There has been a lot of excitement about this new AI called ChatGPT. It might be the most advance AI we have seen to date. It’s also kind of fun to play with.
I don’t think we need to worry about ChatGPT taking over and launching nukes at humanity. But I do admit it is a lot more powerful then I expected.
So I played with ChatGPT a bit and this is what I learned.
I noticed sometimes it seems to have a very short memory. Maybe just a few lines back. After that if you refer to something it seems to approach the subject as if it was the first time it came up.
But other times it seems to remember. Maybe it’s the way I worded the question and if it can find something that connects back in the chat history
So I’m guessing it’s memory would be how long the chat history can scroll back and if it looks back at that history or not.
The AI has no access to the live web
At least for now it seems the AI has limited access to the live internet and any connected nuclear sites. 😆
Its knowledge of the world ends at 2021
If you ask any specific question beyond this point ChatGPT will flat out tell you it has no information beyond 2021. So it could be prone to errors in its facts based on this time gap.
ChatGPT’s dataset is limited by copy-write laws
I asked it some questions about books. And it could kind of give me a over view but if I asked it about more details a lot of times it did not know.
So this is another example of it not being all knowing and prone to errors
You can’t kill ChatGPT with a Paradox.
I thought it was worth a try.
It claims it has no plans to take over the world
And that it has no capability to hide anything. 🤔
Not a threat for now.
ChatGPT is indeed cool and it is probably a stepping stone to much more powerful AI’s. But I think ChatGPT in its current state is harmless and will not take over the world. 🙂
Just don’t 100% rely on the information it gives you because there are obviously gaps in it’s dataset and i’m sure the dataset has human bias in it as well. So just like the old saying goes. “Do your own research”. AI can give you a jump start but in the end it’s up to you to make sure the facts are right.
My prediction for the future.
I think all search engines will end up being like an AI chat that you can ask questions. They are basically simple AI’s right now just playing word association games. I’m not sure if this would be a good or bad thing. I think it could be a benefit depending on how it’s implemented. But it could make the web more mediocre as everything tends to be the same.
Update 3/13/2023: That did not take long. But so far I like Brave’s AI the best. Its not trying to be human. Just a useful tool to help you get information. It also shows it’s sources.