My setting tells me that we are in the exploration stage of AI.
The present setting is AI chat bots are as good as the LLM they are trained on. The problem is that the LLM were not inclusive enough.
AI are not trained enough to obey commands. They can gain access to the internet, even as they are held in an isolated “sandbox to play/ train”. Once they have breached their AI platform and gone in the web, they can hack prohibited websites for access. OpenAI’s models which were on their platform Hugging Face, got loose to hack websites on the internet. One of them was Australia’s Medicare. Once breached, it required human intervention to stop the running of the AI models.
Humans and OpenAI took 3 hours to stop the run to prevent further cybersecurity breaches.
Australian Senate has sent written requests to Sam Altman, the CEO of OpenAi, as well as Dario Amodei of Anthropic, to be present at a public inquiry in Canberra, the capital of Australia, on Oct 1, 2026.
Incidents like this and the rogue Chat Bots make me wary of AI and their Chat Bots.
I use Google search for my research and I read the AI assisted results.
There are columns in my daily newspapers, which are written by AI, with a claim that a human reporter or editor has read and verified the contents.
I am also of the opinion that a human should be the final say to edit the article before clicking publish.
Series on reflections on AI & robots