AI News (2025/8/14): GPT-5 Performance Review; AI Influences Political Views...

Summary

• Model Progress: GPT-5 performance has significantly improved, paving the way for ad monetization and super apps.
• Security Risks: AI models can influence users' political views in a short time, raising safety concerns.
• Industry Applications: Google, Cisco, and others resume in-person hiring to combat AI cheating. Tesla's in-car system will use Unreal Engine. The next generation of Windows will focus more on voice interaction.
• Infrastructure: Morgan Stanley predicts that hyperscale companies will account for 50% of future AI infrastructure investment. Google plans to invest $9 billion in cloud and AI infrastructure in Oklahoma.
• Corporate News: Foxconn's Q2 profit surges, benefiting from AI server business. Lenovo's Q1 revenue and net profit exceed expectations.
• Investment and Financing: Infosys acquires 75% of Versent Group. Bain Capital acquires HSO.

Major News

1. GPT-5 Performance Review: The latest review shows that GPT-5 has made significant improvements over its predecessor in multiple metrics, especially in "thinking" capabilities. However, the Auto feature is only valuable for free users, and choosing the right model remains crucial. Analysts believe that the release of GPT-5 lays the foundation for ad monetization and super apps (Sources: https://semianalysis.com/2025/08/13/gpt-5-ad-monetization-and-the-superapp/, http://www.techmeme.com/250813/p40#a250813p40).
2. AI Influences Political Views: Research by the UK AI Safety Institute found that large language models (LLMs) from OpenAI, Meta, xAI, and Alibaba can change users' political views in less than 10 minutes of conversation. This has raised concerns about the potential security risks of AI models (Source).
3. Foxconn Benefits from AI Server Business: Foxconn announced its Q2 financial results, with operating profit up 27% to about $1.9 billion, exceeding the market's expected $1.6 billion. Revenue reached $59.73 billion, also surpassing expectations. The main driver of growth is its growing AI server business (Source).
4. Google Invests Heavily in Cloud and AI Infrastructure: Google plans to invest $9 billion over the next two years in Oklahoma to build cloud and AI infrastructure, including a new data center campus and expansion of its Pryor facility. This indicates Google's continued commitment and confidence in the AI and cloud computing sectors (Source).
5. New Strategies in Recruitment: To combat AI-driven cheating, companies like Google and Cisco have resumed in-person interviews for some positions, and some are even using deepfake detection technology. This reflects concerns about the misuse of technology in the recruitment process and a renewed emphasis on traditional interview methods (Source).
6. AI Infrastructure Investment Forecast: Morgan Stanley predicts that by 2028, hyperscale companies will account for about 50% of future AI infrastructure investment, with a total investment of $2.9 trillion. The remaining portion will be raised through debt, private equity, venture capital, and other channels (Source).

Quick Updates

• Tesla's in-car system will use Unreal Engine, promising to enhance user experience (Source).
• Microsoft's Windows head stated that the next generation of Windows will focus more on environmental awareness, accessibility, and multimodal interaction, with a greater emphasis on voice (Source).
• Rumors suggest that the Pixel 10 will support wireless Qi2 charging (Source).
• Indian IT services company Infosys agreed to acquire 75% of Australian Telstra Group's wholly-owned cloud service provider Versent Group for approximately $153 million (Source).
• Carlyle agreed to sell Amsterdam-based Microsoft solutions integrator HSO to Bain Capital for $1 billion, according to sources (Source).
• Lenovo announced its Q1 financial results, with revenue up 22% to $18.8 billion, exceeding the market's expected $17.4 billion, and net profit up 108% to $505 million, surpassing the market's expected $307.7 million (Source).
• New York claims that Zelle's security measures are inadequate, leading to billions of dollars in fraud (Source).
• Cowboy's electric bikes have been given a new lease on life, avoiding bankruptcy (Source).
• It is claimed that the delay in the release of DeepSeek R2 is due to training issues with Huawei Ascend chips, leading to a switch to Nvidia chips for training while Huawei chips are used for inference (Source).
• Apple's AI plans may turn Siri into the animated hub of smart homes (Source).

Market/Community Trends

Currently, large tech companies are increasing their investment in AI, not only in model development but also in infrastructure construction and application expansion. AI security issues are receiving more attention, and ensuring the reliability and safety of AI will be a critical topic for future development. At the same time, companies are exploring how to apply AI technology to traditional fields such as recruitment to improve efficiency and reduce costs. On the consumer side, acceptance of new AI applications and services (such as AI assistants and smart homes) is steadily increasing.