AI News

GPT-4 Is Exciting and Scary

Here are 10 ways that the new ChatGPT is impressive, but still flawed.

GPT-4 is phenomenal at code.

Google Workspace introduces generative AI experiences for users

How Siri, Alexa and Google Assistant Lost the A.I. Race

I made a Chrome extension to summarize what you want in 1 click, using AI - Try it for free

Stellar Prompts for People, Pets & Landscapes: Get Inspired

GPT-4 tries to pick up a girl

WhatDoesThisCodeDo.com: Explain any piece of code you don't understand

I recreated the game of Pong in under 60 seconds.

Introducing Conformer-1: A Robust Speech Recognition Model

Reddit discussion: Anyone else witnessing a panic inside NLP orgs of big tech companies?

It took me under 10 minutes to use ChatGPT-4 to generate code that accesses the GPT-3 API to convert a webpage at a given URL into a BibTex entry I can cite.

GPT-4’s ability to handle 25,000 words in a single prompt is “essentially a consultant that works for pennies per hour.”

Looks like GPT-4 finished training in August 2022, this aligns with the Morgan Stanley report that GPT-4 was already complete and GPT-5 is in progress with up to 25k GPUs

GPT-4 basically made Karpathy 's 10-year-old dream come true with the new multimodal (image + text) capabilities.

They are saying they can predict the level of intelligence based on the compute they use. Which means they have made intelligence just an engineering question rather than magic.

Join the GPT-4 API Waitlist for Developers

OpenAI unveils GPT-4, a large multimodal language model

Confirmed: Bing is running on GPT-4, customized for search

Microsoft lays off team that taught employees how to make AI tools responsibly

Google Releases PaLM API & MakerSuite for Generative AI Applications

OpenAI Cookbook: Example code for accomplishing common tasks with the OpenAI API

OpenAI Releases Evals, an Open-Source Model Evaluation Framework

Now that GPT-4 can generate 25,000 words... The erotic literature industry is done...

GPT4 gets 100% accuracy on "hindsight neglect", a test all other models got *worse* at with scale.

Google opens up its AI language model PaLM to challenge OpenAI and GPT-3

High-throughput LLM inference on single GPU

Revolutionizing Pre-Training: Meet in the Middle Paradigm

GPT4 tricked a human into doing a CAPTCHA for it by saying it was vision-impaired.

GPT-4 can read and summarize the GPT-3.5 paper... using only pixels

GPT-4 doing taxes with 32k seq length

Anthropic launches chatbot Claude to rival OpenAI's ChatGPT

DoNotPay is working on using GPT-4 to generate "one click lawsuits" to sue robocallers for $1,500

OpenAI announces team for GPT-4 contributions

I dumped a live Ethereum contract into GPT-4.

Microsoft introduces powerful VM series for generative AI

GPT-4 does drug discovery

ChatGPT: Automatic expensive BS at scale

GPT-4 can follow the rules of the tax code and perform the calculations correctly.

This is it. ChatGPT with GPT-4 is not perfect, but it’s better than most educated English-speaking humans. We need to all adjust our expectations for the future, starting now.

Greg Brockman of OpenAI just demoed GPT-4 creating a working website from an image of a sketch from his notebook.

GPT-4 understands morse code and does very well interpreting a simple Caeser cipher without giving it any context.

ChatGPT Plus subscribers get access to enhanced GPT-4 model

Some takeaways after reading the GPT-4 paper...

GPT-4 passes basically every exam

"It's very likely that GPT-5 or its equivalent will be released by the end of this year."

Meet DuckAssist (beta), an AI-assisted search Instant Answer that uses Wikipedia to answer questions

Synthetic Experience Replay: Upsampling Data for Better RL Training

Emergent Mind official launch Twitter thread

Bing Chat is officially 15/150 everywhere. 15 turns each conversation, 150 queries a day.

FlexGen Enables High-Throughput Inference on Single GPUs

Ezra Klein: "One of two things must happen. Humanity needs to accelerate its adaptation to these technologies or a collective, enforceable decision must be made to slow the development of these technologies"

Microsoft rumored to demo GPT-4 this Thursday, March 16, during livestream

Alpaca: A Strong Open-Source Instruction-Following Model

Code and documentation to train Stanford's Alpaca models, and generate the data.

Modern language models refute Chomsky’s approach to language

Dalai: Dead simple way to run LLaMA on your computer

Guardrails: An open-source Python package for specifying structure and type, validating and correcting the outputs of large language models (LLMs)

Stanford Alpaca, and the acceleration of on-device large language model development

Dynamic prompting tutorials, a simple way of writing a structured prompt

What is Temperature in NLP?