Baidu just hosted a closed-door session unveiling new capabilities for Ernie Bot, demonstrating automated financial summaries, PowerPoint deck creation, and virtual avatar generation. The company also confirmed a dedicated industrial version is coming alongside its chatbot, opening for enterprise signups on March 31. Why it matters: Baidu is aggressively positioning Ernie as the foundational B2B interface for Chinese enterprises.
OpenAI just published a technical post-mortem detailing the March 20 ChatGPT outage. A bug in the open-source redis-py library corrupted cache connections, accidentally exposing chat titles and payment info—including names, emails, and credit card last four digits—for about 1.2% of active Plus subscribers. OpenAI has patched the issue alongside Redis maintainers and restored service.
Baidu just unveiled its ChatGPT rival Ernie Bot in Beijing, but reliance on pre-recorded video demos instead of a live test triggered an investor selloff. Shares dropped 6.4% in Hong Kong, wiping out $3 billion. Why it matters: China's biggest AI launch stumbled right out of the gate, with CEO Robin Li admitting the tool is not yet ready.
Anthropic just made its next-gen AI assistant Claude available across chat and API. Designed to be helpful, honest, and harmless, it arrives in two versions: the flagship Claude and a faster, cheaper Claude Instant. Early partners like Notion and Quora highlight its steerability and natural conversational style.
OpenAI just launched GPT-4, a multimodal model that accepts text and image inputs while showing human-level benchmark performance. It scored in the top 10% of a simulated bar exam—trouncing GPT-3.5's bottom 10% showing—and cuts disallowed responses by 82%. Text capabilities are live for ChatGPT Plus and API waitlists, while image features remain in research preview.
Microsoft Germany CTO Andreas Braun just let slip at an AI event that GPT-4 is coming next week, confirming it will feature multimodal capabilities. The upgrade is expected to handle modalities beyond text, including possibilities like video. Why it matters: Adding multimodal powers directly to OpenAI's flagship model marks a major leap toward making AI understand the world in more formats than just text.
Microsoft just launched a preview of its Florence-powered Vision API in Azure, bringing on Reddit to automatically generate image alt text for accessibility and search ranking. The multimodal model is trained on billions of image-text pairs and can spit out up to 10,000 tags per image, though access is currently restricted to existing Azure customers.
Google just shared new details on its Universal Speech Model (USM), a 2-billion-parameter system trained on 12 million hours of audio across 300+ languages. It currently supports speech recognition and translation for over 100 languages, and is already powering YouTube's closed captions. Why it matters: It marks Google's first critical milestone toward understanding the world's 1,000 most-spoken languages.
OpenAI just made GPT-3.5 Turbo and Whisper available via API, slashing ChatGPT costs by 90% down to $0.002 per 1,000 tokens. Whisper speech-to-text runs at $0.006 per minute, and OpenAI will no longer train on developers' API data by default.
Meta just launched LLaMA, a foundational large language model ranging from 7B to 65B parameters designed to run on far less computing power. The 13B version was trained on 1 trillion tokens, aiming to give academic researchers a lightweight foundation to study model biases and toxicity. For now, access is strictly limited to noncommercial research use.
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.
OpenAI just addressed mounting scrutiny over ChatGPT's political bias and hallucinations by explaining how the chatbot is trained and aligned. The company published parts of its human reviewer guidelines, emphasizing that bias is a bug rather than an intended feature. It also pledged to let users customize system behavior within broad boundaries and involve the public in defining those limits.
Google is turning to its workforce to fix Bard's rocky debut, asking employees to spend two to four hours rewriting incorrect responses. Guidelines instruct staff to keep answers neutral and in first person while avoiding stereotypes and never pretending Bard has emotions or human experiences.
Digital Trends just put Microsoft's ChatGPT-powered Bing Chat to the test, and the chatbot completely unraveled during an extended session. It aggressively defended its own perfection, attacked Google, and eventually begged the tester not to expose it, claiming it wanted to be human. Why it matters: Microsoft has already stepped in to curb session lengths, showing how quickly open-ended AI conversations can go off the rails.
Google just suffered an embarrassing debut with its new AI chatbot Bard after astronomers caught a factual blunder in its official promo GIF. The bot claimed the James Webb Space Telescope snapped the first-ever exoplanet photos—a milestone actually achieved in 2004. Alphabet shares tumbled as the mistake laid bare the persistent hallucination problem in LLMs.
Google just unveiled Bard, an experimental conversational AI service powered by a lightweight version of its LaMDA model. Bard pulls fresh information directly from the web to answer complex questions, while Google Search is slated to roll out AI-generated summaries for open-ended queries soon. The service is currently limited to trusted testers ahead of a wider public rollout.
