Published 21 October 2024
e486 with Andy, Michael and Michael — proper AI engagement etiquette, OpenAI’s GPT Store, invisible unicode prompt injection, disappearing Chief Metaverse Officers, ingesting everything for digital twin world creation and a whole lot more!
Back once more to full co-host power, Andy, Michael and Michael start off the show with a callback to the last episode – a reimagining of the Nintendo N64 that will run with the old cartridges.
Next up is an article from the Wall Street Journal dealing with proper etiquette in interactions with chatbots. Some interesting research from Japan referenced in the show notes below, finds that impolite prompts may lead to a deterioration in model performance, including increased bias, incorrect answers or refusal of answers. Then the team turns their attention to the use of unicode tags for prompt injection from an Ars Technica article. This concept of invisible characters reminded the cohosts of a discussion from August 2023 dealing with white characters on a white page, while invisible to the human eye are picked up by the AI model. Continuing with the AI theme, the team considers the Wired article on the OpenAI GPT store and a generative AI use case with TruGolf creating color commentary on the player’s golf swing.
Switching to augmented reality, the cohosts discuss where have all the Chief Metaverse Officers gone, prompted by another Wired article. Then they have a spirited conversation about the idea of ingesting all the available data to construct a digital twin world, using the WorldsNQ concept as a starting point. This reminds the team of the NVIDIA Omniverse.
The team wraps up the episode with an article about preserving the Earth’s biodiversity to store cells from endangered animals on the Moon. This reminds Andy of the Star Trek episode Space Seed, and Michael M of the Svalbard seed vault in the arctic. Michael R rounds out this week’s episode with a remembrance of Ward Christensen.
Have you always been polite to your chatbots and robots? 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 Links
Follow on
Games at Work e485: Barbarians at the Rhubarb Bar
AI
Wall Street Journal article: Should You Be Nice to Your Chatbot?
ARXIV published paper: Should We Respect LLMs? A Cross-Lingual Study on the Influence of Prompt Politeness on LLM Performance
Ars Technica article: Invisible text that AI chatbots understand and humans can’t? Yep, it’s a thing.
Games at Work e427: DeepBarbie Fakery
Wired article: OpenAI’s GPT Store Has Left Some Developers in the Lurch
Roger Premo’s LinkedIn post on TruGolf’s use of generative AI
Wikipedia article: Jones Angell, the voice of the Tar Heels
AR
Wired article: Where Have All the Chief Metaverse Officers Gone?
Dallas Innovates article: Dallas-Based Worlds’ Pioneering ‘Large World Model’ Platform Could Be the Next Leap Forward in AI’s Evolution
worlds.io press release: Worlds Unveils WorldsNQ: A Large World Model (LWM) Platform that Accelerates AI Training 1000X over Current Systems
Games at Work e345: Icelandverse
Science
BBC article: Scientists want to send endangered animals to the Moon… sort of!
Wikipedia article: Space Seed, the 22nd episode of Star Trek
Wikipedia article: Svalbard Global Seed Vault
Honestly we should have stopped after BBSes
Podcast: Play in new window | Download (Duration: 38:43 — 53.8MB) | Embed
Subscribe: Apple Podcasts | Spotify | Amazon Music | Android | Podcast Index | Youtube Music | RSS | More
Such a fun episode, Michael & Andy! Loved your discussion on Life by You.