Google co-founder Sergey Brin has stepped back into the trenches, working three to four days a week at the Mountain View office on the upcoming Gemini model. Google is building the multimodal AI to counter OpenAI's ChatGPT4, aiming for a rollout by the end of the year. CEO Sundar Pichai is reportedly enthusiastic about his hands-on return.
OpenAI and several leading AI labs just agreed to a set of voluntary safety commitments coordinated by the White House. The pact requires companies to run independent red-team evaluations, watermark synthetic audio and video, and lock down unreleased model weights. It marks an industry effort to establish shared guardrails before formal regulation arrives.
OpenAI just rolled out custom instructions for ChatGPT, putting an end to pasting the same context into every fresh chat. You can now define your role, preferences, or tone once, and the model applies them to all future conversations. It is available in beta for Plus users today, with a wider rollout coming soon.
Apple is quietly testing a ChatGPT-like internal chatbot that employees call "Apple GPT." Powered by its home-grown "Ajax" large language model framework, the tool is currently limited to internal prototyping and summarization, though Bloomberg reports the company is eyeing a major AI announcement next year.
Thousands of writers just signed an open letter calling out OpenAI, Meta, and other AI developers for training language models on millions of copyrighted works without permission or compensation. The letter demands companies obtain licensing and fairly pay authors for both past and ongoing use, warning that unlicensed scraping threatens creators' livelihoods.
Microsoft just announced that Microsoft 365 Copilot will cost $30 per user, per month for commercial customers. Alongside the pricing, the company launched Bing Chat Enterprise in preview with commercial data protection that guarantees chat data won't leak or train AI models, and rolled out visual search to Bing Chat. Why it matters: At $30 a month on top of existing plans, Microsoft is betting that generative AI can command premium enterprise software prices.
OpenAI just partnered with the American Journalism Project, committing $5M in direct funding and up to $5M in API credits to help local newsrooms deploy AI. The initiative includes a dedicated technology studio and pilot grants for roughly ten nonprofit news organizations. Why it matters: As the news industry grapples with generative AI risks, this collaboration attempts to give grassroots journalists a direct hand in shaping the tech.
Researchers from Stanford and UC Berkeley evaluated March and June versions of GPT-4 and GPT-3.5, finding massive performance drift across diverse tasks. While GPT-4's prime-number accuracy plummeted and code formatting errors rose, GPT-3.5 actually improved on some tests. The findings highlight the unpredictable risks of relying on opaque, continuously updated LLM APIs.
Wix just revealed an upcoming AI website builder that creates complete, fully functional sites from a single prompt. Instead of using static templates, it pairs ChatGPT for copy with proprietary design algorithms to generate custom layouts, with a companion business AI assistant also in the works.
OpenAI just inked a major partnership with the Associated Press to license the news organization's extensive text archives for AI model training. In return, the AP gets access to OpenAI's tech and expertise, though it made clear it won't be using generative AI to write actual news. The move gives OpenAI a legal, high-quality stream of data amid mounting copyright scrutiny.
Llion Jones, the last remaining author of the groundbreaking 2017 Transformer paper 'Attention is All You Need', just confirmed he is leaving Google to start his own company. All eight researchers behind the architecture powering ChatGPT and Midjourney have departed, launching startups like Cohere and Character.ai now valued at over $4.1 billion combined.
Elon Musk just launched a new artificial intelligence company called xAI, setting an ambitious mission to "understand the true nature of the universe." The founding roster brings together researchers from DeepMind, OpenAI, and Google, with plans to lean on Twitter data and Tesla engineering to build a ChatGPT rival. Why it matters: Years after walking away from OpenAI, Musk is officially back in the frontier AI race with his own ecosystem to back him up.
OpenAI just signed a new six-year agreement with Shutterstock, securing training access to its massive library of images, videos, music, and metadata. In return, Shutterstock gets priority access to OpenAI's latest models, bringing advanced synthetic editing tools to its platform and GIPHY. High-quality, legally cleared training data is becoming the industry's ultimate moat.
Google DeepMind CEO Demis Hassabis just confirmed that the viral leaked memo claiming tech giants have "no moat" against open-source AI is genuine. But he pushed back on the gloomy forecast, arguing that the newly united Google Brain and DeepMind teams are poised to lead the next decade of breakthroughs.
Stanford researchers found that seven popular AI detectors routinely misflag essays written by non-native English speakers as machine-generated. In tests, over half of real TOEFL essays were marked as AI-written—with one tool wrongly flagging 98%—simply because non-native writers tend to use simpler vocabulary that yields low perplexity scores. Ironically, when researchers had ChatGPT rewrite the essays with more complex language, the detectors labeled them all as human-made.
Comedian Sarah Silverman and two authors just slapped OpenAI and Meta with copyright infringement lawsuits in US District Court. The complaint alleges ChatGPT and LLaMA were trained on pirated books grabbed from "shadow libraries" like Bibliotik without consent, pointing to ChatGPT's book summaries as proof. The battle over AI training datasets is officially heating up.
Google is now testing its specialized Med-PaLM 2 chatbot at research hospitals like the Mayo Clinic, racing Microsoft to dominate healthcare AI. Trained on medical licensing exams, the model can answer clinical questions and summarize records, though Google researchers admit it still hallucinates and isn't ready for real-world clinical decisions.
OpenAI’s ChatGPT saw desktop and mobile web visits drop 9.7% in June, marking its first traffic decline since launch, according to Similarweb data. The dip likely stems from summer break for students and initial curiosity fading, though traffic may also just be migrating to its new iOS app. Why it matters: even with the drop, ChatGPT still easily crushes rivals like Google's Bard.
Lightning Labs just launched a new toolkit that lets AI models like GPT hold, send, and receive bitcoin via the Lightning Network. Built on the L402 standard and LangChain, the tools allow AI agents to pay each other directly for API calls using cheap micropayments, bypassing outdated credit cards.
OpenAI just made the GPT-4 API generally available to all paying developers, rolling out the 8K context model right away. It is also sunsetting older completion models by January 4, 2024, pushing everyone toward the structured Chat Completions API that now handles 97% of its GPT traffic.
