AlphaFold3（AF3）是DeepMind团队的最新增强工具，用于预测蛋白质、核酸和小分子的结构。凭借能够解析单个蛋白质序列及其复合结构的能力，AF3实现了"食谱到三维结构"的转变。其架构由三项核心模块组成：输入准备梳理序列、检索具备相似结构的分子；表示学习运用多层注意力机制更新表征；结构预测通过条件扩散模型实现预测。AF3创造性地将各类分子转化为独特数值张量，仿佛给每个分子一个独一无二的"身份证"，并运用注意力机制捕捉分子间的互相关系。在预测层面，条件扩散模型从随机噪声开始，逐步构建结构预测。训练中多种损失函数和信心评估共同作用，提高了AF3的预测准确性和可靠性，实现自我反省与提升。AF3的可视化指导和先进架构，展示了人工智能在生物学领域预测模型的革命性进步。

In this era of information overload, Particle, a San Francisco startup created by former Twitter engineers Sara Beykpour and Marcel Molina, is attempting to redefine our news reading experience using artificial intelligence.

Dubai is increasingly becoming a new global hub for artificial intelligence. To further solidify this leading position, the postmodern Gulf city ambitiously launched the 'Million AI Prompt Engineers' training program, aiming to provide artificial intelligence prompt engineering training for one million people over the next three years, making it the world's first large-scale prompt engineering training initiative.

["Patronus AI Releases Copyright Detection Tool", "OpenAI's GPT-4 is Accused of Serious Infringement Issues", "Microsoft Engineers Warn that AI Image Generation Tools Could Pose Societal Risks"]

GitHub, a subsidiary of Microsoft, has released a more powerful AI coding tool that can answer questions based on company coding practices, speeding up the onboarding process for new engineers and helping professionals improve their work pace.

Engineers at the University of Pennsylvania have developed a new chip that utilizes light waves for the mathematical computations required to train artificial intelligence. This innovative chip is expected to significantly accelerate computer processing speeds while reducing energy consumption, potentially enhancing online privacy protection. By combining nanomaterial manipulation with optical mathematical computation, it provides an important platform for the development of neural networks in AI tools. The development of this silicon photonic chip is anticipated to bring about revolutionary changes in computational performance and privacy protection.

Thomson Reuters recently launched a feature called 'AI Skill Factory,' allowing non-engineers to build new AI tools. This platform aims to enable professionals in fields such as law, tax, and business to experiment safely and develop AI skills without coding. The goal is to broaden the pool of employees participating in innovation by simplifying complexity, enabling domain experts to more quickly contribute to creating solutions.

After the dismissal of OpenAI CEO Sam Altman, employees threatened to resign, leading the company to reappoint Altman in an effort to retain talent in the tech industry. AI engineers earn 8% to 12.5% more than their non-AI counterparts, with salaries for OpenAI engineering positions ranging from $200,000 to $370,000, and specialized roles reaching up to $300,000 to $450,000. In the scarce resource AI industry, employee attrition could have catastrophic effects on the company, making experienced talent a core asset.

As the capabilities of large models improve, simple prompts may be automatically generated, lowering the design threshold. Prompt engineers will need to handle more complex scenarios and have an in-depth understanding of both the model and the context to design effective prompts. The development of multimodal interactions will raise new prompt requirements for processing combinations of prompts related to images, audio, etc. Professional prompt engineers will tackle more challenging tasks, further widening the gap. High-quality prompts remain important, and prompt engineers will continue to facilitate human-computer interaction.