The AI-ronman 🚀

Greetings, humans! Ready to explore the world where algorithms have more personality than your last date? 💔 🤖 

Quick Takes ⚡

1. Say hello to Gemini Live: Google’s answer to ChatGPT’s voice mode

2. Grok-2 launches: xAI adds image features to its chatbot

3. Ascend 910C: Huawei's answer to Nvidia's H100 in China's AI chip market

4. ChatGPT's secret upgrade: Faster responses and improved quality

5. Iran's cyber tactics: Microsoft exposes AI-driven influence on U.S. elections

6. OpenAI revamps coding benchmark with SWE-bench verified

Deep Dive 🔍

• Say hello to Gemini Live: Google’s answer to ChatGPT’s voice mode🎤
  
  Google has launched Gemini Live, a voice feature aimed at competing with ChatGPT's Advanced Voice Mode. 🎙️ Announced at the Made by Google 2024 event, this tool allows users to engage in detailed voice conversations on their smartphones. With ten new realistic voices and a responsive speech engine, Gemini Live supports hands-free operation and can adapt to interruptions in real time. Although it currently lacks multimodal input, this feature is expected later this year, along with expanded language support and iOS compatibility. Gemini Live is available through the Google One AI Premium Plan for $20 a month. 💰

• Grok-2 launches: xAI adds image features to its chatbot 📸
  
  xAI has rolled out Grok-2 and Grok-2 mini in beta, enhancing the previous Grok-1.5 model with improved reasoning and new image generation features. 🎨 These models are available exclusively to Premium and Premium+ users on 𝕏, with Grok-2 designed for better chat and coding capabilities. However, the image generation feature has raised concerns due to the absence of restrictions on political figures, which could lead to misuse. xAI plans to integrate these models into 𝕏’s features and offer them through an enterprise API soon.
  
  

  

  
  
  

• Ascend 910C: Huawei's answer to Nvidia's H100 in China's AI chip market 🏆
  
  Huawei is close to introducing its Ascend 910C, a new AI chip designed to rival Nvidia's H100 in China's market, despite U.S. sanctions. The chip, which is being tested by major Chinese companies, is claimed to be comparable to Nvidia's offering. U.S. restrictions have prevented Nvidia from selling advanced chips to China, opening the door for Huawei's 910C. Initial negotiations suggest potential orders exceeding 70,000 chips, valued at around $2 billion. Aiming to start shipping in October, the Ascend 910C could reshape the AI chip industry in China. 🚀

• ChatGPT's secret upgrade: Faster responses and improved quality! ⚡
  
  OpenAI has confirmed that ChatGPT has received a secret upgrade with the new GPT-4o model, enhancing its performance significantly. 🚀 Announced on 𝕏, this version offers quicker response times and improved interaction quality. While the specific changes remain undisclosed, it is separate from the recent API update aimed at reducing costs for developers. Users can access the upgraded features through the ChatGPT app, but it’s unclear if the mini version has been updated as well. A ChatGPT Plus subscription is required for full access to the new capabilities. 🎉
  
  

  

• Iran's cyber tactics: Microsoft exposes AI-driven influence on U.S. elections 🗳️
  
  Microsoft has alerted that Iranian actors are stepping up their cyber influence tactics, utilizing generative AI to impact the upcoming 2024 U.S. presidential elections. 🤖 The Microsoft Threat Analysis Center (MTAC) has pinpointed various Iranian groups, including Sefid Flood, which impersonates social and political activists, and Mint Sandstorm, which carries out phishing attacks against campaign staff. The Storm-2035 network is also active, operating fake news sites that disseminate divisive messages. While some groups are experimenting with AI-generated content, many are reverting to simpler, more effective digital manipulation strategies. ⚠️
  
  

  

• OpenAI revamps coding benchmark with SWE-bench verified 👨‍💻
  
  OpenAI has launched ‘SWE-bench Verified,’ a new subset of the SWE-bench coding benchmark developed in collaboration with its authors. 🎉 This human-validated version addresses previous issues, featuring 500 samples verified by developers. GPT-4o achieves a 33.2% success rate, while the top open-source scaffold, Agentless, improves to 16%. Nonetheless, the partnership raises concerns about potential conflicts of interest amid ongoing discussions about ‘Project Strawberry.’ 🔍
  
  

  

That’s it for this edition! As they say, “AI is like a toddler with a PhD”—cute but unpredictable! 🤷‍♂️ Until next time, keep your algorithms sharp and your jokes sharper!

Ciao for now!

Author: Poonam 👧 

Karan 😎 🚀 
CEO, GlazeGPT.com