OpenAI Project Mercury, Microsoft Sales Agent, Google Quantum

In the realm of artificial intelligence, OpenAI is actively developing 'Project Mercury,' an AI system aimed at automating entry-level banking tasks, with over 100 former investment bankers training the models for $150 per hour. This initiative seeks to reduce the demanding hours for junior analysts and commercialize AI for the finance industry. Meanwhile, Microsoft has launched the Sales Research Agent within Dynamics 365 Sales, an AI tool for sales leaders to analyze data through natural language, supported by the new Microsoft Sales Research Bench for evaluating AI responses to strategic sales questions. Airbnb's CEO, Brian Chesky, has paused plans to integrate ChatGPT, citing that the AI is not yet ready for their needs, and instead is leveraging Airbnb's own AI, which resolves customer issues in six seconds. In a unique political development, Virginia Republican lieutenant governor candidate John Reid held a 40-minute debate against an AI-generated version of his opponent, Ghazala Hashmi, after she declined all debate invitations. The event, streamed on YouTube, was criticized by Hashmi's campaign as a 'desperate move.' Google has achieved a significant milestone in quantum computing, with a new algorithm demonstrating 'quantum advantage' by computing a molecule's structure 13,000 times faster than supercomputers, though practical applications are still distant. NVIDIA's AI Day in Sydney highlighted Australia's growing AI ecosystem, focusing on sovereign AI, quantum computing, and high-performance computing. Yelp is enhancing its platform with over 35 new AI features for 2025, including an AI Assistant chatbot and AI-powered menu vision. On the security front, Verizon's Mobile Security Index warns of a 'perfect storm' of AI and human error increasing risks, particularly on mobile devices, with only 17% of businesses having defenses against AI-assisted attacks. Finally, a Thomson Reuters Institute report indicates that 67% of corporate tax departments are not yet using generative AI, largely due to resource limitations, though early adopters are seeing benefits in research and strategic work.

Key Takeaways

• OpenAI is training an AI system, codenamed 'Project Mercury,' with over 100 former investment bankers to automate entry-level finance jobs, paying them $150 per hour.
• Microsoft has introduced the Sales Research Agent in Dynamics 365 Sales and the accompanying Sales Research Bench to help sales leaders analyze data and evaluate AI performance.
• Airbnb CEO Brian Chesky has halted plans to integrate ChatGPT, stating it's not ready for their needs, and is instead using Airbnb's own AI that resolves customer issues in six seconds.
• Virginia Republican lieutenant governor candidate John Reid debated an AI version of his opponent, Ghazala Hashmi, after she declined all debate invitations.
• Google's quantum computer has achieved a 'quantum advantage,' computing a molecule's structure 13,000 times faster than supercomputers.
• NVIDIA AI Day in Sydney showcased Australia's expanding AI ecosystem, focusing on sovereign AI, quantum computing, and high-performance computing.
• Yelp is rolling out over 35 new AI-powered features for 2025, including an AI Assistant chatbot and AI-enhanced menu viewing.
• Verizon's Mobile Security Index reports that only 17% of businesses have defenses against AI-assisted attacks, with AI and human error creating significant security risks on mobile devices.
• A Thomson Reuters Institute report found that 67% of corporate tax departments are not using generative AI due to resource constraints, though early adopters report benefits in research and strategy.
• Taiwan excels in AI hardware manufacturing but faces challenges in developing a competitive AI software and application startup scene.

Virginia candidate debates AI opponent after Democrat declines

In the Virginia lieutenant governor race, Republican candidate John Reid held a 40-minute debate against an AI-generated version of his opponent, Ghazala Hashmi. Hashmi declined all debate invitations, prompting Reid to create the AI debate, streamed live on YouTube. Reid's campaign stated the debate aimed for fairness and to show Virginians a contrast of positions. The Democratic Party of Virginia criticized the move as a 'shoddy gimmick,' while Hashmi's campaign called it 'desperate.'

Virginia candidate debates AI bot after opponent refuses

Republican Virginia lieutenant governor candidate John Reid held a 40-minute debate against an AI version of his Democratic opponent, Ghazala Hashmi, after she declined all invitations. The debate, titled 'The Debate That Should Have Happened,' was streamed live on YouTube. Reid's campaign claimed it was created for fairness to show voters a contrast. However, the Democratic Party of Virginia called it a 'shoddy gimmick,' and a spokesperson for Hashmi labeled it a 'desperate move.'

Virginia candidate debates AI opponent when opponent won't attend

Virginia's Republican lieutenant governor candidate John Reid held a 40-minute debate against an AI version of his Democratic opponent, Ghazala Hashmi, because she refused all debate invitations. The event was streamed live on YouTube and called 'The Debate That Should Have Happened.' Reid's campaign stated the AI debate was a fair way to contrast their positions. The Democratic Party of Virginia criticized the event as a 'shoddy gimmick,' and Hashmi's campaign spokesperson called it a 'desperate move.'

OpenAI trains AI to replace entry-level banking jobs

OpenAI is developing an AI system, codenamed 'Project Mercury,' to automate entry-level banking tasks like building financial models for IPOs and restructurings. Over 100 former investment bankers are helping train the AI, earning $150 per hour. The goal is to reduce the long hours junior analysts typically work. This initiative is part of OpenAI's strategy to commercialize its AI tools for specific industries like finance, as the company has not yet reported a profit.

OpenAI hires ex-bankers to train AI for Wall Street tasks

OpenAI is hiring over 100 former investment bankers to train its AI models for tasks like building financial models and creating presentations. This initiative, aimed at automating routine entry-level finance jobs, could significantly change the nature of junior analyst roles. Experts suggest AI will transform, rather than eliminate, these positions, potentially shifting required skills. While some fear job losses, others believe AI will enhance analyst capabilities, allowing them to focus on more complex work.

Microsoft launches Sales Research Bench for AI sales insights

Microsoft introduced the Sales Research Agent in Dynamics 365 Sales, an AI tool that helps sales leaders analyze data through natural language conversations. To evaluate such AI solutions, Microsoft also launched the Microsoft Sales Research Bench. This benchmark assesses how AI responds to strategic sales questions, with plans to release the full evaluation package publicly. The Sales Research Agent aims to provide quick, trustworthy analysis on pipeline, revenue, and other key metrics, transforming how business decisions are made.

Airbnb CEO pauses ChatGPT deal, reveals own AI solves issues in 6 seconds

Airbnb CEO Brian Chesky announced the company is halting plans to integrate OpenAI's ChatGPT, stating the AI is not yet ready for Airbnb's specific needs. Instead, Airbnb is using its own advanced AI, which can resolve customer issues in just six seconds. Chesky emphasized that Airbnb's internal AI development focuses on improving guest and host experiences and streamlining support. This decision highlights Airbnb's commitment to proprietary AI solutions for better control and immediate impact.

Yelp enhances search and user experience with AI features

Yelp is introducing over 35 new AI-powered features in its 2025 Fall Product Release to improve local discovery. The Yelp Assistant chatbot can now answer questions about businesses and remembers user preferences for services. New feature Menu Vision uses AI to show dish photos when users point their camera at a menu. Yelp also added natural language and voice search capabilities for more relevant results and introduced 'Popular Offerings' to highlight trending items at businesses.

Taiwan's AI paradox: strong hardware, weak startup scene

Despite Taiwan's dominance in AI hardware manufacturing, its software and application startup ecosystem struggles to compete with Silicon Valley. Venture capital experts note this 'AI paradox' as the global AI race accelerates. While Taiwan excels in producing the physical components for AI, developing innovative AI software and applications presents significant challenges for its startup companies.

Google's quantum computer achieves breakthrough over supercomputers

Google announced a significant breakthrough in quantum computing with a new algorithm that performed a task beyond the capabilities of conventional supercomputers. This algorithm, detailed in the journal Nature, computed the structure of a molecule 13,000 times faster than classical computers. While this achievement demonstrates 'quantum advantage,' Google acknowledges that practical applications for quantum computers are still years away. Experts note this is a step towards more powerful quantum machines, potentially impacting fields like medicine and materials science.

NVIDIA AI Day Sydney highlights Australia's growing AI ecosystem

NVIDIA AI Day Sydney brought together over 1,000 attendees to discuss the future of AI and its impact on Australia. Technology leaders like Brendan Hopper from Commonwealth Bank of Australia emphasized how next-generation computing drives AI innovation. The event showcased advancements in sovereign AI, quantum computing, and high-performance computing across various sectors, including finance and design. NVIDIA highlighted its role in building Australia's AI infrastructure and fostering collaboration among startups, industry, and government.

Verizon warns AI and mobile devices create security risks

Verizon's latest Mobile Security Index highlights a 'perfect storm' of AI and human error posing significant security risks, especially on mobile devices. Only 17% of businesses have specific defenses against AI-assisted attacks, despite 93% of employees using generative AI on their phones. The report warns of increased 'smishing' attacks via SMS and notes that mobile device security is often less mature than for desktops. Verizon advises companies to provide AI training and implement stronger mobile security measures.

Tax departments slow to adopt AI due to resource limits

A Thomson Reuters Institute report reveals that 67% of corporate tax departments are not yet using generative AI, primarily due to being under-resourced and focused on reactive compliance tasks. While 57% plan to implement AI, early adopters report benefits in research and strategic work, shifting focus from compliance. The report also notes a confidence gap in forecasting and penalty avoidance between well-resourced and under-resourced departments. AI's accessibility offers opportunities for smaller companies, but overall adoption is hindered by resource constraints.

Sources

• Virginia GOP candidate stages ‘debate’ against AI-generated opponent
• Virginia candidate holds debate with AI bot after opponent declines to participate
• Virginia candidate for lieutenant governor debates AI bot when opponent doesn't show
• OpenAI is Training AI to Replace Entry-Level Banking Jobs
• Sam Altman's OpenAI is coming for Wall Street's grunt workers as AI continues to transform the entry level
• Sales Research Bench: Benchmarking AI-Powered Sales Insights
• Airbnb CEO Freezes ChatGPT Deal--Then Unveils AI That Solves Customer Issues in 6 Seconds
• Transforming Yelp with AI: Smarter search, seamless connections, instant answers and a more conversational experience | Yelp
• Taiwan's AI paradox: startup struggles despite hardware dominance
• Google hails breakthrough as quantum computer surpasses ability of supercomputers
• ‘The Next Generation of Compute Is Driving AI,’ Technology Leader Says at NVIDIA AI Day Sydney
• Verizon: AI, mobile devices are brewing ‘perfect storm’ for security
• Why most tax departments aren't using AI yet