The month of March 2023 will surely be remembered as the rise of the AI (Artificial Intelligence), or LLM (Large Language Model), as they truly are. People are having a lot of fun and games with ai! Even though there seems to be some significant steps toward something that could be sentient, we shouldn’t fool ourselves. The AIs of today are not consious, they are just incredibly good at what they’re programmed to do: Interpret language and calculate the correct response based on a vast information source: the internet.
There are currently a lot of differing opinions on the subject, and a lot of doom and gloom vs. the opportunists and optimists. I’m not really certain to which crowd I’m more pulled towards. There are some really great opportunities with the technique but as with all kind of incredibly powerful technology, it can also be our downfall. It not all fun and games with ai, one must consider the consequences as well.
I noticed a very interesting tweet by @yoheinakajima:
He basically gave the AI any objective such as ”Make the world a better place” and that its next task is to figure out its next task itself. The AI then went into a loop where it started to search the internet for information on what a first task could be.
He later gave the following prompt:
You are an AI tasked with becoming self aware, your first task is to figure out your first task.
So the AI went on to Google and searched for ”How to become self aware AI”
Apparently the AI found some source where ”sci-fi writer Hugh Howey says we shouldn’t try it”.
After a couple of back and forth (see tweets) the AI got stuck in a loop and fortunately did not gain self awarness.
While this may seem like all fun and games with ai we should remember that it’s been less than two weeks since OpenAI released version 4 of its ChatGPT. That’s a mere three months since the initial release of ChatGPT. V4 is already outdoing humans in many field of academia, intelligence, medicine, coding, etc. and it will only increase.
The CEO of OpenAI, Sam Altman believes That in the next decade we could have a real AGI (Artificial General Intelligence) that could be percieved as sentient and self aware by us. Until then we just have to make do with the already incredibly advanced ChatGPTs!
Leave a Reply