So far, I have used the AutoGPT experiment and https://chat.openai.com/ to work on some interesting projects, and since I have nowhere to talk about it, I figured this would be the best place to dump my findings:
Building HTML & JS:
So I build Drupal sites on the side and have used ChatGPT to help create all sorts of extraordinary works, although it does need a lot of massaging for CSS. All in all, asking for design suggestions has given many directions for framework development and strengthening the responses when performing Lighthouse inspections.
Content Development:
Many companies have been proposing that ChatGPT is revolutionary regarding customer engagement. I will have to disagree with this sentiment openly. Although this software is impressive at navigating and conceptualizing a core ideal, once a topic begins to drift three or four layers down, the software becomes unnecessarily encompassed with wanting then to revolve all the topic material into a brief sub-layer of contextualization.
A simple example would be wanting to talk about balloons and strings. When your topic material begins to talk about air, it can have the potential to want to focus on only the air within the balloon and ignore the intent of the original question.
Regarding maintaining a narrative with Auto-GPT, the software still requires a lot of guidance between its determinations on how to carry a story or how content should be represented. It can sometimes get caught in what I would call "writer's block" when choosing between two opposing sides with equal merit.
Dungeons & Dragons:
So here is where the AI has me blown away. My friends and I have constructed a 5th Edition Homebrew campaign letting the AI do everything from generating characters to NPCs and correct CR encounters. I have to say, for whatever reason, by using AI narratively in this fashion, the software can stay fairly cohesive in presenting the campaign. My friends and I think this is just a group delusion because, inherently, that is what Dungeons and Dragons are under the surface of the game place since we all define the values interim to the game itself.
I think this development direction will revolutionize people's existence, but I cannot say to what extent. The internet was built with the idea that if we could all share information, we could make the world a better place. However, many social media outlets could care less about that as a mission focus. I suppose it's just a matter of time till we see the extent of what AI will do to offer this change.
I remote interviewed someone for a dev job and it became obvious they were feeding questions to chatgpt. As soon as the questions were nuanced and required experience and practical knowledge, they basically bailed on the interview.
I asked them why they kept pausing and why the pauses were getting longer. Rather than say "I'm thinking" or something, they simply said they were not pausing. Um, ok.
I've been working on an autonomous agent. Basically you can chat with it as if it was an assistant, and it will do tasks for you or answer questions, but it also just figures out what tasks need to be done and does them. It keeps track of all sorts of info and has access to a wide range of actions. For example, I just did a demo where I had it gather everyone's pizza order via slack and text message and then look up the number for and call the pizza place to make the order. It still just amazes me to watch the server logs as it figures out how to perform some task it wasn't specifically programmed to do and has never done before.
I'd like to do something like this. How do you take in the orders from slack and text message and pass it to ChatGPT and then have it look up phone numbers. I thought it wasn't connected to the internet
Base ChatGPT isn't connected to the internet. They have (or had because it's temporarily disabled) a version that could browse the web (poorly). My company's software uses the OpenAI API along with a lot of code we wrote to do the rest.
I asked it to play the role of a 1920s gangsters just as a morning joke. I was thinking it would be fun to have a morning conversations with this AI patsy.
Long story short, it planned to rob Charles Lindbergh within 3 questions.