Enough About AI

Bursting the Bubble?


Listen Later

Dónal and Ciarán explore the increasingly looming questions about the overinflation of the major AI companies' valuations and the anxiety about whether we are in a bubble - and what might happen if it pops. Relevant to their discussion of the need for AI models to keep the hype levels high, they discuss the muted reception to the release of ChatGPT-5 and some of the emerging strategies to make AI chatbots more palatable to certain audiences who are worried about "woke".

Topics in this episode

  • Are we in an AI bubble? Spoiler alert: Yes - based on any normal metric of what an investment bubble is - but why is the promise of an almost-there superintelligence keeping things from popping?
  • Despite being in a bubble, the lasting impacts of where AI already is on jobs and society is discussed
  • The recent release of ChatGPT-5 has led to negative feedback from the tech press and vocal users - this is contrasted with other recent version releases.
  • Examining how AI companies are trying to find new ways to add value, leading to a discussion of "Third Devices" and AI hardware
  • The limitations and diminishing returns of training on synthetic data - and the apparent slowing down in model progress
  • AI & Ideology - what does it mean to have a non-woke AI?

Resources & Links

  • The Economist story mentioned by Dónal: "AI valuations are verging on the unhinged - Unless superintelligence is just around the corner" (25 June 2025)
  • Article in TheJournal.ie on "Brendan", the AI Dublin Tour Guide
  • ChatGPT's dodgy graph is linked and discussed here: "OpenAI gets caught vibe graphing" (The Verge, 07 August)
  • Sam Altman (OpenAI) tells venture capitalists that he will take billions of their money and build AGI - and then ask it how to make a return on the investment  (Twitter Video, Warren Terra)
  • Some good discussion on the struggles of agentive AI ("AI Agents have, so far, mostly been a dud", Gary Marcus, Substack)
  • Apple's important recent paper on the limitations of "reasoning" within tested reasoning models is available as a PDF here
  • Coverage of Truth Social's deal with Perplexity - to make a non-woke chatbot for the platform

You can get in touch with us - [email protected] - where we'd love to hear your questions, comments or suggestions!

...more
View all episodesView all episodes
Download on the App Store

Enough About AIBy Enough About AI