Hi everyone, I hope everyone is doing well at the hackathon! Unfortunately, I had to leave yesterday to focus on my graduate school studies this weekend as I approach the end of the semester. Out of curiosity, I was thinking of creating a ChatGPT plugin that would query relevant help from StackOverflow and GitHub forums when a programmer has compilation errors. Does this idea have any merit you think? I know GitHub Copilot X exists, but I am currently on the waitlist. Thank you!

anyone know how to make chatgpt wait longer for the response from a plugin? I’m using langchain summarization behind the scene and it takes about 5 mins to process, but chatgpt seems to expect the response in like 2 mins or so

Hey all, So, I'm working on running a Service Safari (basically a human-centered design study that takes someone through a service/ product experience and get their quick reactions on it as a way to better understand user needs and to inspire creative concepts). I'm looking for a service that: 1. Involves (Gen) AI- assisted tech in some way 2. Only accessible in the Bay Area / North America (this might be the kicker) 3. Ideally having to do with Wellness/ Health/ Learning/ Work 4. Easy to record participants' reactions without too much regulatory overhead (e.g. clinical experience is hard because of HIPPA etc.) 5. Something that is adjacent to consumer electronics, but not limited to it For example, self-driving car is a good one, but I need other examples. Any ideas? (Prototypes are good, too).

Hey all, does anyone have experience working with DialogFlow (https://cloud.google.com/dialogflow) or Azure Health bot (https://azure.microsoft.com/en-us/products/bot-services/health-bot/#overview) ?

Hi everyone, is anyone working in the AI interpretability/explainable AI space? I am interested in learning more about the problem of AI interpretability and how current AI regulations rely on this concept that is in its infancy, and how regulations will affect AI development and deployment. Two questions come to mind; first, could interpretability ever be achievable for an AI system? Second, assuming we can accomplish the first, can interpretability be advanced enough to solve alignment issues and provide safe regulations to control AI from acting in unpredictable and potentially malicious ways? Any relevant research in this area would be much appreciated. I am happy to share my thoughts and ideas. Thank you!

@Conner Swann: “This is probably the most production-ready agent implementation” https://github.com/microsoft/JARVIS • I’m spending time this week studying / playing with agent libraries, if anyone wants to jam in Hayes valley DM me!

I say this with extreme caveat that it is not production ready in any way shape or form lol — who is going to make the production agent library? 👀

I'm seeing a lot of buzz around pinecone, in part because it's used in BabyAGI. When I look at BabyAGI, though, it looks like pinecone is really only being used as little more than a simple key-value store. I get the theoretical advances of a vector store for use with LLM's, but are there any actual practical advantages of vector stores like pinecone for a BabyAGI-like use case compared to a more conventional MongoDB or Redis cache kind of store?

Anyone try deploying an app that uses langchain to heroku? I am realizing maybe its super bloated with dependencies to actually make a lightweight web app.

I made some rough estimates this morning on the cost per hour of providing an ai chat representative. The numbers are biased towards customer support or sales support scenarios where the human is having a focused, back-and-forth text chat conversation with an ai "representative" but are probably decent hand-waving estimates for a decent range of chat scenarios. Cost per hour of chat: GPT-3.5-turbo: $1 GPT-4: $7 - $430 Call center employee in the Philippines: $2 Call center employee in the US: $6 Lawyer in the US: $300 - $600 details in 🧵 for math types (edit added): My read on these number is GPT-3.5 is already cheap enough to open up tons of business models that didn't previously make sense with humans, but that GPT-4 is currently about 10x too expensive today to do the same (it's too close to the cost of appropriately educated human labor when run in a steady-state continuous mode - yes, special cases will exist, but it doesn't change the business model of "everything"). That said, the gpt-3.5-turbo release dropped the cost per token for the gpt-3.5 family by 10x. If openAI can pull off the same 10x cost reduction for the gpt-4 family, then pretty much everything opens up as a business model. Personally, I'm guessing there's at least a 3x cost reduction to be had for the GPT-4 family models in the not too distant future, and I wouldn't bet against a 10x or 30x or 100x reduction in the next few years.

Hi guys. A while back I was working on an AI project of my own and quickly realized that user authentication (login, registration, account management, payments and so on..) would take a pretty ridiculous amount of time. I'm looking into this a bit deeper now and would appreciate it if you guys could spare 7 minutes of your day to answer this form: https://698rzrbjveb.typeform.com/to/U7wMzqys

Is it possible to add more team members for the ChatGPT plug-in access? We are partnering with another organisation to build out the plug-in!

Hello everyone! I have a question about the architecture of our AI system. We use a python server in production to handle our core AI logic (together with Azure ML), because of all the data science oriented libraries. We use a sentence transformer package, which has pytorch as a dependency (also nvidia). After checking, these two packages have a combined size of something above 3 Gb, which is a lot - it’s very slow to upload on our container registry + it’s also slow to launch when scaling up. Because we run in kubernetes, we want to be able to scale and downscale fairly fast. I was thinking about extracting the functionality using the huge libraries to a separate microservice so it would scale only if it really had to, but this could potentially create bottlenecks + it would also increase the complexity of the entire system. Maybe it’s better to just tank the huge size when scaling the pods?. Any advice on how to deal with issues like these or what is the best AI System architecture for these kinds of deployments ?

Could anyone help me with a vercel deployment problem? Essentially only my logs from react front end api calls are showing up in my runtime logs. My python logs don't show up. I am using

import logging

and ensured basicConfig was set to logging INFO. That is a subproblem to a larger one - I am getting a lot of function execution timeouts with calls to ChatGPT, but not getting them locally. The logs are to debug why. It's strange because I upgraded to the pro plan which allows for 60s serverless execution function time but it errors out fairly fast

Sharing some notes on how to implement a ReAct agent from scratch with 150 lines of code. Happy to chat about it or answer any questions here too. https://twitter.com/HeyTedC/status/1651805471555346433

Product or project idea? Is anyone making API integrations easier, like what OpenAI is doing with their docs -> plugins tech? Not "are you a making a plug-in", more "are you making it easy for other devs to add integrations to their own projects?" I need to integrate some old and horrible but well documented API's (that I don't control) into a system I'm building. I don't want to do this as a ChatGPT plugin because I don't every API call to incur token costs and multi-second waits. I want to feed the API docs into something once and get back a purpose-built SDK in the language of my choosing that does exactly what I need it to do. Anyone built that or building it? Am I dreaming in thinking someone could build this?

Is there a Chat-GPT plugin, langchain module, huggingface model or whatever else to submit PRs to existing codebases? LLMs are good at suggesting code changes but if you want to revise with them in rapid succession it would be best that they make their suggested changes as PRs to your repo than just pasting code snippets in the chat

More thoughts on LLM pricing: gpt-3.5-turbo and gpt-4 launched around the same time. gpt-3.5-turbo offered a huge price reduction from the previous model and gpt-4 represented a huge quality increase from the previous model. Can we make guesses about whether gpt-4 already has the gpt-3.5-turbo style optimizations, and if it doesn't have it can we make estimates of what kind of cost reductions might be coming in a hypothetical future gpt-4-turbo? tl;dr - it looks like gpt-4 already has the "turbo" style optimizations baked in. There will certainly be more optimizations to come but it doesn't look like there is an "easy" 10x turbo optimization waiting ready to dropped onto gpt-4. Math in 🧵

I slightly broke GPT-4! Check out these formatting oddities. I think it's a cool insight to how it thinks about code and instructions vs regular actions. https://drive.google.com/file/d/1ysE4yQv4fDN9SH8rT2X9kFxeAj0AlZsY/view?usp=sharing

I read this NYT article and wanted to share it with you all. Geoffrey Hinton worked in machine learning and AI at my university (CMU), so a friend shared it with me. Please let me know what your thoughts are. It is definitely an interesting time to be in AI. https://www.nytimes.com/2023/05/01/technology/ai-google-chatbot-engineer-quits-hinton.html

Has anyone built a chatbot for Microsoft Teams and/or listed it on their marketplace? I have a quick question.

Anybody have any rabbit holes/libraries/case studies they recommend for utilizing group chat to interact with their Chatbot? Use case example for more context. We connected our chatbot to a twilio we'd love to be able for users to add our assistant to the group chat.

Anyone have success generating more photorealistic images that don't have the slightly warped look without training their own models? If so, how are you generating the prompts, and what kind of service are you using for that? We tried DALL-E, but we want to focus on more realistic rather than artistic images Our use case is generating images to add to custom work instructions content for the manufacturing industry

Have another question 🙂 Have anyone been able to context switch in a chatbot convo that uses embeddings for info retrieval from a set of documents? Essentially this question: https://community.openai.com/t/dealing-with-context-switching-in-a-conversation-that-uses-embeddings-for-information-retrieval/87836/2

i came across a paper or library sometime ago that did generative memes - text/images. but can’t find it anymore! anyone else remember what it was?*

What are the workarounds for OpenAI completion API of 4k tokens? I am asking LLM to summarize the content in a particular format (aka JSON schema), which takes 2k tokens in the prompt without the actual text. I found that JSON is not very token-efficient and already switched that to the YAML as input and output, but I still need to keep the structured format. Sample of the schema:

attr_1:
  type: number
  minimum: 0
  maximum: 30
attr_2:
  type: string
  enum: [value_1, value_2, etc]

and 50 more attributes like that. How are you optimizing long-structured schema prompts? What are the potential workarounds? Would love to hear from the community 🤖

For everyone looking for access to ChatGPT-Developer-Plugins https://github.com/SamurAIGPT/ChatGPT-Developer-Plugins

Anyone have a recommendation for load testing Slack chat bots?