AI's reasoning problems -- why 'thinking' models may not actually be smarter
The latest releases from the major players in artificial intelligence, including OpenAI, Anthropic, Alphabet and DeepSeek, have been models with reasoning capabilities. Those reasoning models can execute on tougher tasks by "thinking," or breaking problems into logical steps and showing their work.
Now, a string of recent research is calling that into question.
In June, a team of Apple researchers released a white paper titled "The Illusion of Thinking," which found that "state-of-the-art [large reasoning models] still fail to develop generalizable problem-solving capabilities, with accuracy ultimately collapsing to zero beyond certain complexities across different environments."
In other words, once problems get complex enough, reasoning models stop working. Even more concerning, the models aren't "generalizable," meaning they might be just memorizing patterns instead of coming up with genuinely new solutions.
"We can make it do really well on benchmarks. We can make it do really well on specific tasks," said Ali Ghodsi, the CEO of AI data analytics platform Databricks. "Some of the papers you alluded to show it doesn't generalize. So while it's really good at this task, it's awful at very common sense things that you and I would do in our sleep. And that's, I think, a fundamental limitation of reasoning models right now."
Researchers at Salesforce, Anthropic and other AI labs have also raised red flags about reasoning models. Salesforce calls it "jagged intelligence" and finds that there's "significant gap between current [large language models] capabilities and real-world enterprise demand."
The constraints could indicate cracks in a story that has sent AI infrastructure stocks like Nvidia booming.
"The amount of computation we need at this point as a result of agentic AI, as a result of reasoning, is easily a hundred times more than we thought we needed this time last year," Nvidia CEO Jensen Huang said at the company's GTC event in March.
To be sure, some experts say Apple's warnings about reasoning models may be the iPhone maker shifting the conversation because it is seen as playing catch up in the AI race. The company has had a series of setbacks with its highly-touted Apple Intelligence suite of AI services.
Most notably, Apple had to delay key upgrades to its Siri voice assistant to sometime in 2026, and the company did not make many announcements regarding AI at its annual Worldwide Developers Conference earlier this month.
"Apple's putting out papers right now saying LLMs and reasoning don't really work," said Daniel Newman, Futurum Group CEO on CNBC's "The Exchange." Having Apple's paper come out after WWDC "sounds more like 'Oops, look over here, we don't know exactly what we're doing.'"
Watch this video to learn more.
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Yahoo
43 minutes ago
- Yahoo
Wedbush Reaffirms ‘Outperform' Rating on Apple Inc. (AAPL) with $270 PT
Apple Inc. (NASDAQ:AAPL) is included in our list of the 14 Best Warren Buffett Stocks to Invest in. 10 Smartphones with the Best Cameras and Battery Life On August 7, 2025, Apple Inc. (NASDAQ:AAPL) announced an additional $100 billion investment, raising its total domestic commitment to $600 billion over the next four years. This move follows the company's February pledge of $500 billion and 20,000 new jobs. With this announcement, Apple extends its commitment to support its American Manufacturing Program, expanding U.S. manufacturing and supply chain capacity amid continued tariff tensions. Furthermore, Apple Inc. (NASDAQ:AAPL) announced the expansion of its partnership with Corning. With this partnership, the company intends to increase smartphone glass production at Corning's Kentucky plant, increasing domestic component sourcing. Following these critical developments, Wedbush reaffirmed its 'Outperform' rating on Apple Inc. (NASDAQ:AAPL) with a $270 price target. Apple Inc. (NASDAQ:AAPL) manufactures smartphones, personal computers, tablets, wearables, and accessories. It also sells related software and services globally. It is included in our list of the best Warren Buffett stocks. While we acknowledge the potential of AAPL as an investment, we believe certain AI stocks offer greater upside potential and carry less downside risk. If you're looking for an extremely undervalued AI stock that also stands to benefit significantly from Trump-era tariffs and the onshoring trend, see our free report on the best short-term AI stock. READ NEXT: 13 Best Oil Refinery Stocks to Buy Right Now and 7 Best Potash Stocks to Buy According to Analysts. Disclosure: None. Sign in to access your portfolio
Yahoo
an hour ago
- Yahoo
AI Development Is Accelerating: 1 Vanguard ETF to Buy Right Now
Key Points The fund provides exposure to the AI hardware and software supply chain while maintaining an ultra-low 0.09% expense ratio. Top holdings, including Nvidia, Microsoft, and Apple, represent companies dominating different segments of the AI revolution. The fund has nearly doubled the S&P 500's returns over the past 15 years. 10 stocks we like better than Vanguard Information Technology ETF › Everyone's hunting for the next Nvidia (NASDAQ: NVDA), but they're looking in the wrong place. The real artificial intelligence (AI) opportunity lies in owning the companies supplying the chips, cloud platforms, and enterprise software powering today's AI gold rush. With $300 billion in infrastructure spending from tech giants this year alone, modern prospectors are investing in computing and networking. The companies selling that infrastructure are the ones striking gold. For investors who prefer to avoid single-stock risk, the Vanguard Information Technology ETF (NYSEMKT: VGT) provides exposure to the AI technology supply chain. Read on to find out more about this top AI vehicle. The accidental AI fund This exchange-traded fund (ETF) wasn't designed as an AI play when it was launched in 2004, years before deep learning went mainstream. But through market-cap weighting and technology's winner-take-all dynamics, it has become a strong proxy for the AI build-out. Nvidia commands 16.7% of the fund, reflecting its stranglehold on the GPU market that powers AI training. Microsoft (NASDAQ: MSFT) represents 14.9%, with its Azure cloud platform hosting everything from OpenAI's ChatGPT to thousands of enterprise AI applications. Apple rounds out the top three at 13%, preparing to push AI features to over 2.35 billion active devices. The concentration gets more interesting further down. Broadcom, at 4.6%, designs custom AI chips for hyperscalers and supplies the networking gear that keeps data flowing between thousands of processors. Oracle, at 2%, has reinvented itself as an AI cloud provider, winning customers with specialized infrastructure for training large models. These aren't speculative bets. Microsoft's cloud revenue hit $40.9 billion last quarter, up 21% year over year. Nvidia's data center revenue reached $39.1 billion in its most recent quarter, up 73% relative to the same period a year ago. The AI gold rush is generating real revenue, and this fund captures the companies mining the most gold. Beyond the mega-caps and long-term performance The ETF's reach extends beyond its mega-cap holdings, with 319 stocks providing AI exposure across the technology stack. Palantir Technologies (1.6%) delivers AI solutions to government agencies and Fortune 500 clients. Cisco Systems (1.6%) supplies the networking hardware that links together vast AI computing clusters. IBM (1.6%) deploys its Watson AI platform to help enterprises integrate advanced analytics and automation into their operations. Over the past 15 years, the fund has delivered 19.7% annual total returns (including distributions), turning $10,000 into roughly $155,000. That's about 9 percentage points per year ahead of the benchmark S&P 500 over the same period. That's the power of owning the leaders in the most important technological innovation of our time. Upside potential, fees, and risk profile Analysts remain optimistic about the ETF's largest holdings, thanks to AI spending. Morgan Stanley estimates that Microsoft, Amazon, Alphabet, and Meta Platforms will invest about $300 billion in AI infrastructure in 2025. That level of investment should directly benefit the ETF's hardware and semiconductor holdings. On the fee front, investors get this exposure for just $9 per year on a $10,000 investment -- a fraction of what most actively managed tech funds charge. The ETF has some blind spots. As a fund in the information technology sector, it excludes Amazon and Alphabet entirely, leaving out Amazon Web Services and Google Cloud despite their dominant role in AI. It is also heavily weighted toward Apple, Microsoft, and Nvidia, so weakness in any of these companies could weigh on performance. Moreover, technology-sector drawdowns can be severe. The fund declined more than 50% in 2008 and 30% in 2022. Even so, decades of compounding have more than offset the occasional sharp downturn. The innovation premium The Vanguard Information Technology ETF doesn't need to guess the next AI breakthrough. It owns the companies with the capital, talent, and market dominance to adapt to whatever comes next. Periods of volatility are inevitable, but history shows that technology leaders often emerge from downturns even stronger. For investors willing to ride out the drawdowns, the reward has been decades of market-beating compounding. So, with AI adoption accelerating and the sector's fundamentals in their best shape in years, this Vanguard ETF offers a low-cost way to capture the gains from this platform shift. Do the experts think Vanguard Information Technology ETF is a buy right now? The Motley Fool's expert analyst team, drawing on years of investing experience and deep analysis of thousands of stocks, leverages our proprietary Moneyball AI investing database to uncover top opportunities. They've just revealed their to buy now — did Vanguard Information Technology ETF make the list? When our Stock Advisor analyst team has a stock recommendation, it can pay to listen. After all, Stock Advisor's total average return is up 1,069% vs. just 184% for the S&P — that is beating the market by 884.49%!* Imagine if you were a Stock Advisor member when Netflix made this list on December 17, 2004... if you invested $1,000 at the time of our recommendation, you'd have $660,783!* Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you'd have $1,122,682!* The 10 stocks that made the cut could produce monster returns in the coming years. Don't miss out on the latest top 10 list, available when you join Stock Advisor. See the 10 stocks » *Stock Advisor returns as of August 13, 2025 George Budwell has positions in Apple, Microsoft, Nvidia, Palantir Technologies, and Vanguard Information Technology ETF. The Motley Fool has positions in and recommends Alphabet, Amazon, Apple, Cisco Systems, International Business Machines, Meta Platforms, Microsoft, Nvidia, Oracle, and Palantir Technologies. The Motley Fool recommends Broadcom and recommends the following options: long January 2026 $395 calls on Microsoft and short January 2026 $405 calls on Microsoft. The Motley Fool has a disclosure policy. AI Development Is Accelerating: 1 Vanguard ETF to Buy Right Now was originally published by The Motley Fool Error in retrieving data Sign in to access your portfolio Error in retrieving data Error in retrieving data Error in retrieving data Error in retrieving data
USA Today
an hour ago
- USA Today
Stop leaving your friends on read by marking the text as unread: Video
We've all been there. Someone sends us a text while we're in the middle of something else. We glance at it and tell ourselves we'll respond later. The next thing we know, it's been days and we never replied. Oops! A visual cue is one of the best ways to remind us when we unlock our devices. So the next time you need to remember to reply to someone, take a couple of seconds and mark the text as unread. Or type a text now and schedule it to send later! Watch this video to see how to mark a text as unread. How to mark a text message as unread There are two ways to mark a text unread on your iPhone: The red notification bubble will appear on the Messages app but will not appear as a notification on your lock screen. If the person you're texting has read receipts turned on, it will still show that you've read the message, even if you mark it as unread on your phone. These steps also work if you're trying to mark a message as unread in WhatsApp. If you have an Android, you can turn on reply reminders. The feature will prompt you to respond to the messages you haven't replied to or show you outgoing messages you haven't followed up on. The feature is part of the Nudges setting and is turned on by default. To turn it off or check your setting: The nudge will appear in the message thread and your inbox. Shop top-rated related products: How to pin a text message thread You can also pin messages to the top of your inbox as a reminder to respond or to quickly access the people you text most often. Just hard-press on the conversation and tap Pin. Those people will then appear as larger circles at the top of your messages inbox. Our team of savvy editors independently handpicks all recommendations. If you purchase through our links, the USA Today Network may earn a commission. Prices were accurate at the time of publication but may change.
