ChatGPT: First Impressions

GPT-3 Tweaks are trending.

Hey Everyone,

It’s been a really interesting week for the Zeitgeist of A.I. adoption. Even as we await announcements on GPT-4 and more Microsoft funding for OpenAI. Now we have Davinci-003 and ChatGPT.

Alex Kantrowitz

recently covered this trend here. Twitter also seems enthusiastic about it.

Okay maybe not that enthusiastic:

Fascinating @fasc1nate

Throwback to this Gorilla in a pool dancing to Maniac.

ChatGPT is OpenAI’s newest language model based on the GPT-3.5 series of LLMs, optimised for dialogue.

During the research preview, usage of ChatGPT is free. Try it now at chat.openai.com.

Michelle Huang

michelle huang @michellehuang42

i trained an ai chatbot on my childhood journal entries - so that i could engage in real-time dialogue with my "inner child" some reflections below:

‘It felt like I was reaching into the past and giving her a giant hug, and I felt it ripple back into the present’

I’m not sure how genuine the story but it is intriguing.

While we’ve been waiting years for “chatbots” and Conversational A.I., we may be getting a bit closer with GPT-3 tweaks.

Guy Parsons @GuyP

OK so @OpenAI's new #ChatGPT can basically just generate #AIart prompts. I asked a one-line question, and typed the answers verbatim straight into MidJourney and boom. Times are getting weird...🤯

Davinci-003 and ChatGPT do show GPT-3 continues to be tweaked in meaningful ways even years after its release as a foundational model.

Someone on LessWrong write about their first impressions of ChatGPT.

Travis Smith @_tks98

This is actually so wild, I really didn't think we'd get to this point so quickly with language models where they can analyze source code for bugs and write suggested fixes? #chatGPT openai.com/blog/chatgpt/

Combined with details about Google’s project Pitchfork this week (Insider), it’s a lot to digest.

Even OpenAI seems to be impressed with the demand for demos of its chat bot spin-off.

Sam Altman @sama

there is a lot more demand for ChatGPT than we expected; we are working to add more capacity

ChatGPT you see is trained using Reinforcement Learning from Human Feedback; human AI trainers provide supervised fine-tuning by playing both sides of the conversation. However it can still get into some moral dilemmas of course.

It’s been 10 years now since HER, the movie about the human-AI interaction that changed how a lot of us hope A.I. might become in terms of dialogue. It’s stunning to see the decline Amazon Alexa, Siri, Cortana among others during this period.

ChatGPT is a new spin on what a “chatbot” can do. People try to push them to the limits for a good story too:

Alice und Bob @alice_und_bob

AIs are all like: "na babe I'm not that kind of AI" but once you start role-playing they go full Armageddon on you. #ChatGPT

People training a chatbot on their old diaries to have a conversation with their younger self is not something I would have considered amusing. I think what it shows is we crave an A.I. that’s not just good at dialogue, but relatable like a counselor or personal assistant.

michelle huang @michellehuang42

i kept diaries for about 10+ years of my life, writing almost everyday — about my dreams, fears, secrets the content ranged from complaining about homework, to giddiness i felt from talking to my crush some days were very mundane, some rather insightful