Callista AI Weekly (April 21 - 27)
New AI Use Cases
Real-world deployments of AI are accelerating across industries. In logistics, generative AI “agents” are automating complex operations. Global freight forwarder C.H. Robinson revealed its AI agents have executed over 3 million shipping tasks, from instant price quotes to processing orders, cutting workflows from hours to seconds and freeing employees for higher-value work. In retail, Walmart is leveraging OpenAI’s GPT models via Azure to power a conversational shopping assistant that serves customers a personalized list of recommended products, making online browsing more intuitive and tailored. Telecom giant Vodafone reports that AI chatbots and digital assistants now resolve about 70% of customer inquiries, reducing call handling times by over a minute per call – a huge efficiency gain in customer service.
Financial services are similarly embracing AI-driven automation. Yes Bank in India built an AI assistant (“Ask Genie”) for its relationship managers using GPT models, enabling staff to answer client queries instantly with accurate, up-to-date information. In Europe, Zurich Insurance developed an AI solution for risk assessment that automates underwriting decisions with higher accuracy. This has sped up policy underwriting significantly and boosted customer satisfaction by shortening turnaround times. In the public sector and fintech, innovative use cases are emerging: a Swiss nonprofit created a multilingual AI agent to help refugees manage personal finances and paperwork, assisting thousands of people daily in navigating banking and government services.
Even traditionally cautious professions like law and accounting are now on board. According to a new Thomson Reuters survey, 22% of professional services firms are actively using generative AI – nearly double the rate from a year ago. These firms are applying AI to draft documents, research cases, summarize records, and prepare advisory reports, augmenting their experts’ productivity. Notably, 95% of corporate professionals surveyed expect AI to be central to their workflows within five years. In practice, large consultancies and tech providers are offering “AI as a service” to help any company implement such solutions. For example, Kyndryl (the IT services spinoff of IBM) launched new AI Private Cloud offerings to help enterprises identify high-ROI AI opportunities, build custom models, and deploy them securely on private infrastructure. Collectively, these use cases show AI moving from pilot programs to mainstream operations – improving customer experiences, cutting costs, and unlocking new business value across logistics, retail, finance, telecom and beyond.
Major Vendor Updates and New Models
OpenAI is making waves with both product releases and strategy shifts worth mentioning again. Earlier this month, OpenAI launched GPT-4.1, an upgraded suite of GPT models boasting major improvements in coding ability, long-text comprehension, and following complex instructions. The flagship GPT-4.1 can ingest up to 1 million tokens (roughly 800,000 words) of context, far more than prior models. This lets businesses feed massive documents or datasets into the AI for analysis, enabling applications like extensive contract review or large-scale data summarization. GPT-4.1 is also 21% better at coding tasks than previous GPT-4 versions, making it ideal for enterprise software development support. Importantly, OpenAI cut the model’s operating cost and rolled out “mini” and “nano” variants for affordability. All of this means companies can integrate powerful AI into products and workflows more cheaply and at greater scale than before.
Now OpenAI is looking beyond its API business to something truly new: an open-source AI model. In a surprising move, CEO Sam Altman revealed plans to release a freely downloadable “open” model – OpenAI’s first in roughly five years. Slated for early summer 2025, this model is expected to rival the likes of Meta’s LLaMA in performance. What really sets it apart is a proposed “handoff” capability: the open model could automatically call on OpenAI’s cloud-hosted larger models when it needs extra reasoning power. In essence, a lightweight local AI could delegate tough questions to a heavyweight AI in the cloud – much like a junior analyst asking a senior expert for help. This agent-like feature (inspired by Apple’s hybrid on-device AI approach) would let companies use the open model on-premises for speed and privacy, while still tapping into GPT-4-level power on demand. If executed, OpenAI’s strategy could pull more developers into its ecosystem and set a new standard for AI interoperability. The company is training this model from scratch and considering a very permissive license for it, signaling a notable philosophical shift toward openness. For businesses, an open, free GPT-quality model – with optional cloud boosts – could be a game-changer, reducing dependence on closed APIs and expanding AI use cases behind corporate firewalls.
Microsoft latest Wave 2 is signalling a shift from a single AI helper to a hub where humans delegate work to specialised AI agents. Highlights worth noting:
Redesigned Copilot app – the home screen now opens in Chat, mirroring the consumer version and adding adaptive memory so Copilot remembers preferences and context.
Copilot Search – an enterprise-grade search that answers in natural language and can pull signals from third-party platforms like Slack, Jira, Google Drive and ServiceNow.
Agent Store – a built-in marketplace for Microsoft, partner or custom agents (e.g., the new Researcher and Analyst deep-reasoning agents) that organisations can roll out via the Frontier early-access programme.
Notebooks & Pages – project-centric workspaces that let teams feed Copilot a curated set of docs, links and chats, then co-edit results in real time—think Loop meets AI.
Create canvas – an evolution of Microsoft Designer powered by OpenAI’s GPT-4o, generating images, videos and surveys directly inside Office docs.
Roll-out & access – new features begin shipping to enterprise tenants from late May 2025; users can summon Copilot with the dedicated keyboard key or Win + C shortcut on Windows 11 PCs.
Microsoft frames this release as the foundation for the “Frontier Firm,” where every worker becomes an “agent boss”—building, delegating to and supervising fleets of AI agents to amplify impact. For readers, the message is clear: Copilot is evolving from an assistant to a full-fledged AI operations console for modern work.
Alphabet Highlights AI in Strong Earnings: Google’s parent Alphabet reported robust Q1 2025 results, with profit up 50% year-on-year and revenue topping forecasts. On an earnings call, Google emphasized that its heavy investments in AI are “powering returns” in its core ad business, helping drive ad revenue growth despite economic uncertainties. Executives reaffirmed ambitious AI expansion plans and maintained a $75 billion capital expenditure outlook (largely for AI and cloud infrastructure), signaling to investors that the AI boom at Google is far from finished. Google Cloud’s growth slowed slightly (28% YoY), but the company soothed analysts by linking future ad and cloud growth to continued AI innovation and tools like its Gemini model. Google also revealed that users are rapidly embracing its AI-powered search features. For example AI Overviews – the AI-generated summaries at the top of search results – now serve over 1.5 billion users each month, reflecting two years of steady expansion of this feature. (AI Overviews compile web results into a quick answer, for queries like “What is generative AI?”, although publishers have noted some traffic cannibalization.) Google also recently began testing an “AI Mode” in Search that allows conversational follow-up questions, its answer to chat-style search engines. Additionally, since late 2024 Google has integrated ads into AI Overview results, indicating the company’s confidence in monetizing these AI search experiences.
Smaller AI labs are making headlines too. Anthropic, known for its Claude chatbot, recently unveiled models with an emphasis on lengthy, step-by-step “reasoning.” In fact, Anthropic’s latest Claude iteration can maintain extremely long conversations or document summaries without losing context, aiming to function as an AI that “thinks as long as you want.” This caters to enterprise needs for analyzing long reports or multi-hour meeting transcripts. Meanwhile, Elon Musk’s xAI is ramping up fast. In a major finance story, Musk’s newly merged entity xAI Holdings (which now encompasses the X social platform as well) is reportedly in talks to raise a staggering $20 billion in fresh funding. Such an infusion would value Musk’s AI venture at over $120 billion, catapulting it into the top tier of AI players. Musk has folded Twitter (now “X”) into this venture to combine a rich real-time data source with AI development. For businesses, xAI’s trajectory is worth watching: the capital would likely fuel development of a “TruthGPT” or similar large model with uniquely broad social media training data. Musk’s vision is an AI that is maximally truth-seeking and which could power everything from better content discovery on X to Tesla’s autonomous systems. If xAI succeeds, it could spawn new AI services plugged into the X platform or offered via Tesla and SpaceX ecosystems, adding a competitive alternative to OpenAI/Google for enterprise developers.
In China, the AI arms race is prompting rapid advancements from tech giants. Baidu – China’s search and AI leader – this week launched its latest generative models, ERNIE 4.5 Turbo and a companion reasoning model ERNIE X1 Turbo. ERNIE 4.5 Turbo is Baidu’s newest foundation model for Chinese (and multi-language) generative AI, delivering improved performance in tasks like open-ended Q&A and content creation. ERNIE X1 Turbo is optimized for logical reasoning and decision support. Critically, Baidu has also slashed the cost of using its AI models, aiming to undercut rivals and drive adoption. According to Bloomberg reporting, the new Ernie versions are offered at significantly lower price points than previous APIs, making AI-powered services more affordable for businesses in China. The timing is strategic: competition is fierce with other Chinese players (like Alibaba, Tencent, and startup DeepSeek), and U.S. firms are somewhat hampered in China. Baidu is further strengthening its position by open-sourcing parts of its technology; it announced plans to open-source the Ernie model series by mid-year, allowing developers to freely build on its code. For Chinese enterprises, these moves lower barriers to deploying AI in products and may reduce reliance on Western models that face regulatory hurdles.
On the hardware front, Huawei is tackling the critical challenge of AI chips, which power model training and deployment. This week we learned Huawei will begin mass shipments of its new AI accelerator chip, the Ascend 910C, as soon as May. The 910C GPU is designed as a domestic alternative to Nvidia’s high-end processors, which are now heavily restricted for China. Huawei achieved performance on par with Nvidia’s A100/H100 by cleverly combining two earlier chips into one package, doubling the computing power and memory. Chinese AI firms, starved of top-tier Nvidia silicon due to U.S. export controls, are poised to adopt Huawei’s chip as their main AI engine. In fact, Huawei is already prototyping an even more advanced Ascend 910D chip that it hopes will exceed Nvidia’s flagship H100 in performance. If successful, this could be a breakthrough: the Wall Street Journal reports Huawei is recruiting local tech companies to test the 910D soon. For the global business community, Huawei’s progress signals that China’s AI sector may overcome hardware bottlenecks, ensuring continued development of competitive AI models. Multinationals operating in China could benefit from a growing ecosystem of local AI cloud services not reliant on U.S. chips. Conversely, U.S. firms may face new Chinese competitors whose entire AI stack – from chips to apps – is homegrown.
AI Governance
With AI’s rapid deployment come urgent questions of governance, regulation, and ethical guardrails. Policymakers worldwide took notable steps (and stances) this week to shape an environment where innovation can thrive without undue risk.
In Europe, the landmark EU AI Act is moving from legislation to implementation – and officials are fine-tuning its impact on businesses. The European Commission is now seeking ways to ease compliance burdens on startups and small firms under the AI Act. A newly unveiled “AI Continent Action Plan” proposes to “minimize the potential compliance burden of the AI Act, particularly for smaller innovators.” . This reflects concern that an avalanche of AI rules could stifle young companies. While the AI Act (agreed in principle late last year) imposes strict transparency and safety requirements on “high-risk” AI systems, EU tech officials want to ensure the rules don’t unintentionally become red tape for Europe’s nascent AI startups. The Commission has opened consultations with industry on how to simplify and clarify obligations. For example, EU regulators this week published draft guidelines for general-purpose AI (foundation models like GPT) ahead of the Act’s full enforcement. These preliminary guidelines, released on April 22, outline how providers of large AI models should handle transparency, data governance, and risk management in practice. They will inform a formal Code of Conduct expected by August. The message to business: Europe remains committed to “trustworthy AI” but is listening to feedback to avoid a one-size-fits-all approach. Companies operating in the EU should prepare for new reporting and testing duties for their AI systems – but also possible exemptions or streamlined requirements if they are small or using low-risk AI.
Transatlantic tensions around AI regulation are also coming into focus. The United States (which has taken a lighter-touch, voluntary approach so far) is wary of Europe’s prescriptive rules. In fact, recent reports revealed the U.S. government quietly urged Brussels to scrap or soften its upcoming AI code of practice – a voluntary rulebook that goes beyond the EU AI Act’s baseline. U.S. officials argue that Europe’s draft “AI rulebook” (being developed with industry and due by April’s end) could stifle innovation and impose extralegal requirements, such as mandates for third-party model audits and full training data disclosure. Essentially, Washington fears Europe adding “teeth” to voluntary guidelines that could de facto become global standards, much as GDPR did for privacy. American tech firms share these concerns, pointing out that some obligations in the draft code – like extensive transparency on model training – exceed what the AI Act law itself will require.
In addition, the White House pushes AI Education: In the policy arena, a new U.S. Executive Order prioritizes AI education and workforce development. President Trump established a White House Task Force on AI Education to coordinate efforts across agencies to “promote AI literacy” nationwide. It also creates a Presidential AI Challenge . By integrating AI concepts early and upskilling educators, the initiative seeks to cultivate the next generation of AI-skilled workers and ensure the U.S. remains competitive in the AI era.
In China, AI governance is intertwined with the state’s strategic and social objectives. President Xi Jinping this week emphasized the need to “speed up” laws and regulations for AI to ensure it remains “safe, reliable, and controllable”. In a April 25 Politburo session dedicated to AI, Xi doubled down on China’s dual mandate: rapidly advance indigenous AI capabilities (to rival U.S. tech supremacy) and tightly manage AI’s societal impacts. He lauded domestic innovators like startup DeepSeek, which recently unveiled a powerful reasoning AI built with fewer computing resources – evidence that China can narrow the gap despite chip sanctions. At the same time, Xi warned of AI’s risks and called for a “risk warning and emergency response system” for AI incidents. We can expect China to roll out new guidelines for AI ethics, content moderation (especially with generative media), and perhaps licensing requirements for providers. Indeed, last year China implemented rules requiring recommendation algorithms and deepfake tools to register with authorities and abide by certain values. For businesses in China, compliance with government mandates on data security and censorship in AI models will remain a core requirement. Yet Xi also noted AI shouldn’t be a “game of rich countries” and urged international cooperation on AI governance. This suggests China will continue participating in global AI forums (even as it charts its own course at home).
Lastly, Switzerland is carving out its approach to AI governance, aiming for a pragmatic middle ground. In mid-April, the Swiss federal government indicated it will not rush into sweeping new AI regulations, preferring targeted measures that address clear risks without hampering innovation. The Swiss Federal Council has observed the EU’s moves but signaled a “wait and see” stance – strengthening existing frameworks (e.g. data protection, product safety laws) to cover AI where possible, and only introducing new rules if absolutely necessary. This balanced approach is meant to protect citizens from harms like biased algorithms or unsafe AI in healthcare, while encouraging Swiss companies to continue pioneering AI solutions. As a highly innovative economy, Switzerland recognizes the competitive advantage of AI integration and thus seeks to avoid overregulation that could disadvantage its firms. That said, Swiss authorities are actively engaging in international AI standards discussions and assessing where adaptations might be needed, for instance in liability law or transparency obligations for AI systems. Companies operating in Switzerland can likely expect guidance and best practices from the government in lieu of hard regulations in the short term, along with support for AI research and talent development to keep Switzerland at the forefront.
Breakthrough Research
AI research continues to break new ground, with implications that promise to further empower businesses. A notable development this week came from a collaboration of the Institute for Basic Science and Max Planck Institute: researchers unveiled a brain-inspired AI method that makes computer vision models both more accurate and significantly more efficient. Dubbed Lp-Convolution, the technique allows AI image recognition systems to dynamically adjust how they “focus” on visual features, mimicking the human visual cortex. In tests, this innovation boosted image classification accuracy on standard benchmarks while using fewer computational resources. The practical upshot is that future AI vision applications – from quality inspection cameras on factory lines to retail inventory scanners – could run faster and on cheaper hardware, yet see and interpret the world more like a human expert. For businesses, that means lower infrastructure costs and more robust performance in AI-powered vision systems (which translates to fewer errors in detecting defects or recognizing products, even under suboptimal conditions).
Another broad trend in research is the rise of smaller, specialized models that rival giant models. A state-of-industry report from Stanford University’s AI Index (discussed in Nature this month) found that thanks to algorithmic advances, a modern AI model with 100× fewer parameters can now achieve performance comparable to a top model from just two years ago. This dramatic leap in efficiency is enabling what one might call “right-sized AI” – models that are not absurdly large, but smart enough for specific tasks and far easier to deploy. For companies, this is hugely encouraging: rather than only relying on trillion-parameter behemoths owned by a few Big Tech firms, they can increasingly train or fine-tune mid-sized models that meet their needs, whether it’s a customer-service bot or a supply chain optimizer. These sleek models require less data and compute power, lowering the barrier to entry for AI-driven improvements. We’re already seeing this in the open-source community, with projects like Meta’s LLaMA and DeepSeek’s R1 model showing it’s possible to deliver high performance without hyper-scale resources.
Research is also pushing AI into new frontiers of autonomy and decision-making – so-called agentic AI. The initial wave of generative AI (typified by ChatGPT) was mostly about content generation and single-turn interactions. Now, researchers and startups are exploring AI agents that can plan multi-step goals, take actions, and continuously learn. OpenAI’s integration of a “handoff” mechanism in its upcoming model is one example of building more agent-like behavior (by letting an AI decide to consult a stronger AI). In the open-source world, experimental systems like “AutoGPT” (which chains GPT calls to iteratively work on tasks) have garnered attention among developers. Enterprises are keen on this evolution: a recent industry survey found 96% of large companies are expanding their use of AI agents in 2025, reflecting the belief that autonomous AI processes can handle operations at a scale and speed beyond human abilities. These agents could manage workflows such as triaging IT tickets, monitoring financial transactions for fraud, or even autonomously conducting market research. While still early, the operational impact is expected to surpass that of static generative AI. Businesses experimenting in this area report significant efficiency gains – for instance, the logistics AI agents at C.H. Robinson not only execute tasks but also learn from each freight shipment, getting “more capable every day” as they optimize routes and decisions. Ongoing research into safe and reliable agents will be critical, but their progress hints at a future where many back-office and analytical chores could be offloaded entirely to AI driven processes.
Finally, AI continues to accelerate scientific discovery in ways that can spur new industries. This week, an interdisciplinary team at Emory University described a new AI model that identifies candidate materials for high-temperature superconductors in record time. And in biotech, AI systems are now proposing protein structures and drug molecules that might have taken humans years to find. One striking example reported recently: an AI system analyzing cancer research data independently hypothesized a treatment combination that was later confirmed in the lab – essentially converging on a genuine scientific discovery on its own. Such breakthroughs underscore the point that AI is becoming a partner in R&D. Companies in pharmaceuticals, materials engineering, and energy are beginning to use AI “co-pilots” in their labs to test ideas virtually at high speed, drastically shortening innovation cycles. The business impact of these research advances could be profound: faster development of new products, reduced R&D costs, and even the creation of entirely new markets (e.g. quantum materials discovered by AI or personalized medicines designed by AI). While these are early days, the trajectory of AI research suggests that the competitive gap will widen between organizations that harness these AI-driven breakthroughs and those that do not. Savvy businesses are already forging partnerships with AI research labs and adopting tools from cutting-edge studies to stay on the forefront of innovation.
Swiss AI Developments
Switzerland, known for innovation, is making notable moves in AI on both the industry and research fronts. A new Swiss National AI Institute, jointly led by ETH Zurich and EPFL Lausanne, was announced with the ambitious goal of creating a “trustworthy Swiss AI” – essentially a homegrown large language model tailored to high standards of reliability and multilingual Swiss context. To power this effort, Switzerland has deployed ALPS, a world-class public AI supercomputer at the national supercomputing center (CSCS) in Lugano. With ALPS providing massive GPU horsepower, Swiss researchers intend to train advanced foundation models locally, giving Switzerland strategic autonomy in AI similar to how it has in cryptography. This initiative is expected to spin off AI capabilities that benefit Swiss industry and government (for example, language models fluent in French, German, Italian, and Romansh to serve all national languages). It also emphasizes “trustworthiness” – aligning with Swiss values of precision and safety, which could make the outputs particularly attractive for sensitive sectors like healthcare or finance that demand high reliability and data privacy.
Swiss companies, meanwhile, are early adopters of AI solutions – outpacing many of their global peers. According to Microsoft’s Work Trend Index 2025 study, 52% of Swiss enterprises are already using AI-based agents to automate workflows, compared to about 46% globally. From Swiss banks employing AI for fraud detection to manufacturing firms using AI for predictive maintenance, adoption is widespread. Swiss business leaders appear convinced of AI’s transformative impact: about 80% of Swiss executives expect to significantly adapt their business models and strategies in response to AI advances by the end of this year. This proactive stance is supported by a workforce that is relatively AI-savvy (44% of Swiss employees claim familiarity with AI tools, higher than international averages)). In practice, we see examples like Zurich Insurance (mentioned above) implementing AI in underwriting, and Swiss multinational Nestlé using AI to optimize supply chains and personalize marketing. The country’s vibrant startup scene has also produced AI firms in robotics, medtech, and fintech that draw on Switzerland’s strengths in precision engineering and data security.
However, Switzerland is also navigating external challenges. One pressing issue is access to cutting-edge AI chips. Recent U.S. export control decisions – initially by the Biden administration and now continued under President Trump – will limit Switzerland’s access to advanced Nvidia AI chips from mid-May 2025. The U.S. sees even neutral Switzerland as part of a broader strategy to prevent certain high-end technologies from possibly flowing to adversaries. Swiss industry groups, like economiesuisse, have raised concern that such restrictions could hamper Swiss AI research and competitiveness if not addressed. They urge diplomatic solutions to ensure Swiss companies and universities can still obtain the latest GPUs essential for training AI models. The Swiss government is now in talks with Washington to seek exemptions or alternatives, underscoring the importance of hardware sovereignty in the AI era. In the interim, Switzerland’s investment in the ALPS supercomputer (which presumably was provisioned with advanced chips before the cutoff) is a timely asset.
Overall, Switzerland’s approach to AI can be characterized by enthusiastic adoption and cautious stewardship. The country is leveraging its talent and infrastructure to build AI solutions that align with its values and needs, from multilingual trustworthy models to enterprise AI integrations. Simultaneously, Swiss regulators are monitoring developments to ensure that while innovation flourishes, risks are kept in check via existing laws and a possible “middle path” regulatory stance. For Swiss businesses, this balanced environment – high innovation, high trust, and measured oversight – provides a strong foundation to capitalize on AI advancements.
Conclusion
The week of April 21–27, 2025 showcased how fast the AI landscape is evolving – and how deeply it is entwining with business strategy. On the ground, companies big and small unveiled tangible AI-driven gains: from automated shipping operations and smart retail assistants to AI copilots for bankers and lawyers. These real use cases underscore that generative and agentic AI are no longer science projects, but practical tools delivering efficiency, cost savings, and new capabilities in day-to-day operations. The competitive pressure to adopt AI is mounting, as early movers report significant ROI (studies now peg the return at about $3.70 for every $1 invested in GenAI, on average). Businesses that successfully weave AI into their workflows are improving customer satisfaction and unlocking capacity for growth.
At the same time, the AI product ecosystem is in overdrive. The major AI vendors – OpenAI and its allies, the tech titans of Silicon Valley, and counterparts in Asia – are racing to outdo each other with more powerful models, more accessible services, and integrated solutions. For enterprises, this means an expanding menu of options: be it choosing a cloud AI service, deploying an open-source model on-premises, or leveraging built-in AI functions in software they already use. The emphasis on agentic capabilities in new offerings (like Microsoft’s proactive Copilot or OpenAI’s handoff-enabled model) hints that the next generation of AI tools will be increasingly action-oriented. We can expect enterprise software soon to feature AIs that don’t just advise, but actually execute tasks under human oversight – effectively co-workers that take initiative. Wise companies will pilot these autonomous agents carefully, harnessing their productivity benefits while establishing guardrails (to intervene if the AI goes off course).
In summary, the business impact of this week’s developments can be captured in a single theme: AI is increasingly agentive, pervasive, and governed. Forward-looking businesses are deploying AI not just as a tool but as a collaborative agent in their operations. Vendors are delivering ever more embedded and autonomous AI features. Regulators are shaping the boundaries within which all this innovation will unfold. And researchers are expanding AI’s capabilities, ensuring the technology’s trajectory remains steep. For business leaders, staying updated on these fast-moving trends is not optional – it’s essential to crafting strategies that harness AI’s transformative power while managing its risks. The winners of this new era will be those who can integrate AI’s latest advances into their core processes, adapt to the evolving regulatory landscape, and cultivate trust among customers and stakeholders in how they use AI. As April 2025 shows, the future is racing toward us, and it’s powered by artificial intelligence.
Sources:
Maxwell Zeff, “OpenAI wants its ‘open’ AI model to call models in the cloud for help.” TechCrunch, April 24, 2025.
Kyle Wiggers, “OpenAI seeks to make its upcoming ‘open’ AI model best-in-class.” TechCrunch, April 23, 2025.
Deborah Sophia, “OpenAI launches new GPT-4.1 models with improved coding, long context comprehension.” Reuters, April 14, 2025.
Jeffrey Dastin, “Microsoft, turning 50, dials up Copilot actions to stay in AI game.” Reuters, April 4, 2025.
Jeffrey Dastin, “Google rebrands Bard chatbot as Gemini, rolls out paid subscription.” Reuters, Feb 8, 2024.
“Baidu launches new AI model amid mounting competition.” Reuters, April 25, 2025.
Fanny Potkin and Che Pan, “Exclusive: Huawei readies new AI chip for mass shipment as China seeks Nvidia alternatives.” Reuters, April 22, 2025.
Reuters News Wire, “Elon Musk’s xAI in talks to raise $20 billion from investors – Bloomberg.” Reuters, April 26, 2025.
James Pomfret and Summer Zhen, “China’s Xi calls for self sufficiency in AI development amid U.S. rivalry.” Reuters, April 26, 2025.
Foo Yun Chee, “Europe wants to lighten AI compliance burden for startups.” Reuters, April 8, 2025.
Lucas Mearian, “US wants to nix the EU AI Act’s code of practice, leaving enterprises to develop their own risk standards.” Computerworld, April 25, 2025.
“Brain-inspired AI breakthrough: Making computers see more like humans.” ScienceDaily, April 22, 2025.
Press Release: “At C.H. Robinson, Artificial Intelligence Has Now Performed Over 3 Million Shipping Tasks.” C.H. Robinson Newsroom, April 16, 2025.
Press Release: “From Incubation to Integration: Generative AI Adoption Nearly Doubles in Professional Services.” Thomson Reuters, April 15, 2025.
Filip Sinjakovic, “Schweizer Unternehmen sind Vorreiter beim Einsatz von KI.” Netzwoche (Switzerland), April 24, 2025.
