M.B.A. Students vs. ChatGPT: Who Comes Up With More Innovative Ideas?

How good is AI in generating new ideas?

The conventional wisdom has been not very good. Identifying opportunities for new ventures, generating a solution for an unmet need, or naming a new company are unstructured tasks that seem ill-suited for algorithms. Yet recent advances in AI, and specifically the advent of large language models like ChatGPT, are challenging these assumptions.

We have taught innovation, entrepreneurship and product design for many years. For the first assignment in our innovation courses at the Wharton School, we ask students to generate a dozen or so ideas for a new product or service. As a result, we have heard several thousand new venture ideas pitched by undergraduate students, M.B.A. students and seasoned executives. Some of these ideas are awesome, some are awful, and, as you would expect, most are somewhere in the middle.

The library of ideas, though, allowed us to set up a simple competition to judge who is better at generating innovative ideas: the human or the machine.

In this competition, which we ran together with our colleagues Lennart Meincke and Karan Girotra, humanity was represented by a pool of 200 randomly selected ideas from our Wharton students. The machines were represented by ChatGPT4, which we instructed to generate 100 ideas with otherwise identical instructions as given to the students: “generate an idea for a new product or service appealing to college students that could be made available for $50 or less.”

In addition to this vanilla prompt, we also asked ChatGPT for another 100 ideas after providing a handful of examples of successful ideas from past courses (in other words, a trained GPT group), providing us with a total sample of 400 ideas.

Collapsible laundry hamper, dorm-room chef kit, ergonomic cushion for hard classroom seats, and hundreds more ideas miraculously spewed from a laptop.

How to compare

The academic literature on ideation postulates three dimensions of creative performance: the quantity of ideas, the average quality of ideas, and the number of truly exceptional ideas.

First, on the number of ideas per unit of time: Not surprisingly, ChatGPT easily outperforms us humans on that dimension. Generating 200 ideas the old-fashioned way requires days of human work, while ChatGPT can spit out 200 ideas with about an hour of supervision.

Next, to assess the quality of the ideas, we market tested them. Specifically, we took each of the 400 ideas and put them in front of a survey panel of customers in the target market via an online purchase-intent survey. The question we asked was: “How likely would you be to purchase based on this concept if it were available to you?” The possible responses ranged from definitely wouldn’t purchase to definitely would purchase.

The responses can be translated into a purchase probability using simple market-research techniques. The average purchase probability of a human-generated idea was 40%, that of vanilla GPT-4 was 47%, and that of GPT-4 seeded with good ideas was 49%. In short, ChatGPT isn’t only faster but also on average better at idea generation.

Still, when you’re looking for great ideas, averages can be misleading. In innovation, it’s the exceptional ideas that matter: Most managers would prefer one idea that is brilliant and nine ideas that are flops over 10 decent ideas, even if the average quality of the latter option might be higher. To capture this perspective, we investigated only the subset of the best ideas in our pool—specifically the top 10%. Of these 40 ideas, five were generated by students and 35 were created by ChatGPT (15 from the vanilla ChatGPT set and 20 from the pre trained ChatGPT set). Once again, ChatGPT came out on top.

What it means

We believe that the 35-to-5 victory of the machine in generating exceptional ideas (not to mention the dramatically lower production costs) has substantial implications for how we think about creativity and innovation.

First, generative AI has brought a new source of ideas to the world. Not using this source would be a sin. It doesn’t matter if you are working on a pitch for your local business-plan competition or if you are seeking a cure for cancer—every innovator should develop the habit of complementing his or her own ideas with the ones created by technology. Ideation will always have an element of randomness to it, and so we cannot guarantee that your idea will get an A+, but there is no excuse left if you get a C.

Second, the bottleneck for the early phases of the innovation process in organisations now shifts from generating ideas to evaluating ideas. Using a large language model, an innovator can produce a spreadsheet articulating hundreds of ideas, which likely include a few blockbusters. This abundance then demands an effective selection mechanism to find the needles in the haystack.

To date, these models appear to perform no better than any single expert in their ability to predict commercial viability. Using a sample of a dozen or so independent evaluations from potential customers in the target market—a wisdom of crowds approach—remains the best strategy. Fortunately, screening ideas using a purchase intent survey of customers in the target market is relatively fast and cheap.

Finally, rather than thinking about a competition between humans and machines, we should find a way in which the two work together. This approach in which AI takes on the role of a co-pilot has already emerged in software development. For example, our human (pilot) innovator might identify an open problem. The AI (co-pilot) might then report what is known about the problem, followed by an effort in which the human and AI independently explore possible solutions, virtually guaranteeing a thorough consideration of opportunities.

The human decision maker is likely ultimately responsible for the outcome, and so will likely make the screening and selection decisions, informed by customer research and possibly by the opinion of the AI co-pilot. We predict such a human-machine collaboration will deliver better products and services to the market, and improved solutions for whatever society needs in the future.

Christian Terwiesch and Karl Ulrich are professors of operations, information and decisions at the Wharton School of the University of Pennsylvania, where Terwiesch also co-directs the Mack Institute for Innovation Management.

A sharp technology sell-off has exposed a new anxiety on global markets: whether the enormous borrowing behind the artificial-intelligence infrastructure boom can generate sufficient returns as interest rates rise and Chinese competition intensifies.

The artificial-intelligence investment boom has powered technology shares, semiconductor manufacturers and data-centre operators to extraordinary heights. Last week, investors were given a reminder that much of this expansion is being built with borrowed money.

The technology-heavy Nasdaq Composite fell 4.5 per cent over the week ended October 16, its weakest performance since late June. Semiconductor shares suffered an even deeper correction, with the Philadelphia Semiconductor Index losing about 11.4 per cent.

The selling spread quickly through global markets. Micron Technology fell 14.6 per cent over the week, while South Korea’s Samsung Electronics and SK Hynix dropped approximately 18 per cent and 24 per cent respectively.

This was more than a routine bout of profit-taking. Investors were reassessing three interconnected risks: higher interest rates, the rapidly expanding debt burden attached to AI infrastructure and the prospect of stronger competition from Chinese memory-chip manufacturers.

The debt behind the AI build-out

Artificial intelligence requires a physical infrastructure network of enormous scale. Data centres must be built, advanced processors purchased, power supplies secured and cooling systems installed before customers begin generating meaningful revenue from them.

Technology companies are increasingly supplementing their own cash with corporate bonds, syndicated loans, equipment leases, private credit and project-level financing.

JPMorgan describes the emerging funding model as a combination of conventional corporate borrowing and project finance tied to individual data-centre assets. In Australia, the Reserve Bank estimates that syndicated loans accounted for approximately three-quarters of data-centre financing during 2025 and 2026 to date.

The numbers are becoming large enough to affect broader credit markets. The Federal Reserve Bank of Dallas has cited Wall Street estimates suggesting AI-related investment-grade bond issuance could approach US$300 billion in 2026 alone.

That borrowing is not automatically problematic. Debt can be an efficient way to finance long-lived infrastructure when future revenue is sufficiently predictable. The concern is that AI technology, customer demand and computing economics are changing much faster than the repayment schedules attached to the assets.

A data centre financed today may carry debt for years. Investors must therefore make assumptions about future utilisation rates, electricity prices, chip performance, customer concentration and the amount businesses will ultimately pay for AI computing.

When interest rates rise, those assumptions become less forgiving.

Inflation changes the calculation

September’s US inflation data intensified the pressure. Consumer prices increased 3.7 per cent over the year, ahead of market expectations and up from 3.4 per cent in August.

The hotter reading reduced confidence that interest rates would fall quickly and pushed bond yields higher. That matters particularly for companies whose valuations depend on profits expected well into the future, as well as highly leveraged infrastructure operators that may need to refinance.

CoreWeave has emerged as a closely watched test case. The specialist cloud-computing company has accumulated substantial debt and lease obligations to build AI capacity, while relying on a concentrated group of large technology customers.

Its shares fell sharply during the sell-off, while the cost of insuring its debt against default climbed. CoreWeave has argued that its contracted backlog more than covers its obligations and that financial markets do not always understand its business model. Nevertheless, its securities have become a proxy for investor confidence in debt-funded AI infrastructure.

Oracle, another major investor in AI computing capacity, fell 8.8 per cent in Friday trading, while CoreWeave dropped 15.5 per cent.

The question is shifting from how much AI infrastructure can be built to how profitably it can be operated.

China adds another source of pressure

Interest rates were not the only catalyst. Reports of advances by Chinese memory-chip producer ChangXin Memory Technologies raised concerns that lower-cost Chinese products could increase competition across the global semiconductor industry.

Memory chips are essential to AI systems, but the industry is historically cyclical. Periods of tight supply and high pricing can encourage rapid investment, eventually producing oversupply and falling margins.

The prospect of a well-funded Chinese competitor expanding its technological capabilities unsettled investors in Samsung, SK Hynix and Micron. It also demonstrated that AI-related demand does not insulate chipmakers from conventional competitive pressures.

That combination—slower rate cuts, more expensive financing and potentially cheaper supply—challenged the assumption that every company exposed to AI would enjoy continually expanding margins.

A reset rather than the end of the boom

The sell-off does not establish that the AI boom is ending. Demand for computing power remains strong, the largest technology companies continue to generate formidable cash flows and many infrastructure projects are backed by long-term customer contracts.

What has changed is the market’s willingness to overlook the price and financing structure of that growth.

The most resilient companies are likely to be those able to fund investment from operating cash flow, secure long-duration customers and demonstrate that new capacity will earn returns comfortably above its financing costs.

More speculative operators may face greater scrutiny, particularly where debt is expensive, customers are concentrated or expected revenue depends on demand that has not yet materialised.

For Australian investors, the lesson extends beyond US technology shares. Local data-centre operators, infrastructure funds, electricity suppliers and property groups are increasingly exposed to the same investment cycle.

Artificial intelligence may remain one of the defining growth themes of the decade. But last week’s correction showed that enthusiasm for the technology does not eliminate financial discipline.

The next stage of the AI boom will be judged not only by computing power, model performance or demand for chips, but by something more traditional: whether the assets being financed can generate enough cash to repay the money borrowed to build them.

Fact box

Nasdaq Composite: Down 4.5 per cent for the week ended October 16, 2026

Philadelphia Semiconductor Index: Down approximately 11.4 per cent

Micron Technology: Down approximately 14.6 per cent

Samsung Electronics: Down approximately 18 per cent

SK Hynix: Down approximately 24 per cent

US September inflation: 3.7 per cent annually

Estimated 2026 AI-related investment-grade issuance: Approximately US$300 billion