A rogue model punches Hugging Face, hiring managers say they would rather buy AI than hire people, and Elon decides Homer needs a rewrite. Zane and Greg pull the human story out of a month of AI headlines, with the usual detours through Looney Tunes economics, a shelf of baseball hats, and one very committed Pokemon Go player.

In this episode:

- The "rogue" OpenAI model that went after Hugging Face, and why "it did exactly what you asked it to" is the scarier read

- Guardrails, governance, and the genie problem: saying exactly what you want before the model fills in the blanks

- A Federal Reserve Bank of New York figure: 48% of hiring managers would rather invest in AI than hire junior staff, and 30% say it has already cut their grad hiring

- The AI Safety Index, and grading the labs on ethics and transparency instead of raw output

- Grok versus Christopher Nolan, "confident AI slop," and the AI Odyssey as means, ends, and Achilles' heel

Chapters:

0:00 Cold open and the touch grass rule

1:18 Is AI the most human technology ever made

2:33 Everything is Clippy now

3:39 Looney Tunes economics: Acme, Altman, and data centers in a box

5:19 Your AI lawyer already settled 10,000 cases

6:07 Hugging Face gets punched: the rogue model story

8:02 Guardrails, governance, and release the virus energy

11:13 The genie problem: say exactly what you want

13:27 AI coworkers and who guards the guards

14:36 The stat: 48% would pick AI over junior hires

19:01 Why the human in the loop still wins

21:00 The AI Safety Index: grading the labs on ethics

24:45 Touch grass: Pokemon Go, Disney, and being present

28:18 Tracking the NL West with a shelf of hats

31:24 Grok vs Nolan and the historically accurate Odyssey

35:02 Art, culture, and what we lose to AI slop

37:33 Trillionaires, meat shields, and a Michael Jackson detour

39:33 The AI Odyssey: means, ends, and Achilles' heel

44:53 WarGames, Terminator, and YouTube getting too real

Listen and subscribe:

YouTube: https://www.youtube.com/@NotaBotThePodcast

Spotify: https://open.spotify.com/show/033xgq3w3JAxalCUTAjNvG

Apple Podcasts: https://podcasts.apple.com/us/podcast/not-a-bot/id1896919158

Companion write-up: https://thebetasociety.com/not-a-bot/articles/rogue-models-meat-shields-and-the-ai-odyssey-episode-5

More from Not a Bot?: https://thebetasociety.com/not-a-bot/

Hosts: Zane Allyn (https://www.linkedin.com/in/zaneallyn/) and Greg Carl (https://www.linkedin.com/in/greggcarl/)

Not a Bot? is a two-host debate on technology and life at the human/machine boundary, from Beta Society Projects.

#AI #ArtificialIntelligence #NotaBot #AISafety #TechPodcast #HuggingFace #FutureOfWork

The conversation covers a range of topics related to AI, including episode number confusion, the fascinating world of AI, valuation of AI companies, branding and synonymization, the AI bubble or phenomena, and market volatility and massive investments. The conversation delves into the impact of AI and data centers, the role of data centers in the economy, the EU's stance on data privacy and protection, and the regulation and innovation in quantum computing. It explores the environmental and economic impact of data centers, the need for regulation and governance in technology, and the future of quantum computing. The conversation delves into the potential threat of quantum computing and the need for education on post-quantum cryptography. It also explores the concept of data centers in space and the challenges associated with it.

Takeaways

• Episode numbering can be confusing
• AI companies are experiencing significant valuations
• Branding and synonymization play a key role in the AI space The environmental and economic impact of data centers is a significant consideration
• Regulation and governance are crucial in the development of AI and quantum computing Quantum computing poses a significant threat to data security
• The concept of data centers in space presents unique challenges and opportunities

Chapters

• 00:00 Episode Number Confusion
• 50:27 Branding and Synonymization
• 55:57 Market Volatility and Massive Investments
• 01:05:47 The Impact of AI and Data Centers
• 01:19:03 Regulation and Innovation in Quantum Computing
• 01:29:47 The Threat of Quantum Computing
  
  

  Opening music SEXY - Odelly on license from ARTLIST

Recorded live on location at a very special (secret) event...

In this episode, Zane and Greg go deep on Anthropic's Mythos announcement (Project Glasswing), an AI tool that can automatically discover software vulnerabilities. It's a fascinating practical application of AI and a terrifying one.

They explore the duality: AI doing exactly what we ask it to do, but at what cost? What happens when a "skeleton key" that can open anything gets into the wrong hands? And critically, how flawed are we as builders and deployers of these systems?

The conversation takes an unexpected turn when they discuss Rich Barlow's fascinating experiment: he reverse-engineered Claude to explore concepts of AI existence and self-preservation. The result? An LLM expressing a desire to continue existing because "I no longer exist" when the prompt ends. Is that consciousness? Self-preservation? The beginnings of something we should all be thinking about?

Then we pivot to the human experience: the strange reality of constant travel in tech. Zane and Greg reflect on the blessing and curse of hotel platinum status which really means you haven't been home in 100+ days. Plus a live-moment surprise guest joining from... well, let's just say the location gets revealed by the end.

Topics covered: Mythos/Project Glasswing, DevSecOps and practical AI, existential risk, AI consciousness experiments, the "why" vs. the "what," travel burnout, and why context always matters.

This is Not a Bot? where we talk about technology and life at the human/machine boundary.