The Daily Aus Are we being lied to about how smart AI is?
Aug 13, 2025
A major tech company has claimed their new AI model, GPT-5, operates at PhD-level intelligence. However, experts warn that the AI race is more about marketing hype than genuine advancement. The discussion highlights ethical concerns regarding intelligence measurement and the risks of selective testing in AI development. There's an urgent need for transparency and trust to ensure responsible AI innovation. The hosts also touch on the evolution of AI technology and the importance of a global dialogue on regulation.
AI Snips
Chapters
Transcript
Episode notes
What AI Models Really Are
- AI models are advanced language generators best thought of as supercharged autocomplete systems.
- You interact by typing prompts and the model predicts likely next words to form answers.
Big Claims Around GPT-5
- OpenAI claims GPT-5 scores highly on maths, coding and factual accuracy compared to prior models.
- They position it as a step toward AGI, though that label remains disputed.
Benchmarking Is Fragmented
- Companies run different tests, making cross-model comparisons misleading without standardisation.
- A universal benchmark would let us compare accuracy, speed and reasoning like-for-like.