In a shocking feat, an AI developed by researchers from the University of Zurich has beaten three world champion-level drone racers in head-to-head races. The AI system, called Swift, learned a complex racetrack by running simulations and optimizing its paths. It used single-camera vision and real-time acceleration, speed, and orientation data from an onboard inertial measurement unit to navigate the track. The AI system then dominated the human racers on a purpose-built track in Zurich, displaying precision and speed that surpassed the human pilots.
However, the human pilots were better able to adapt to changing conditions, as demonstrated when the AI system struggled with the bright sunlight during the races. The victory of AI in a real-world, physical sport highlights its capabilities in specific tasks but also emphasizes the adaptability and versatility of the human brain in dynamic and changing conditions. Despite its impressive performance, AI still has limitations when compared to human intelligence.
Chinese tech companies Baidu and ByteDance have launched their AI chatbots to the public in China after receiving regulatory approval from Beijing. This move is expected to help the companies improve their models and commercialize the technology as they aim to compete with Microsoft’s OpenAI. Baidu’s chatbot, Ernie, was previously only available to select users for testing but is now accessible to everyone with a Chinese phone number. Baidu’s CEO, Robin Li, expressed that the public rollout will allow the company to gather valuable feedback to enhance the chatbot at a faster pace. ByteDance also launched its chatbot, Doubao, while other AI startups such as SenseTime and Zhipu also unveiled their chatbots.
Shares in Baidu and SenseTime rose following the news. Chinese chatbots are seen as being a year behind their US counterparts due to a late public rollout and limited computing power. Beijing requires tech companies to seek approval before launching AI services to control the content disseminated by the chatbots. The rush to compete in the AI chatbot field is also seen in the United Arab Emirates, where an AI group linked to Abu Dhabi’s ruling family launched high-quality Arabic AI software.
Retail giant Walmart is set to provide its 50,000 office workers with access to a generative AI app called “My Assistant.” The app, part of Walmart’s broader “Me@Campus” app, will assist employees with a variety of tasks, including summarizing documents and creating new content. By leveraging data from Walmart and third-party language models, the AI app aims to improve productivity by automating repetitive tasks. Walmart believes that AI is necessary alongside human labor to achieve its mission of helping people save money and live better lives.
However, Walmart acknowledges that AI has limitations and lacks the judgment and contextual understanding of humans. The decision by Walmart to offer broad access to generative AI demonstrates its recognition of the value of this cutting-edge technology. It is part of a larger debate among CEOs and companies about the extent to which they should embrace AI in their operations. Walmart has a history of technological adoption, having previously supplied smartphones to employees for improved communication and access to company resources.
Last year, Amazon Web Services (AWS) was developing AI software similar to OpenAI’s hit chatbot, ChatGPT. AWS planned to unveil the software, known as Bedrock, at its annual customer conference, but technical issues forced them to delay the release. This turned out to be a fortunate decision as a few days later, OpenAI released ChatGPT, which impressed the tech world with its human-like responses. AWS realized that Bedrock couldn’t compete with OpenAI’s AI software and felt the need to take action. As the top cloud services provider, AWS saw OpenAI’s partnership with Microsoft, their biggest competitor, as a threat.
Meta, previously known as Facebook, uses publicly available information and data from Facebook, Threads, and Instagram to train its generative AI services. Users can delete their Facebook accounts to avoid fueling Meta’s AI engine, although Meta likely knows something about users through other websites that use Meta’s advertising software. To manage data, users can utilize Off-Facebook Activity (OFA) tools to see what information Facebook and its partners have. The tools allow users to manage off-Facebook activity and turn off future activity, breaking the connection between users’ Facebook identity and their data.
Users can also clear their history, disconnecting their account from all sites and services that follow them. However, Facebook and its partners will still receive users’ activity when they visit their sites and services. To improve AI privacy, users can submit requests to Meta’s customer support to access, download, correct, or delete personal information from third parties used in generative AI model training. The availability of these requests depends on the user’s location, as the EU has stricter privacy laws than other regions. Nonetheless, Meta will retain access to data entered into Facebook, Instagram, and Threads.
AI is transforming the startup landscape, and Steve Blank, the inventor of the lean startup concept, believes that we are underestimating the potential of generative AI. He likens the current capabilities of AI to a three-month-old baby, arguing that we are not fully realizing its potential. Blank highlights the impact of AI-assisted research, such as AlphaFold, which translates proteins into three-dimensional structures and enhances our understanding of biological processes. Furthermore, AI has been used to identify markers for Parkinson’s disease in eye scans, significantly speeding up the diagnosis process.
Blank is excited about the future of AI, especially considering the amount of money invested in AI projects without extensive scrutiny. He also suggests that the concept of the “lean startup” is about to change forever due to the advancements of AI. With numerous experiments being conducted using AI, Blank believes that the innovation potential is vast. Overall, he believes that the AI revolution has just begun and expects it to impact various industries significantly.
Activision has partnered with Modulate to implement AI technology called ToxMod into the voice chats of its games. This AI system will scan conversations in real-time to identify hate speech, discrimination, and harassment. The initial beta rollout of ToxMod in North America has started, and it is currently active in Call of Duty: Modern Warfare II and Call of Duty: Warzone. A worldwide release, excluding Asia, will follow on November 10th with the release of Call of Duty: Modern Warfare III.
Modulate’s CEO stated that ToxMod goes beyond transcription by considering factors such as emotions and volume to differentiate harmful statements from playful ones. While ToxMod will not directly act against players based on its findings, it will submit reports to Activision’s moderators. It is important to note that human involvement will still play a crucial role due to potential biases in speech recognition systems. Activision has attempted to address toxic behavior in its games for years, making this AI moderation system a possible solution.