Germany’s Aleph Alpha just released benchmarks showing its 70-billion-parameter Luminous-Supreme model matches OpenAI's 175B GPT-3 across most tasks with less than half the parameters. The European startup is also testing a 300-billion-parameter model dubbed Luminous-World, slated for release later this year to challenge GPT-4.
Scammers are flooding the App Store and Google Play with fake ChatGPT apps designed to steal user data and drain bank accounts. An investigation by Top10VPN revealed that several clone apps log private chat histories, track device fingerprints, and charge users up to $15,000 a year for features OpenAI offers for free. Users are urged to check their phones and delete them.
A Twitch channel called Nothing, Forever has been broadcasting an endless, AI-generated Seinfeld parody for over a month. Powered by GPT-3 and generative algorithms, the system scripts dialogue, voices lines, and directs camera angles in real time. It is glitchy and deeply surreal, but it offers a fascinating look at the future of infinite, automated entertainment.
Meta’s chief AI scientist Yann LeCun just pushed back on ChatGPT hype, arguing the viral bot is not particularly innovative in terms of underlying science. LeCun credited OpenAI with doing a stellar engineering job combining pre-existing technologies like Transformers and self-supervised learning, noting that established giants hold back because they have much more to lose when models invent facts.
OpenAI reportedly relied on outsourced workers in Kenya earning less than $2 an hour to review horrific text—including child abuse, torture, and self-harm—to keep ChatGPT clean, according to a Time investigation. Workers reported enduring lasting mental trauma before the contractor cut ties early. The takeaway: generative AI's impressive polish still relies on poorly paid human labor bearing the dark side of the web.
OpenAI has built a working prototype to secretly watermark ChatGPT text using cryptographic token tweaks invisible to humans. The signature needs just a single paragraph of text to flag AI-generated content—and holds up even if sentences are reordered or lightly edited. Why it matters: It could give teachers and platforms a reliable way to bust AI plagiarists and automated bot farms.
Former Meta engineer just launched QuickVid, a tool that chains together GPT-3, DALL-E 2, and Google text-to-speech to churn out TikTok and YouTube Shorts from a single prompt. It automates scripts, b-roll, voiceovers, and overlays in one go. Why it matters: It turns content creation into a push-button assembly line, raising fresh concerns over automated feed spam.
Researchers at Drexel University just demonstrated that OpenAI's GPT-3 can predict early-stage Alzheimer's disease with 80% accuracy simply by analyzing spontaneous speech transcripts. The text-embedding method beat traditional acoustic tests and outperformed standard models by 20% in cognitive score predictions. The team plans to turn the concept into a lightweight pre-screening web app for clinics and homes.
Princeton researchers just introduced ATLA, a method that uses GPT-3-generated descriptions of tool shapes and functions to train robotic arms. Providing language context allowed simulated arms to master unfamiliar tools significantly faster. Why it matters: large language models are proving to be an effective shortcut for giving physical robots common-sense adaptability.
Tome just rolled out generative AI features powered by OpenAI's GPT-3 and DALL-E 2 to build presentations from scratch. Users type a simple prompt to get a structured narrative, draft text, and matching images in seconds. Why it matters: It turns blank-page dread into an editable first draft, and it is currently free to try in beta.
