Recent studies and developments in the field of Artificial Intelligence (AI) reveal both progress and limitations. Apple researchers have released findings indicating that current AI models, including those from OpenAI, Anthropic, and DeepSeek, struggle with complex reasoning and problem-solving, particularly when faced with unfamiliar scenarios. Their performance declines as the complexity of tasks increases, suggesting that AI's 'thinking' ability may be an illusion. Despite these limitations, AI is being integrated into various applications. Apple's Xcode 26 now includes ChatGPT for coding assistance, and the company has introduced an AI-powered Shortcuts app. However, an AI-powered Siri update has been delayed to ensure quality standards are met. In the business world, Meta is considering a significant investment in Scale AI, while Broadcom's semiconductor business is increasingly driven by AI chip sales. AI is also transforming retail, with AI-driven insights boosting sales performance. Concerns around AI security and compliance are driving growth in private cloud investment. From a societal perspective, U.S. bishops have urged Congress to develop AI in a way that serves humanity and respects human dignity. However, the use of AI in political campaigns, such as the AI-generated images used by a Queens City Council candidate, raises ethical questions. Furthermore, disagreements over AI regulation and energy policy may be causing Democrats to lose support from the tech industry. On a positive note, a PwC study suggests that AI is creating more jobs, increasing productivity, and enhancing workers' value.
Key Takeaways
- Apple researchers found that current AI models struggle with complex reasoning and problem-solving, especially in unfamiliar situations.
- AI models' accuracy decreases as the complexity of tasks increases, suggesting limitations in their 'thinking' abilities.
- Apple's Xcode 26 now integrates ChatGPT for coding assistance, and an AI-powered Shortcuts app has been introduced.
- An AI-powered Siri update has been delayed to ensure it meets Apple's quality standards.
- Meta is considering a multi-billion dollar investment in Scale AI.
- Broadcom's semiconductor business is increasingly driven by AI chip sales.
- AI is transforming retail by providing insights that boost sales performance.
- Concerns about AI security and compliance are driving growth in private cloud investment.
- U.S. bishops are advocating for AI development that serves humanity and respects human dignity.
- A PwC study indicates that AI is creating more jobs, increasing productivity, and enhancing workers' value.
Apple study generative AI can't think like humans
Apple researchers found that advanced AI models, like Claude 3.7 Sonnet Thinking and DeepSeek-R1, struggle with complex problem-solving. While these models outperform standard AI on moderately complex tasks, they fail completely as the difficulty increases. The study suggests current AI has limits in generalizing reasoning like humans do. Researchers noted that models sometimes shortened their reasoning even with enough computing power, and performance depended on training data familiarity.
Apple's research says AI reasoning has limits
Apple researchers released a paper that says AI reasoning isn't as good as people think. The study found that AI models from OpenAI, Anthropic, and DeepSeek struggle with new and complex puzzles. Their ability to reason increases with problem difficulty but then declines. The models face accuracy problems when dealing with things they haven't seen before.
Apple says AI models still lack human-level reasoning
Apple researchers found that current AI models can't reason like humans. Leading AI models like ChatGPT and large reasoning models (LRMs) have limitations. Apple's study, 'The Illusion of Thinking,' showed that these models struggle with reasoning beyond standard math and coding tasks. The researchers tested Claude Sonnet, OpenAI's o3-mini and o1, and DeepSeek-R1 and V3 chatbots and found that LRMs' accuracy drops with complexity.
Apple study finds AI models fail with complex problems
Apple researchers discovered that advanced AI models struggle with complex problems. The study found that standard AI models perform better on simple tasks, but all models fail when tasks become highly complex. Large reasoning models (LRMs) even reduced their reasoning effort as they neared failure. The Apple paper questions the race to achieve artificial general intelligence (AGI).
Apple paper casts doubt on AI industry's reasoning claims
Apple researchers released a paper questioning the reasoning abilities of top AI models. The paper suggests that the AI industry overstates how well models like OpenAI's o3, Anthropic's Claude 3.7, and Google's Gemini can 'reason'. Apple's team says that the ability to reason is just an 'illusion of thinking'. The researchers found that these models struggle with problems beyond a certain level of complexity.
AI reasoning models fail with complicated problems, says report
A new report by Apple researchers says large reasoning models (LRMs) fail when faced with complex tasks. These AI models, designed to 'think' step-by-step, don't perform better than standard models when problems become too complicated. Researchers tested the models using puzzles like the Tower of Hanoi. They found that the AI's performance dropped significantly as the puzzles became more complex.
Apple study says reasoning AI doesn't really reason
Apple's new study reveals that reasoning AI models don't actually reason. The tests showed that AI models like ChatGPT GPT-4 and Claude 3.7 Sonnet can't adapt to unfamiliar problems. Apple used puzzles like Tower of Hanoi to test the models. The AI models gave up on solving the puzzles at harder levels, suggesting they don't truly think.
Apple says AI 'thinking' is just an illusion
Apple AI researchers found that the 'thinking' ability of large reasoning models collapses when things get complicated. The study tested models like OpenAI’s o1/o3, DeepSeek-R1, Claude 3.7 Sonnet Thinking, and Gemini Thinking. These models break problems into smaller steps but struggle with complex challenges. The researchers designed a puzzle environment to test the models and found they often failed at higher complexity levels.
Apple integrates ChatGPT and AI models into Xcode
Apple's Xcode 26 now includes OpenAI's ChatGPT for coding and documentation. Developers can also use API keys to add AI models from other providers. This allows developers to use AI to write code, fix errors, and more. Apple also launched on-device foundation models, requiring only three lines of code to access them.
Apple delays AI-powered Siri update
Apple didn't announce an AI-powered Siri update at WWDC 25. The update, first introduced at WWDC 24, is expected to make Siri more personalized. Apple's SVP of Software Engineering said the update needs more time to meet their quality standards. The new Siri was supposed to understand personal context and take action across apps.
Apple introduces AI-powered Shortcuts app at WWDC 2025
Apple introduced an AI-powered Shortcuts app at WWDC 2025. The new Shortcuts app uses Apple Intelligence and allows developers to access Apple’s on-device large language model (LLM). Users can create shortcuts that use AI to summarize text or create images. The app also includes intelligent actions enabled by Apple Intelligence.
Meta may invest billions in Scale AI
Meta is reportedly in talks to invest billions of dollars in Scale AI. The deal could be worth over $10 billion. This would be Meta's largest AI investment and one of the biggest funding events for a private company.
AI insights boost retail sales
AI can help retail sales representatives by simulating customer interactions. This allows them to create better sales pitches and increase sales. ETBrandEquity and Sharpsell.ai explored how AI is changing sales in the retail industry. AI helps sales representatives understand customer needs and improve their sales techniques.
Bishops say AI must serve humanity
U.S. bishops wrote a letter to Congress about artificial intelligence. They said AI should be developed to serve all of humanity. The bishops want to ensure AI development respects human dignity and the common good. The letter addresses AI's impact on family life, labor, healthcare, and more.
AI security drives private cloud growth
Private cloud growth is outpacing public cloud investment due to AI security and compliance concerns. Companies are moving AI workloads to private clouds for better security and to meet compliance rules. A survey found that private cloud spending will increase significantly this year. Companies want more control over their AI testing and deployment.
Broadcom focuses on AI chips
AI is driving Broadcom's semiconductor business, accounting for over half of its sales. The company's AI chip revenues are increasing, including switch ASICs and XPU ASIC design. Broadcom is working with hyperscalers and cloud builders to create custom compute engines. They expect AI revenue to continue growing in the coming years.
AI images used in Queens campaign
A Queens City Council candidate, Ruben Wills, appears to be using AI-generated images on his campaign website. Experts identified distorted faces and unnatural details in the photos. The images raise questions about AI's role in political campaigns. Wills' website was designed by C&P Creative, a company that focuses on AI in political advocacy.
Democrats risk losing AI leadership
Democrats may be losing support from the tech industry due to disagreements on AI regulation and energy policy. Tech leaders are increasingly supporting conservative candidates. Democrats' focus on antitrust and AI regulation is causing tension with tech companies. To avoid ceding AI leadership, Democrats need to collaborate with AI executives and support infrastructure development.
AI creates more jobs and higher pay, study says
A new study by PwC found that AI is creating more jobs and increasing productivity. Companies that have adopted AI are seeing higher job creation and increased pay. AI is also helping people without university degrees perform tasks previously reserved for knowledge workers. The study suggests that AI is enhancing workers' value and boosting their productivity.
Sources
- Apple says generative AI cannot think like a human - research paper pours cold water on reasoning models
- Apple’s new research paper says AI reasoning isn’t all it’s cracked up to be.
- AI models still far from AGI-level reasoning: Apple researchers
- Cutting-edge AI models ‘collapse’ in face of complex problems, Apple study finds
- Apple Researchers Just Released a Damning Paper That Pours Water on the Entire AI Industry
- Report: Reasoning AI Models Fail When Problems Get Too Complicated
- Breakthrough Apple study shows advanced reasoning AI doesn't actually reason at all
- Apple AI boffins pour cold water on reasoning models
- Apple brings ChatGPT and other AI models to Xcode
- Still no AI-powered, 'more personalized' Siri from Apple at WWDC 25
- At WWDC 2025, Apple introduces an AI-powered Shortcuts app
- Meta reportedly in talks to invest billions of dollars in Scale AI
- Enabling the last mile with AI generated insights
- U.S. Bishops Tell Congressional Leadership that AI Must Serve All of Humanity
- AI security and compliance concerns are driving a private cloud boom
- Broadcom At The Crossroads Between Merchant And Custom Silicon
- A Queens council candidate appears to be using AI campaign images – should it matter?
- Opinion | Democrats resist AI at their peril
- New Research Finds That AI Is Creating More Jobs and Higher Pay