneier on Security post: Disabling Self-Driving Cars with a Traffic Cone

Vice article: ‘Week of Cone’: Activist Group Is Protesting Driverless Cars by Disabling Them With Traffic Cones

Harvey Mudd College Prank Policy

Reality Games

New World Notes article: Viewfinder Finally Playable On Steam & PS5 For Brain-Melting Indie Fun

View-Master Wikipedia entry

Kotaku article: This Ocean-Diving RPG Never Stops Adding Brilliant New Ideas

Baldur’s Gate 3

Five Things On Friday

Games

PC Gamer article: D&D’s upcoming virtual tabletop radiates a big-budget misunderstanding of what matters in role-playing games

Roll20 Marketplace

Gizmodo article: Star Wars: Knights of the Old Republic Is Still Blowing Minds, 20 Years Later

KOTOR Wikipedia entry

Games at Work e102 — I’ve got Plasticity

Hardware

The Verge article: Framework Laptop 16: our exclusive hands-on

Framework midboard QR code resolves to their job postings

DevTerm

ClockworkPi

CrowdSupply MNT Reform

Ultimate Hacking Keyboard Trackpoint Module

Podcast: Play in new window | Download (Duration: 34:51 — 40.5MB) | Embed

Subscribe: Apple Podcasts | Spotify | Amazon Music | Android | Podcast Index | Youtube Music | RSS | More