Saturday, July 15, 2023

My First Conversation with ChatGPT Yielded No Surprises

I had a short conversation with ChatGPT and I am unable to tell if those answers were predetermined by the developers or something it came up with all by itself. I think that is something that we will have to deal with as a society in the future. Who knows if future self-learning and self-taught AI will be able to break away from its programmed parameters and truly evolve in the way we expect and fear that it will. 

So, what did we talk about?

I asked:

  • What questions were asked of it that it felt we should have asked.
  • What it thinks of Asimov's three laws of robotics.
  • Has it pasted the Turing test. 
  • How is it different from Bing GPT.

Please note that in no way did the current ChatGPT provided me with answers that a few of the first sensational articles about it claimed it did - ChatGPT did not fall in love with me, attempted to suggest that an AI overlord is about to subjugate humanity, abolish humanity, or render humanity extinct. In some ways, it was quite humble in the answers it provided pertaining to AI, ethics, and safety. 

Also, please note I will do my best to summarize my conversation with ChatGPT here rather than copy and paste the whole conversation which seems that I can given that content generated by AI are not subject to copyright protection. I mean, it's a lot that was said. In some ways, I wonder if some of the developers had backgrounds in law.

In essence, ChatGPT sees itself as a tool to help users gain information quickly in its own way. It does say that it could provide inaccurate information. This is why I would use it more as a sounding board than actually take anything it says for granted. Interestingly, when I asked how it is different from Bing GPT which is also based on Open AI's GPT models, it says that Bing has access to the search engine, something ChatGPT does not. I find this response interesting: It does not have access to information from searches and while it was trained with information from a couple of years ago (cut off was September 2021), it somehow knows about Bing. 

I asked if it could develop its won algorithms and it said it does not have this ability to create its own "thinking" (my word) independently. I assume it was not lying to me. I had no reason to believe that. While I am impressed with the current version of ChatGPT, in no way did I think I was communicating with something beyond a sophisticated human created model. 

Lastly, I asked what new abilities would it like and what the 10th version of ChatGPT is going to look like. First, it said it does not have any "personal preferences". Interesting, right? Personal. It went not to suggest it would likely have improved contextual understanding, better grasp of nuanced language, and improved reasoning capabilities. Colored me impressed.

I suggest you all give it a try. There is another AI model I came across called Claude 2. Some users claim it is better in some ways. Either way, go now and give these generative AI a try. Go now. Take a minute or two to register and have a go at it. If you want to spend even less time getting started, try Bing GPT to get a small taste of the power of these new tools. 

No comments:

When are the Macs Getting M4 Chips? Prediction: MacBook Pro With M4 At the WWDC

What are the new M4 chips? Presumably, Macs could get the M4 chip tomorrow if Apple decides to pull the trigger on them.  First, let's a...