Callista AI Weekly (May 5 - 11)
AI Use Cases and General News
Robotaxis on Uber: Pony.ai and Uber announced a strategic partnership to deploy Pony’s autonomous robotaxis on the Uber ride-hailing app. Later this year in a Middle Eastern city, Uber users may be offered a self-driving Pony.ai car (with a safety driver initially) for certain rides. This collaboration marries Pony’s advanced autonomous driving AI with Uber’s platform to make driverless ride-hailing a reality at scale, expanding to more regions over time. Uber’s CEO said the deal will help “introduc[e] the benefits of autonomous mobility to the world” via Uber’s global network.
Automating Construction Bids: Zurich-based startup Scalera – an ETH Zurich spin-off – secured a $6.5 million seed investment to grow its AI-powered tendering platform. Scalera’s SaaS uses AI as a “bidding assistant” for construction projects, automating tedious procurement tasks like parsing thousand-page PDFs and emailing suppliers. By eliminating this “madness,” Scalera helps construction firms and governments act faster, cut costs, and streamline the supply chain. The fresh funding will fuel expansion into Germany and Austria, reflecting investor confidence in AI solutions for the infrastructure sector.
Customer Service & Retail: Yelp began piloting AI voice agents to answer calls for restaurants and home services. These voice bots can handle routine inquiries – like reservations or business hours – and even add customers to waitlists or send follow-up texts. Importantly, they pass off complex questions to human staff and provide call transcripts afterward for quality control. Early tests indicate such AI agents could help businesses capture sales they’d otherwise miss when phones go unanswered.
Biotech R&D: In a significant enterprise collaboration, Benchling (a cloud platform for life sciences) and vaccine pioneer Moderna expanded their partnership to create an “AI-ready” research platform. Announced May 7, this effort will bring hundreds of Moderna’s scientists onto a unified digital R&D system built for AI-driven analysis. The goal is to streamline experimental workflows and standardize data so that AI models can more rapidly assist in drug discovery. Both companies see AI as crucial for faster innovation in medicine – but only if researchers have the right data infrastructure in place.
Media & Content: Streaming giant Netflix is enhancing how users find content through generative AI. On May 7, Netflix announced a revamp of its TV and mobile apps, including a new AI-powered search feature in testing. Subscribers will soon be able to ask Netflix for movie or show recommendations in natural language – for example, “I feel like watching an upbeat comedy” – and get tailored suggestions. This conversational search aims to make it easier to discover content matching a user’s mood or niche interests, illustrating how AI can improve customer experience in entertainment platforms.
Swiss Banks Embrace AI: A new study from Lucerne University (HSLU), released May 9, found that most Swiss retail banks are now actively using AI. About 70% of surveyed banks reported deploying AI in some forms. The most common uses are in internal processes – for example, automating payment processing, document handling, and data analysis for risk management. A smaller but growing number of banks are also using AI for customer-facing functions like chatbots in call centers or basic financial advice on accounts and cards. The report notes that AI is enabling more efficient operations (such as faster credit checks or personalized service offers) and that banks plan to integrate AI further into areas like fraud detection and wealth management. However, the authors caution that human expertise remains essential, especially for sensitive client interactions or complex financial decisions. They also advise banks to invest in robust IT security, as reliance on AI and third-party tech platforms could introduce new vulnerabilities. Overall, the Swiss banking sector – traditionally conservative – appears to have overcome its initial reticence and is now riding the AI wave to stay competitive.
Economic Outlook for AI Adoption: A study by IBM referenced in Swiss media this week highlighted a sobering reality: only 25% of companies worldwide are seeing the expected profits from their AI investments so far. Many businesses have invested heavily in AI software and talent, yet a majority haven’t realized significant returns yet, due to factors like implementation challenges and skills gaps. Swiss businesses are no exception – adoption is high (85% of executives surveyed remain optimistic AI will deliver efficiency gains by 2027), but translating pilot projects into bottom-line impact is work in progress. This pragmatism is evident in Switzerland’s approach: local firms tend to run careful trials (for example, insurance companies using AI to augment risk assessments, or manufacturers deploying AI for predictive maintenance on a small scale first). The takeaway is that while AI is clearly the future, Swiss business leaders are tempering enthusiasm with realistic assessments of what it takes to integrate AI successfully.
Major Vendor Updates
The past week saw significant moves from AI industry leaders – from new models and features to high-profile partnerships – underscoring an accelerating race in AI capabilities:
OpenAI’s Global Push and Microsoft Restructuring: OpenAI unveiled a new “OpenAI for Countries” initiative to help nations build AI infrastructure and customized AI models “rooted in democratic…values.” Announced May 7, the plan offers partnerships to set up local datacenters and tailor ChatGPT for public services like healthcare and education – a bid to counter “authoritarian versions of AI” and expand AI’s benefits globally. At the same time, OpenAI is amid a corporate restructuring that will strengthen its non-profit governance. Leaked details indicate OpenAI aims to reduce Microsoft’s equity revenue share from 20% to 10% by 2030, revising the terms of its multibillion-dollar partnership with Microsoft. This move – part of a shift from a capped-profit model to a public-benefit corporation – would give OpenAI’s original non-profit more control over strategy. Microsoft, which integrates OpenAI’s tech across Azure and Office, is negotiating to ensure long-term access even as OpenAI seeks more independence. The developments highlight OpenAI’s balancing act: pursuing its mission and global growth while managing its key investor relationship.
Anthropic’s Claude Gets “Agentic”: Anthropic, another major AI lab, rolled out new capabilities for its Claude AI assistant that inch closer to autonomous “agent” behavior. On May 7 it launched a web search API that lets Claude automatically browse the internet for up-to-date information. When enabled, Claude can decide a user query needs fresh data, then generate search queries, retrieve and analyze results, and reply with cited sources. In essence, Claude can now conduct research on the fly, refining its searches in multiple steps – even performing “multiple progressive searches” agentically to answer complex prompts. This upgrade, along with new plugin-like “Integrations” to connect Claude with third-party apps (announced last week), shows vendors racing to give their AI models more tools and autonomy. Anthropic’s moves keep it competitive with OpenAI and Google by offering clients an AI that can tap real-time knowledge beyond its training data.
China’s AI Titans Regroup: Leading Chinese AI companies are making news with new models and inventive use-cases. Baidu, which launched its Ernie 4.5 model last month, drew attention with a more whimsical initiative – filing a patent for an AI that could translate animal sounds into human language. The system would analyze an animal’s vocalizations, behavior and vitals, then use AI to infer the creature’s emotional state and “semantic meaning” – potentially letting humans “talk” with pets or livestock in the future. While still in research phase, Baidu’s patent (published this week) underscores the expansive horizons of AI R&D. Meanwhile, Alibaba is reportedly preparing to debut Qwen 3, an upgraded version of its flagship generative model, aiming to reclaim leadership in China’s AI race amid rising domestic competition. And Elon Musk’s new venture xAI, though quiet this week, recently joined an investment partnership to build out AI supercomputing infrastructure – a sign that major players (from NVIDIA to Big Tech and Musk’s outfit) are collaborating on the “AI factories” and cloud capacity needed for next-gen models. In sum, global vendors large and small are doubling down on AI – whether through better models, novel features, or strategic tie-ups – as the demand for advanced AI capabilities shows no signs of cooling.
AI Governance
As AI adoption surges, regulators and governments are scrambling to set rules or leverage the technology for public benefit. This week brought developments on multiple fronts:
US Eases Chip Export Ban – For Now: In Washington, the Trump Administration moved to recalibrate U.S. export controls on advanced AI chips. President Trump suspended strict chip export restrictions that were introduced in January (under the prior Biden Administration) and would have taken effect mid-May. Those rules would have set a tight cap on computing power exported to nearly every country (Switzerland included) – a policy Swiss officials warned could “severely affect” Switzerland’s innovation and economy. After heavy lobbying by Switzerland and other allies, the White House lifted the blanket restrictions for “certain countries” including Switzerland, Israel, Mexico and Portugal. A spokesperson admitted the old rules were too complex and stifled American innovation. New, simpler regulations will be drafted to prevent adversaries’ access to AI chips without unduly harming U.S. industry. Meanwhile on Capitol Hill, lawmakers are still focused on tightening security around chip exports to China. U.S. Senator Tom Cotton introduced a “Chip Security Act” bill on May 9 that would require export-controlled AI chips to have built-in location tracking to detect any unauthorized diversion or smuggling. Exporters would also have to report if their chips end up in unexpected places. This one-two punch – loosening ally restrictions while boosting monitoring – illustrates the fine line regulators walk to protect national security without hindering the global AI supply chain.
Government AI Initiatives: In Massachusetts, Governor Maura Healey used the IBM Think conference to announce a $31 million state grant for AI R&D infrastructure and named an official director for the state’s new AI Hub, aiming to make Massachusetts a leader in AI innovation. And in Europe, the EU’s sweeping AI Act is steadily advancing – with initial provisions on banned AI practices and transparency starting to take effect this year, and new guidelines emerging for foundation models. Policymakers are increasingly vocal: this week U.S. senators held hearings on “Winning the AI Race,” where OpenAI CEO Sam Altman and others urged a balanced approach to AI competitiveness and safeguards. The overarching trend is clear – AI governance is now a top-tier agenda item. We see a mix of encouragement (funding AI talent and public projects) and caution (drafting rules on chips, usage, and safety) as governments try to harness AI’s benefits for society while managing its risks.
Breakthrough AI Research
Rapid progress in AI research continues, with new academic and corporate breakthroughs unveiled this week that promise to enhance what AI can do and how efficiently it can do it:
AI-Designed DNA for Gene Control: In a milestone for biotech, researchers at the Centre for Genomic Regulation in Spain reported the first successful use of generative AI to design DNA sequences that control gene expression. Using a custom AI model, the team “dreamed up” synthetic DNA fragments not found in nature, which could switch a target gene on or off only in specific cell types. When they introduced these AI-generated sequences into mouse cells, the cells behaved exactly as the AI predicted – fluorescing in the intended cells while leaving others unaffected. This proof-of-concept demonstrates a powerful and precise new way to give cells specific instructions, akin to “writing software for biology.” In the future, such AI-designed regulators could let gene therapy developers boost or silence genes only in the desired tissues, making treatments more effective with fewer side effects. It’s an exciting step in the emerging field of “generative biology,” showing how AI’s creative ability can unlock innovations in medicine.
Cutting AI’s Energy Cost: On the computing side, engineers at Oregon State University unveiled a new chip design that could halve the energy consumption of running large AI models. They presented a prototype at an IEEE conference that uses on-chip AI techniques to dramatically improve data transmission efficiency. Large language models like GPT-4 consume huge power shuttling data within data centers; the researchers tackled this by training an on-chip machine learning classifier to clean up signal errors in data transfers, replacing the power-hungry circuits normally used. In tests, their experimental chip used 50% less energy than conventional designs for high-speed data communication. This approach – essentially embedding a tiny AI to optimize the larger AI’s hardware performance – earned a Best Paper Award and could be key to greener AI. If such efficiency gains scale up, data centers running AI services might handle the same workloads with a fraction of the electricity, addressing mounting concerns about AI’s carbon footprint and cost. It’s a reminder that AI advances are not only about algorithms, but also the behind-the-scenes engineering that makes those algorithms sustainable at scale.
Toward Trustworthy AI in High-Stakes Fields: Researchers are also tackling how AI can be made more reliable for critical applications. One study this week introduced a method to have AI systems better express uncertainty, instead of just giving a confident answer every time. By conveying nuance – essentially saying “I’m not entirely sure” when appropriate – AI predictions could become safer and more useful in fields like medicine or finance where human experts need to know the level of confidence to make decisions. Other teams demonstrated AI tools to improve manufacturing safety and to detect cancers earlier, showing the breadth of “AI for good” research happening now. Many of these breakthroughs are still in the lab, but they foreshadow next-generation AI systems that are more efficient, transparent, and capable – innovations that will likely fuel new business products and services in the coming years.
Conclusion
In just a week, we’ve seen AI news ranging from autonomous taxis and construction SaaS in the business realm, to major strategic shifts by AI powerhouses, to policymakers grappling with how to regulate (and leverage) AI, and scientists pushing its boundaries in health and hardware. The common thread is acceleration – AI is rapidly becoming more embedded in industry and society. For Swiss businesses, these developments carry concrete implications. New AI use cases (like Scalera’s) hint at opportunities to streamline operations or create novel services. Vendor advances mean more powerful AI tools will be at enterprises’ disposal – but also signal fiercer competition, as companies globally adopt the latest models and agentic capabilities. The governance moves, from Washington to Bern to Brussels, remind us that compliance and ethical considerations are evolving alongside the tech; staying ahead may require adjusting strategies to meet upcoming standards or export rules. And the research breakthroughs suggest that today’s challenges – whether the cost of AI or its limitations in sensitive jobs – will gradually be addressed, making AI an even more transformative force. In summary, the AI landscape as of May 2025 is one of dynamic innovation coupled with strategic caution. Business leaders would do well to monitor both tracks: what new AI solutions can do for them, and what guardrails or game-changers are emerging around those very technologies. The pace shows no sign of slowing, and neither should our readiness to adapt.
Sources
ServiceNow unveils AI-powered CRM and autonomous AI agents at its Knowledge 2025 conference.
Axios – “OpenAI launches global push for democratic AI” (Ina Fried, May 7, 2025)
OpenAI – “Introducing OpenAI for Countries” (OpenAI blog, May 2025)
Tech Monitor – “OpenAI to rework Microsoft partnership, reduce revenue share” (May 7, 2025)
TechCrunch – “Anthropic rolls out an API for AI-powered web search” (Kyle Wiggers, May 7, 2025)
Anthropic – “Introducing web search on the Anthropic API” (May 7, 2025)
Reuters – “China’s Baidu looks to patent AI system to decipher animal sounds” (Liam Mo & Brenda Goh, May 9, 2025)
Globe Newswire – “Pony.ai and Uber Announce Strategic Partnership to Advance Autonomous Mobility” (Press release, May 6, 2025)
Inside-IT (CH) – “Schweizer KI-Startup Scalera sichert sich 6,5 Millionen Dollar” (Katharina Jochum, May 9, 2025)
Inside-IT (CH) – “USA heben KI-Chip-Beschränkungen für die Schweiz auf” (Christian Wingeier, May 8, 2025)
Reuters – “US senator introduces bill calling for location-tracking on AI chips to limit China access” (Deborah Sophia, May 9, 2025)
ScienceDaily – “AI-designed DNA controls genes in healthy mammalian cells for first time” (Center for Genomic Regulation, May 8, 2025)
ScienceDaily – “New chip uses AI to shrink large language models’ energy footprint by 50%” (Oregon State Univ., May 8, 2025)
ScienceDaily – “Making AI models more trustworthy for high-stakes settings” (May 6, 2025)
