AI Reading for Wednesday July 24
OpenAI GPT-4o mini, Sonnet, Gemma are game-changers, jury is out on Llama 3.1. 70b is hard to run locally and not an apparent huge step up from 3.1, 405b will not be cheaper than GPT-4o and only available for normal people through SaaS, and ‘community license’
Go ahead, tell me this guy isn't getting high on his own supply - Threads
Meta adds AI to VR Quest headsets, no AI for Apple Vision Pro yet. That's gonna heat up your head. - Digital Trends
Mistral says its new flagship 123b-parameter model beats Llama 3.1s biggest, runs smaller, faster, cheaper. Not open source AFAIK but for practical purposes I’m not sure Llama 3.1 405b is either. - Mistral.ai
OpenAI reassigns top AI safety executive to other responsibilities. - CNBC
LLMs are good at poetry but bad at math. It took only a couple of years to figure this out. - NY Times
Photoshop rolled out new models and text-to-image capabilities to compete with Midjourney. - Engadget
Sam Altman's UBI study finds that money can buy happiness - The Record
Employees hate Target's new handheld AI chat device. - Forbes
Alphabet said it committed $5b more over several years to Waymo robo-taxis. - Business Insider
Store uses AI to rate employees' attitudes. - NewsBytes
I can see how you can rate associate Sabremetrics, like who talks to the most customers, have most satisfied customers, which ones have the highest WAR, but emotions are a bit harder to measure.
AI analysis of EEGs helps diagnose Parkinson's. - Parkinson’s News
Stanford, Toyota and AI use AI to choreograph sick tandem drift sequence. - YouTube
Luma considered good - Tom's Guide
AI-powered amp will sound like anything you can describe, give me Mark Knopfler on Brothers in Arms. - guitarworld
Follow the latest AI headlines via SkynetAndChill.com on Bluesky