🗞 OpenAI Develops CriticGPT to Improve AI-Generated Code 🖥️

🔎 The Latest on the AI Frontier:

• OpenAI Develops CriticGPT to Improve AI-Generated Code 🖥️

• Meta Creates Compact AI Model for Mobile Devices 📱

• Microsoft Maintains AI Access in China, Diverging from OpenAI 📊

• Microsoft Expands AI Education in Hong Kong Schools 🏫

• Amazon's AI-Powered Alexa Upgrade Faces Delays 🗣️

• Samsung Wins Advanced AI Chip Order from Japanese Firm 💻

• Google AI Expert Predicts Human-AI Merge by 2045 🧠

• Ex-Meta Scientists Launch Giant AI Protein Design Model 🧬

• Healthcare Experts Urge Caution on AI Implementation 🏥

• Former Google Employee Launches AI Video Creation Studio 🎬

🖥️ OpenAI develops CriticGPT to catch errors in ChatGPT-generated code. Link

• CriticGPT is a fine-tuned version of GPT-4 trained to critique code generated by ChatGPT, outperforming human evaluators in catching bugs and producing better critiques.

• This is part of OpenAI's efforts on "scalable oversight" to help evaluate and improve AI model outputs as they become more advanced.

• Human judges preferred CriticGPT's critiques 80% of the time compared to human-generated ones, suggesting it could be valuable for future AI model training and improvement.

📱 Meta develops compact AI language model for mobile devices. Link

• MobileLLM is designed for smartphones and resource-constrained devices, with fewer than 1 billion parameters compared to trillion-parameter models like GPT-4.

• The model outperforms previous models of similar size by 2.7% to 4.3% on common benchmark tasks, using innovations like prioritizing model depth and novel weight-sharing techniques.

• This development aligns with growing interest in efficient AI models, potentially enabling more advanced AI features on personal devices in the future.

📊 Microsoft won't follow OpenAI in blocking China's access to AI models. Link

• Microsoft Azure will continue providing access to its AI models for eligible customers in China, even as OpenAI plans to block API access in the country starting July 9th.

• Microsoft operates in China through a joint venture and has stated publicly that its AI models remain available to customers there.

• This contrasts with OpenAI's reported move to restrict access in China, highlighting the different approaches of the two companies in the Chinese market.

🏫 Microsoft expands AI in Hong Kong schools as OpenAI pulls back. Link

• Microsoft partnered with a Hong Kong primary school to provide AI tools to 220 students in grades 5-6 for science classes, using OpenAI's GPT and DALL-E models.

• The company plans to expand AI offerings to all Hong Kong schools in the coming years, maintaining access to OpenAI's models despite API restrictions in the region.

• This educational push comes as Microsoft faces scrutiny over its China operations back in the US, with the company offering to relocate 700-800 employees from China to other markets.

🗣️ Amazon's promised AI-powered Alexa upgrade seems delayed. Link

• In September 2023, Amazon demoed an advanced AI-powered Alexa with more human-like conversation abilities.

• 10 months later, there's little visible progress on this "conversational Alexa," while competitors like Apple and Google push forward with AI assistants.

• Amazon claims to have integrated some generative AI into Alexa components, but provides no details on noticeable improvements.

• The company reaffirms its commitment to developing Alexa as "the world's best personal assistant" using generative AI, but the timeline remains unclear.

💻 Samsung secures cutting-edge AI chip order from Japan's Preferred Networks. Link

• Samsung Electronics has won an order to produce AI chips using its advanced 2-nanometre foundry process and chip packaging service for Japanese AI company Preferred Networks.

• The chips will utilize high-tech gate all-around (GAA) architecture and integrate multiple chips in one package to enhance speed and reduce size.

• This marks Samsung's first disclosed order for its cutting-edge 2-nanometre chip manufacturing process, showcasing its competitive edge in advanced semiconductor production.

🧠 Google's AI visionary predicts merging of human and artificial intelligence by 2045. Link

• Kurzweil believes AI will reach human-level intelligence by 2029 and surpass it shortly after.

• He envisions nanobots interfacing our brains with AI, expanding intelligence "a millionfold" by 2045.

• Kurzweil also predicts radical life extension technologies, allowing humans to potentially live indefinitely and even "resurrect" the dead digitally.

🧬 Ex-Meta scientists debut giant AI protein design model, raise $142 million. Link

• EvolutionaryScale's ESM3 model, trained on 2.7 billion protein sequences and structures, is one of the largest AI models in biology.

• The model successfully created new fluorescent proteins, demonstrating its potential for designing novel biological molecules.

• With $142 million in new funding, the company aims to apply the technology to drug development, sustainability efforts, and other biological applications.

🏥 AI's role in healthcare under scrutiny: experts call for balanced approach. Link

• Cancer specialists warn against overhyping AI as a "magic bullet," urging focus on fundamental issues like staffing and proven treatments.

• New DSIT secretary Peter Kyle aims to make the department a hub for digital expertise, enhancing government services through technology.

• HR platform Lattice plans to onboard an AI "employee," potentially revolutionizing how businesses integrate AI into their operations.

🎬 Ex-Googler and filmmaker launch DreamFlare, an AI-generated video studio. Link

• DreamFlare aims to help creators make and monetize short-form AI-generated video content using third-party AI tools like Runway and Midjourney.

• The platform offers two types of content: "Flips" (comic book-style stories) and "Spins" (interactive choose-your-own-adventure films).

• DreamFlare has raised $1.6 million in funding and claims partnerships with entertainment industry executives, though some remain anonymous due to AI content controversy.

Have any feedback? Send us an email