Exploring AI Applications and Networking

Deep Hack Demo Day Invite

  • Invitation to a demo day
  • Sales pitch to share with friends in Bengaluru
  • One member wishes they could attend but will be flying out that day

Jobot, a ChatGPT-powered bot is live at Jovian

  • Announcement of a new ChatGPT-powered bot
  • One member asks if the demos will be streamed

Introductions

  • One member introduces themselves as an anthropologist and senior fellow with the Max Planck Institute for Social Anthropology
  • Another member introduces themselves as a new media artist and educator, with a focus on using AI, robotics, and code to build experiences
  • A third member introduces themselves as a search engineer at Wolfram|Alpha

AI Projects and Demos

  • Discussion of a video project that uses OCR, content search, style transfer, and projection mapping
  • Sharing of a GitHub repository for a blind visual aid using GPT
  • Discussion of a demo for a work-in-progress project that integrates ControlNet to alter per specific requirements for a particular color-blind person
  • Sharing of a news article about a project that uses goggles for the blind
  • Discussion of using serverless GPUs and accelerate to offload to CPU
  • Discussion of using GPT to write and create episodes for the final season of Game of Thrones
  • Discussion of GPT’s ability to handle degeneracies and reason about problems in its distribution but not in its data
  • Discussion of GPT’s performance on new Leetcode problems past its training date
  • Sharing of a blog post about cheating in coding using LLM-powered assistants
  • Discussion of the potential for LLM-powered coding assistants to automate much of the code we write and the skills that will become more important as a result

Miscellaneous

  • Discussion of a text-to-video tool for camera movements over still images
  • Sharing of a ControlNet web app and API
  • Discussion of the eligibility of remote submissions for a prize

The description and link can be mismatched because of extraction errors.

  • URL: Not provided. The message is related to the introduction of the sender, who is a new media artist and educator focused on making cutting edge technology accessible. The sender also runs a website focused on addressing societal challenges of technology in the Indian context. The question “is it done via stop motion?” is not related to the given context.
  • https://github.com/justanotherlad/blindvisaidgpt: The author mentions trying out generative video models and shares a link to their GitHub repository. They also introduce themselves as a search engineer at Wolfram|Alpha.
  • https://m.economictimes.com/tech/ites/tech-mahindra-tests-goggles-for-blind-plans-variant-for-cars/articleshow/45302948.cms - The link is related to a discussion about a project done for TechMahindra that involved creating goggles for the blind and a possible variant for cars. The message also suggests that the project was not open source and should not be shared outside, but mentions that similar projects could be done today using phones with decent computing power.
  • If anyone is interested in making ControlNet work on iPhone, this is a good start: https://github.com/apple/ml-stable-diffusion/issues/131
  • https://aisnakeoil.substack.com/p/gpt-4-and-professional-benchmarks - Context: Aravind’s post about contamination and GPT-4’s performance.
  • https://about.sourcegraph.com/blog/cheating-is-all-you-need: The author discounts Github’s excellent search product which has similar or maybe larger advantage than Sourcegraph, while someone on Twitter says gpt-4 does much worse than leetcode hard on newer Euler problems.
  • https://niranting.substack.com/ (The author’s Substack, discussing the potential of LLM-powered coding assistants and referencing a Google rant from 2011)
  • https://www.buildt.ai - a website that potentially uses LLM powered coding assistants and has search capabilities on codebases.
  • https://github.com/features/preview/copilot-x: Github Copilot X (available for preview only) is mentioned as a tool that is going to change how we work, and the author is working on something using LLMs.
← AI and Deep Learning in eCommerce and Beyond
OpenAI's Development Trajectory and the AI Dilemma →