What AI Still Can't Do — And Why It Matters More Than Ever
AI will get smarter. That's inevitable. But it still can't do what makes us human, and that's what ... More matters most.
When Sal Khan introduced GPT-4 into Khan Academy's tutoring platform in 2023, it rocked the education world. For the first time, a generative AI model could walk students through complex AP problems, provide Socratic-style coaching and give instant feedback across subjects. It was fast, accessible and, in many ways, effective.
But as The New York Times reported, there were mixed reactions to Khan Academy's AI tutor, Khanmigo. Educators voiced concern that the bot sometimes does 'too much of the thinking work,' potentially hindering students' critical thinking. Others said the AI would be more effective if it could identify misconceptions and respond with open-ended questions — much like a human teacher would.
A study from The Wharton School, summarized by Axios, found that students who used AI tools during practice 'performed significantly worse on exams, where they can't rely on artificial intelligence.' The National Council of Teachers of Mathematics noted that while these AI tools are helpful, 'students will need teachers to help them create a bridge between prior knowledge, new knowledge and shared knowledge.'
It's a strong reminder that even in a world where AI can correct grammar in milliseconds and potentially transform global education by breaking language barriers and expanding access to underserved communities, according to UNESCO, there are still many things it can't do.
According to experts like Selva Pankaj, joint CEO of the Regent Group, as our relationship with AI deepens, the gap between what AI can do and what only humans should do is where our focus should be. One of the most critical steps to take, Pankaj told me in an interview, is modernizing our education systems to prepare students for a world where algorithms and natural intelligence work together in harmony.
Rethinking Skills In An AI-Driven World
AI tools are no longer just nice-to-haves. They now function as intelligent assistants — managing tasks, scheduling meetings, drafting emails, generating artwork and parsing vast datasets — often with a single prompt. As Ron Cameron noted in a previous article on Forbes, the average AI system has reached the stage where it functions as an 'intelligent assistant' that can handle repetitive tasks without any adjustment to the level of human supervision.
That trend isn't slowing down and it's already reshaping the job market in many ways. According to The World Economic Forum's 2023 Future of Jobs Report, the demand for roles such as data entry clerks and bank tellers will decline quickly. While that doesn't necessarily mean that hard skills like graphic design or data analysis are disappearing, it certainly means they're no longer sufficient on their own.
Selva Pankaj, co-CEO of Regent Group
Pankaj believes the global curriculum must evolve to reflect AI's growing impact on us. 'Education has largely remained unchanged for the last 500 years. We are at the cusp of a fourth educational revolution where we can leverage AI to improve education for all.'
When I asked how that looks in practice, he explained how AI is already helping educators: 'An AI system can quickly analyze test results and provide insights into where students are struggling and why,' he said. 'This saves time and helps teachers take targeted action.'
One article by the WEF argues that the future of education lies in hybrid intelligence, where AI and humans complement one another. 'By combining the analytical power of AI with the irreplaceable human element of teaching, we can truly transform education for the next generation,' the report said.
Preparing For A Post-AI Classroom
Many universities have introduced AI-related courses, but the core of most curriculums still rewards memorization and standardized testing — skills that machines now perform better and faster. According to Pankaj, 'that needs to change very quickly.'
To stay relevant, educators must move students beyond passive learning toward empathy, curiosity, adaptability and the kind of moral reasoning no algorithm can replicate. These are the skills that future employers and society will value most.
In an article published on Harvard Business Impact, Cherie Werhun, associate vice provost of human development and potential at Sheridan College, emphasized the need to nurture what she calls 'self-intelligence.' 'If we don't devote attention to students' self-intelligence skills,' she wrote, 'we are equipping them with only half of the education they will need to not only survive, but thrive.'
Neglecting those capabilities doesn't just shortchange students; it could lead to a wider unemployment crisis. Today's jobs increasingly reward human strengths like emotional intelligence and problem-solving, not rote knowledge.
At the 2025 World Governments Summit, Pankaj called for a new definition of education itself. 'We need to stop boxing students into a one-size-fits-all approach,' he said. 'Learning must become personalized, and pedagogy should evolve with real-world outcomes in mind.'
That means encouraging students to evaluate, reflect and creatively solve problems, not just absorb information.
A Future Built On Intelligence Augmentation
As many experts across the industry have already noted, the future isn't man versus machine. Rather, it's going to be man and machine working together. But the key to seeing that relationship flourish is knowing exactly how to do that.
Humans must stop viewing AI as a threat and start seeing it as a co-pilot that can enhance how we think, learn and create. Because despite AI's speed and memory, even the most advanced models still lack deeply human traits like emotional depth, creative intuition and moral judgment. Across education institutions, coursework should evolve to reflect that reality, pushing students to analyze complex issues and respond with innovative solutions.
The real opportunity isn't to outpace machines, but to understand where human intelligence leads and to shape a future that builds on those strengths. As Pankaj put it, 'some jobs like data entry may fade, but new roles like business intelligence specialists will emerge. And only those who leverage AI to upskill themselves will be in contention.'
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Gizmodo
9 minutes ago
- Gizmodo
RV Camping This Summer? This Power Station With Solar Panels Is at a Record Low Price
If you're heading out for camping this summer, having a reliable power station is already a smart move. But pairing your power station with a solar panel for true off-grid independence is even better—and right now, the Jackery Solar Generator 300 with a 100W panel is available at a record low price on Amazon, down by 40%. At just $298, this bundle is a steal compared to its original $499 price tag, which makes it one of the best (early Prime Day) deals you'll find for a portable solar power setup. This setup isn't just for summer adventures, either: When storms are closing in and winter looms with a threat of power loss, this power station will keep your essentials charged. See at Amazon At just 7.1 pounds, it's light enough to carry in a backpack but contains a whopping 293Wh lithium-ion battery that can supply up to 300W of clean energy. The device features two pure AC outlets so you can safely power sensitive electronics without risking voltage spikes and interference. If you desire a power station with flexibility, this model is able to charge a maximum of six devices simultaneously. The output diversity is incredible: two AC outlets, a 60W PD USB-C port (input/output compatible), a fast charge 3.0 USB-A port, and a DC car port. Recharging the Jackery Solar Generator 300 is simple: With the wall socket and the 60W PD USB-C port combined, you can top up 80% of the battery in two hours. If you're not stationary, you can use the included car charger cable or simply plug in the 100W panel for clean, green energy. The solar panel will connect with the power station seamlessly, thanks to the onboard MPPT controller. This deal is one of the best you'll find on a trusted brand like Jackery, and with the current discount, it's hard to imagine a better time to invest in this. See at Amazon
Forbes
15 minutes ago
- Forbes
AI In E-Commerce: Are Brands And Customers Aligned?
Rytis Lauris is the cofounder and CEO of Omnisend, a marketing automation platform built for e-commerce. I don't think there is a single industry AI hasn't touched, and e-commerce is certainly no exception. Chatbots, product recommendations, segment creation and churn prediction have been well underway for some time, but these tools are only the beginning. It seems like every day we read stories of another brand implementing new and improved AI on their site. However, while brands see the value and promise in using AI to revolutionize the customer experience, do consumers want it, and if so, for what purposes? What Online Shoppers Want From AI While times will surely change, currently, many consumers want AI to improve their shopping experiences, not transform them. A consumer survey from my company learned that the AI-powered features shoppers found most useful were improved product recommendations and faster product discovery. Nearly half also said they want to see improvements in customer service. However, product recommendations, whether on-site or in marketing emails, aren't always relevant, much to the dismay of shoppers. This is an opportunity for brands to use AI to give shoppers what they want. The more relevant the recommendations, the easier it is to find a product, leading to a better chance of capturing a sale. For example, take Muse by Wayfair, where users can type a description of what they're looking for (e.g., "modern chic dining room with hints of lime") and find design inspiration. Upon clicking on the 'muse' they like, shoppers can click on the Wayfair products pictured in the image. The challenge for brands with customer service is that many chatbots, though improving, are clunky and often frustrating. This needs to change. For more complex inquiries, there needs to be a seamless escalation procedure that directs to human representatives without losing the chat history or background from the customer. If chatbots or other AI customer service tools can't provide a more intuitive level of support, it creates friction and becomes a detriment to the business. This can't happen. Agentic AI, where autonomous decisions are made, is a rapidly advancing area for e-commerce brands. Maybe the most newsworthy example is Amazon's Buy For Me feature, which allows users to discover products through the Amazon site that are not sold on Amazon and then complete the purchase directly with the off-site retailer on their behalf. Amazon isn't alone in using agentic AI to drive purchases. Visa and Mastercard got into the fray with tools that can make purchases on people's behalf. Admittedly, we are at the early stages of this type of use case for agentic AI, and it's going to be interesting to see where it leads. But right now, most consumers aren't quite ready. In my view, the current behavior for most shoppers is to verify all details, from size and color to shipping information, before clicking 'complete purchase.' Trusting AI to select the right size, color and other variables feels like a large behavioral and trust issue that needs to be overcome. Agentic AI holds a lot of promise for e-commerce brands on the customer-facing and operational sides of the business. It can be used to make product recommendations, sift through product reviews, solve customer service inquiries and track and find lower prices. Operationally, it can be used to track and forecast inventory, analyze data, personalize marketing campaigns and detect fraud. What Brands Can Do Now Consumers' hesitation to use AI for more than basic improvements to their shopping experience isn't irrational. The online purchase process is mostly the same as it was 20 years ago: search, click, browse, add to cart and check out. Expecting consumers to abandon this instinctive behavior overnight is an impossible ask, especially when it comes to making purchases for them. Trust in AI needs to be established and built through use and consistency. But like ecommerce's early days, it's going to be an evolution—one step at a time. If AI applications are consistently useful for shoppers, they'll trust and adopt them. If they're not, they won't. Brands can look to AI to help in various ways, but what they can't lose sight of is that, within this, shopping is often enjoyable for people. Leaving this to programmatic tools removes the thrill of the hunt to find the perfect item. Striking this balance will be a challenge. As such, e-commerce brands should adopt AI thoughtfully and use it to put consumers first. While experimenting with new tools is fine, the goal should be to give shoppers better experiences, whether it's support, product discovery or personalization. If brands can't execute the fundamentals and instead create friction, asking them to change their instinctive shopping behaviors will be a pipe dream. The two 'simplest' ways brands can begin building consumers' trust in AI is through chatbots and product search. Chatbots have come a long way in the past several years, moving from a tool that provides help article links to one that can accurately provide answers and triage customer concerns. By refining customer service chatbots to handle customer queries accurately, consumers can realize their value and trust them further. This trust in one tool gives them a reason to trust the next one. On-site search is another area where brands can use AI to support the customer experience. As search moves from keyword matching to an intuitive prediction model, shoppers' ability to discover additional products and have an expanded, non-linear shopping experience improves. A non-linear shopping experience creates a more enjoyable one, similar to the surprise-and-delight experience of in-store shopping. If AI-powered search results stay relevant and helpful, shoppers are more likely to see AI as a helpful asset vs. an annoyance. The Takeaway Is AI in e-commerce going anywhere? Absolutely not. Will it transform the way online shopping is done? Probably. Are consumers clamoring for radical changes instead of everyday improvements? Not yet. It's OK to dream big, and I'm excited for what's to come, but don't lose sight of what got you here: your customers. Forbes Business Council is the foremost growth and networking organization for business owners and leaders. Do I qualify?
Forbes
15 minutes ago
- Forbes
When AI, Energy Demands And Capital Costs Collide
Charles Yeomans is the chairman and founder of Atombeam. In a testimony to Congress in April 2025, Google's former CEO Eric Schmidt addressed not only the future of AI but also the dramatic power consumption associated with it. Although the dramatic increase in energy used by data centers associated with AI workloads is well known—and my last article touched on this crucially important issue—the sheer scale of the challenge is becoming clearer. Uncharted Territory More specifically, Schmidt noted that data center energy consumption could quickly increase from 3% of the power generated today to 99% of that current value, with data centers requiring 67 more gigawatts of power by 2030. He also subsequently put that point in perspective by pointing out that the average nuclear power plant creates about 1 gigawatt of power, even as technology providers plan for 10-gigawatt data centers. Although those figures are subject to some debate—the U.S. Department of Energy, for example, noted that data centers consumed about '4.4% of total electricity in 2023 and is expected to consume from 6.7 to 12% of total U.S. electricity by 2028'—one thing is certain: The proliferation of AI and the more powerful computational capabilities fast-evolving use cases, algorithms and models demand brings with it a power generation challenge of unprecedented proportions. Not surprisingly, to date, nuclear power is seen as the most viable, although still arguably insufficient, answer—a reality made clear by the nuclear ambitions of the largest technology providers. In early May 2025, Elementl Power announced its agreement with Google to build 'three sites for advanced reactors,' as reported by CNBC. Notably, though, Elementl Power, founded in 2022, has yet to build any nuclear sites. We're in uncharted territory. Coming To Terms Clearly, questions abound not just about national security—something Schmidt also stressed—and the feasibility not just of how data centers will be powered but also how quickly they can be built to keep up with AI. This is a point I also made in my last article while pointing to October 2024 research from McKinsey analysts, who found that 'to avoid a deficit, at least twice the data center capacity built since 2000 would have to be built in less than a quarter of the time.' One point, though, has yet to get the attention it fully deserves as we collectively come to terms with the dramatic demands associated with AI. That point is the sheer scale of the capital needed to facilitate the building boom required to create the infrastructure AI demands. But that, too—like the scale of the power consumption Schmidt brought into focus—is beginning to become more clear as April 2024 research, also from McKinsey analysts, reveals. That study predicts that by '2030, data centers are projected to require $6.7 trillion worldwide to keep pace with the demand for compute power.' Consider for a moment the awe-inspiring reality of that number. In 2025, total federal outlays in the U.S. are predicted to reach $7.0 trillion, or 23.3% of the GDP. Viewed differently, that same amount represents nearly 38% of China's GDP of $17.79 trillion, according to the World Bank. And we're talking just about the creation of new data centers. Cost Considerations Of course, data centers are obviously but one piece of the networks needed to use, move and store data associated with AI-powered use cases and workloads. The power consumption and capital costs associated with the infrastructure AI demands also apply to the pipes that connect everything, from the wired infrastructure to the cellular and satellite networks that pull it all together and enable AI outputs to be used in real-world applications. All of those critical components of AI infrastructure require massive capital investments and expenditures. Subsea cable, for example, costs $40,000 per mile, or between $200 million and $250 million for a transatlantic connection. Despite the increased economy, launching satellites remains a costly endeavor, with outlays being highly variable and impacted by numerous factors, from the orbit itself—geostationary Earth orbit costs more than low Earth orbit—to the weight of the satellite, the launch vehicle involved and, of course, the cost of making the satellite itself. Questions also abound about how costs will evolve and how quickly in light of significant advancements in space delivery technologies. Cellular networks are no less complex and also require significant capital investments. For example, in just one deal with U.S. Cellular in 2024, Verizon confirmed earmarking $1 billion for the purchase of additional spectrum alone. The Legacy Problem And then there's the issue of legacy communications and industrial networks that remain in use across many industries. These systems will need to be upgraded to handle large AI workloads if commercial applications of AI are to progress as envisioned. This includes the supervisory control and data acquisition systems (SCADA) that remain prevalent in the utility sector and across industries like manufacturing, with many now two and three decades old. But the issue of legacy systems isn't confined to any one sector. Link 16, the U.S. military's most-used digital datalink, is now in its fourth decade of use. Focusing On The Data Clearly, given these factors, the costs of the modernization and expansion efforts AI demands aren't feasible as viewed currently, particularly in a time when public-sector and private-sector investments are contracting, not growing. Nor is it realistic to believe we collectively have the people and businesses needed to keep pace with AI's rapid expansion and building demands. That's why we must, now more than ever, be pragmatic not only about our response to the investment AI calls for but also the kinds of AI data we use and how we manage, store, share and use that information. It's time to focus on data itself. Does it need to be this big? Can we make it more efficient and reduce the strain on resources? Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
