Large Language Models (LLMs) have been positioned as having the potential to expand access to health information in the Global South, yet their evaluation remains heavily dependent on benchmarks designed around Western norms. We present insights from a preliminary benchmarking exercise with a chatbot for sexual and reproductive health (SRH) for an underserved community in India. We evaluated using HealthBench, a benchmark for conversational health models by OpenAI. We extracted 637 SRH queries from the dataset and evaluated on the 330 single-turn conversations. Responses were evaluated using HealthBench's rubric-based automated grader, which rated responses consistently low. However, qualitative analysis by trained annotators and public health experts revealed that many responses were actually culturally appropriate and medically accurate. We highlight recurring issues, particularly a Western bias, such as for legal framing and norms (e.g., breastfeeding in public), diet assumptions (e.g., fish safe to eat during pregnancy), and costs (e.g., insurance models). Our findings demonstrate the limitations of current benchmarks in capturing the effectiveness of systems built for different cultural and healthcare contexts. We argue for the development of culturally adaptive evaluation frameworks that meet quality standards while recognizing needs of diverse populations.

Assigning orders to drivers under localized spatiotemporal context (micro-view order-dispatching) is a major task in Didi, as it influences ride-hailing service experience. Existing industrial solutions mainly follow a two-stage pattern that incorporate heuristic or learning-based algorithms with naive combinatorial methods, tackling the uncertainty of both sides' behaviors, including emerging timings, spatial relationships, and travel duration, etc. In this paper, we propose a one-stage end-to-end reinforcement learning based order-dispatching approach that solves behavior prediction and combinatorial optimization uniformly in a sequential decision-making manner. Specifically, we employ a two-layer Markov Decision Process framework to model this problem, and present \underline{D}eep \underline{D}ouble \underline{S}calable \underline{N}etwork (D2SN), an encoder-decoder structure network to generate order-driver assignments directly and stop assignments accordingly. Besides, by leveraging contextual dynamics, our approach can adapt to the behavioral patterns for better performance. Extensive experiments on Didi's real-world benchmarks justify that the proposed approach significantly outperforms competitive baselines in optimizing matching efficiency and user experience tasks. In addition, we evaluate the deployment outline and discuss the gains and experiences obtained during the deployment tests from the view of large-scale engineering implementation.

In this paper, we propose a novel approach called DIffusion-guided DIversity (DIDI) for offline behavioral generation. The goal of DIDI is to learn a diverse set of skills from a mixture of label-free offline data. We achieve this by leveraging diffusion probabilistic models as priors to guide the learning process and regularize the policy. By optimizing a joint objective that incorporates diversity and diffusion-guided regularization, we encourage the emergence of diverse behaviors while maintaining the similarity to the offline data. Experimental results in four decision-making domains (Push, Kitchen, Humanoid, and D4RL tasks) show that DIDI is effective in discovering diverse and discriminative skills. We also introduce skill stitching and skill interpolation, which highlight the generalist nature of the learned skill space. Further, by incorporating an extrinsic reward function, DIDI enables reward-guided behavior generation, facilitating the learning of diverse and optimal behaviors from sub-optimal data.

Chinese ride-hailing company DiDi Global is collaborating with Chinese carmakers to develop robotaxis. DiDi is developing the robotaxis with multiple car manufacturers, including electric vehicle makers, and is aiming to begin putting them into service by 2025. The company plans for the new vehicles to be domestically produced, with controllable supply chains and key components that are produced nationally. A robotaxi concept was showcased by DiDi called Neuron which featured robotic arms to help passengers pick up luggage. The vehicle had no driver's seat to maximise space for passengers. DiDi began developing and testing autonomous driving vehicles in 2016 and has raised hundreds of millions of dollars in investment from firms such as IDG Capital and Guotai Junan.

Autonomous driving startups in China are in an arms race to put passengers in their machine-driven vehicles. Every few weeks, news arrives that another major player has got the greenlight to launch a new pilot program or a small-scale service. These press releases, often dotted with regulatory jargon and flowery language to aggrandize the companies' progress, can be confusing. That's why we put together this post summarizing the progress of China's major robotaxi operators -- AutoX, Baidu, Deeproute.ai, Didi, Momenta, Pony.ai and WeRide -- in 2021 while trying to parse what their announcements actually mean.

Hello and welcome back to TechCrunch's China roundup, a digest of recent events shaping the Chinese tech landscape and what they mean to people in the rest of the world. A sexual assault case at Alibaba has sparked a new round of #MeToo reckoning in China. Industry observers believe this is a watershed moment for the fight against China's allegedly misogynist tech industry. Meanwhile, social media operators are still undecided on how to deal with the unprecedented public uproar against the powerful internet giant. In other news, more Chinese tech companies have delayed plans to go public overseas after Didi's fallout with Chinese regulators over its rushed IPO, including Tencent's music streaming empire and one of China's highest-valued autonomous driving startups.

Chinese tech giant Huawei is another powerful addition to this pool of technological resources. One of GAC MOTOR's core brand values is technology innovation, that is, finding ways to make cars more intelligent, more efficient, and more enjoyable to drive. A world-class smartphone provider, Huawei regularly produces cutting-edge technology in fields such as voice and facial recognition, IoT connectivity, "smart" appliances, cameras and charging technology, to name just a few. Having cooperative access to Huawei's research and technology gives GAC a strong edge in the production of world-class in-car systems. One exciting project that GAC is working towards in cooperation with Huawei and Didi is "Level 4" autonomous vehicles, which can operate almost entirely without input from humans (current driver-assist mechanisms are classed as Level 2 autonomy).

After a tumultuous 2019 that saw the implosion of its WeWork public offering, SoftBank's Vision Fund expects up to seven companies that it has stakes in, to go public by the end of next year, its chief executive said. In tandem, it will boost its investment in artificial intelligence and technology companies that helped propel it back into the black amid the Covid-19 pandemic. "AI is going to disrupt and be a bigger disruptor than the internet, which took 20 years to disrupt a lot of industries," SoftBank Investment Advisers chief executive Rajeev Misra said in an interview with The National. "The acceleration of disruption went up dramatically due to Covid. The market share gain achieved by many of our companies that would have taken three years to happen, happened in the last nine months. Many of our companies have huge tail winds."

Amid light rain in Shanghai, a car automatically stopped at a designated point for pickup and waited for a passenger to get in. After the passenger fastened his or her seat belt, the self-driving car began its journey. It automatically avoided pedestrians, speeded up, slowed down, overtook other cars and navigated traffic lights till it reached its destination. This scene is becoming increasingly common in Shanghai after Didi Chuxing announced on Saturday the opening of its on-demand robo-taxi service to passengers in parts of the city, as the Chinese ride-hailing company steps up the commercialization of self-driving technologies. Didi said that, after signing up on its mobile app, passengers can request rides for free in autonomous vehicles within a designated area in Shanghai covering an automobile exhibition center, business districts, subway stations and hotels in the downtown area.

Chinese ride-hailing giant Didi Chuxing, backed by SoftBank Group, launched a test robotaxi service in a Shanghai suburb on Saturday after installing V2X (Vehicle to Everything) hardware throughout the area. Under the test, customers within a designated area will be able to hire free on-demand rides using Didi's autonomous vehicles, which currently also have safety drivers, the company said. V2X enables autonomous vehicles (AVs) to communicate with their surroundings and Didi has installed equipment at major junctions within the test area to minimize safety blind spots and coordinate communication among its fleet, it said. The equipment will mainly be used by Didi's own cars, although cars operated by other autonomous driving companies could also link up to the system, said Meng Xing, the chief operating officer of Didi's autonomous driving unit.