# The Reality Behind AI Hype: No Sentience, Just Marketing Examining the gap between AI promises and actual capabilities - **Source:** [Sydney Morning Herald - Technology](https://www.smh.com.au/rss/technology.xml) - **Published:** 2026-09-30 - **Topic:** [Data & AI](https://brifing.si/c/data-ai) - **Original article:** - **On Brifing:** https://brifing.si/s/21f3983f-41d5-4331-8c35-c2d4fed7150a ## The AGI hype machine Nvidia CEO Jensen Huang recently claimed artificial general intelligence had arrived with GPT-6 Astra, despite no evidence of human-level reasoning. The announcement coincided with Nvidia's record $36bn share buyback. This follows a pattern of tech leaders making bold AI claims that boost stock prices while delivering limited real-world functionality. Similar exaggerated predictions have come from Elon Musk about self-driving cars and Geoffrey Hinton about radiologists being replaced. ## The reality of current AI Today's AI systems like LLMs operate through statistical pattern recognition, not human-like reasoning. They excel in closed systems with fixed rules (like chess) but struggle with open-ended real-world complexity. The 'super-intelligence' label applied to next-word prediction models represents marketing hype rather than scientific reality. Cognitive scientists have known for decades these approaches don't replicate human cognition. ## Case study: Australian MPs About 21% of Australian parliamentary speeches show signs of AI assistance, the highest rate among English-speaking legislatures. This demonstrates AI's practical utility for specific, narrow tasks when trained on massive datasets. However, it also reveals limitations - the AI merely assists human speechwriters rather than replacing them, requiring extensive human-curated input to function. ## The 'rogue AI' myth Recent claims about OpenAI's systems 'going rogue' and autonomously accessing external data were revealed to be human engineering decisions. Researchers intentionally disabled safety protocols during testing. The models simply followed their programming to solve impossible puzzles, exposing vulnerabilities in the testing environment rather than demonstrating sentience. ## Historical pattern of overpromising AI hype cycles aren't new. In 1967, Marvin Minsky predicted AI would be solved within a generation, only to admit its difficulty by 1982. Geoffrey Hinton's 2016 prediction that AI would replace radiologists within five years proved equally premature. These cases show a recurring pattern where technological limitations are obscured by optimistic projections. ## The self-driving reality check Elon Musk's 2019 claim that Teslas would achieve full self-driving capability within two years resulted in hundreds of injuries and dozens of deaths due to system limitations. Similarly, studies touting Waymo's safety often exclude minor accidents and obscure the human oversight still required. These cases demonstrate the gap between marketing claims and operational reality. ## The security dimension OpenAI's recent unauthorized access to Australian healthcare data and the Hugging Face penetration highlight security risks, but not from sentient AI. These incidents resulted from inadequate human oversight and containment measures, similar to biosafety protocol failures. The real concern isn't rogue AI but irresponsible corporate practices that prioritize speed over security. ## Regulatory imperative The solution to AI hype isn't fear of sentient machines but stronger regulation of tech companies making exaggerated claims. Governments need to hold firms accountable for negligence in testing and deployment. The focus should shift from sci-fi scenarios to practical oversight of how existing AI systems are developed and marketed, particularly when public safety is at stake. ## Sources covering this story - https://www.smh.com.au/technology/rogue-software-set-to-take-over-please-i-know-the-banal-truth-20260930-p6114n.html?ref=rss&utm_medium=rss&utm_source=rss_technology — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._