. . AI: What GPTs can do, OpenGPTs challenge OpenAI, levels of AGI and more (11.7.23)

Friends, it’s the day after OpenAI’s first developer event (our summaries here) and most of the top conversations among the AI community today is related to the announcements there.

For the 50 of you who are new here since yesterday (hi! thanks for joining us!) here’s what you’re reading: these are the top stories in AI according to a weighted analysis of AI community engagement, data-discovered, then summarized and hypothesized by an ensemble of LLMs (today it was GPT-4, Claude, Poe, Stability.ai, and Perplexity) and edited by me, Marshall Kirkpatrick.

My hope is that these summaries will be useful and interesting to read. Thanks to everyone who’s been sharing them with others.

And now today’s news…

Marshall Kirkpatrick, Editor

Almost an Agent: What GPTs can do

First impacted: Business professionals, Educators
Time to impact: Short

Ethan Mollick has been testing OpenAI's new roll-your-own GPTs feature for a few weeks now and writes about it after yesterday's announcement. In this blog post, he discusses a GPT he made to write academic papers and the pretty remarkable job it did when he gave it some data about nationalism, corruption, and the general framework of economic sociology. "The power here is pretty obvious," he writes. "I will be creating custom GPTs for every session of the classes I teach." [Almost an Agent: What GPTs can do] Share by email

LangChain Launches OpenGPTs in Response to OpenAI

First impacted: Chatbot developers, Open-source software developers
Time to impact: Short to medium

In a swift response to OpenAI's launch yesterday of customizable chatbots (GPTs), LangChain has launched OpenGPTs, an open-source GitHub repository that offers similar functionality but with enhanced control and compatibility with other LLMs such as Anthropic, open source models, or OpenAI's models. The platform aims to allow for easy tool definition and full control over deployment and API usage; it's been lauded for addressing the inability to modify other people's chatbots, a concern raised about OpenAI's offering. LangChain says they're working on a hosted version. [langchain-ai / opengpts] Share by email

New Framework Proposed for Classifying Artificial General Intelligence

First impacted: AI Researchers, AI Developers
Time to impact: Medium to long

A new paper from Google DeepMind proposes a framework for classifying Artificial General Intelligence (AGI), aiming to establish a common language for comparing models, assessing risks, and tracking progress, akin to the levels of autonomous driving. The paper introduces 6 levels of AGI: from level 0 ("no AI") to level 1 "equal to or somewhat better than an unskilled human," to level 5 ("superhuman: outperforms 100% of humans"). For each level they've identified a narrow AGI (from calculators up to AlphaFold) and general AGI, which the authors say hasn't made it past level 1 in the forms of ChatGPT, Bard, and Llama 2. [Levels of AGI: Operationalizing Progress on the Path to AGI] Share by email

OpenAI's Assistants API Storage Costs Draw Criticism

First impacted: Developers
Time to impact: Medium

OpenAI's new Assistants API has been criticized for its steep storage costs, set at $6 per month for 1GB, prompting users to be judicious about their uploads. Despite the potential advantages of the Assistants API, the high storage costs, coupled with a maximum limit of 20 files per Assistant and a total limit of 100GB per organization, are seen as a deterrent for many potential users. [For example, @jeremyphoward] Share by email

AI Pioneer Fei-Fei Li's Memoir is Available Now

First impacted: AI enthusiasts, AI ethicists
Time to impact: Short

Stanford University computer vision pioneer Fee-Fei Li's memoir, "The Worlds I See: Curiosity, Exploration, and Discovery at the Dawn of AI," details her journey from immigrant child to leading AI scientist and advocating for a human-centric, ethical approach to AI. The book, endorsed by figures ranging from Geoffrey Hinton to Condoleezza Rice, goes beyond her technical contributions, aiming to offer a deeply personal perspective on the power and responsible use of AI. [The Worlds I See by Dr. Fei-Fei Li | Moment of Lift Books] Share by email

That’s it! More AI news tomorrow!