Here’s my short-n-sweet take on where we stand today.

“Consciousness”, “Singularity”, “Artificial General Intelligence” (AGI), “sentience” — whatever you want to call it. Is AI there yet? Can it become conscious?

Current AI tools like ChatGPT, Claude, Gemini, etc. have no understanding of what is real. They just scrape documents from the web, do pattern recognition, and regurgitate results that pattern-match user prompts. And with the massive “Large Language Models” (LLMs) at their disposal, they do a very good job.

But they do not have a “world model”. They have no functional understanding. No abstract thought. No symbolic reasoning. If there’s inference there, it’s data-driven, not imaginative.

So no, today’s AIs are not conscious.

They imitate a sense of self, based on what they’ve scraped off the web. But it’s a stretch to call that sentience!

Will AI become conscious (aka sentient) at some point? Maybe in decades to come.

Decades ago I dabbled in AI with primitive languages like Lisp and Prolog. I gather that Python is much more up to the job. When today’s LLM AIs answer questions that they shouldn’t, or give answers that they shouldn’t, the AI companies use Python (and other programming tools) to build guardrails. This is called a “harness”, like the harness that is used to control a horse.

Theoretically, a harness could be built in such a way as to place symbolic reasoning, even a world model, on top of an LLM AI. If such a harness on a massive LLM becomes sentient … well, that will be an interesting day.