As AI continues to gain widespread adoption, industry experts predict the hype is about to spiral … [+]
AFP via Getty Images
Artificial intelligence has been the darling of the technology world for over two years, attracting several billions of dollars in capital. Yet, as 2025 unfolds, some experts believe this is the year where the “AI bubble” may burst.
It’s not the first time experts have made such predictions, urging caution as businesses jump on the AI train and companies jostle to carve their AI empires. As I wrote in a previous article on Forbes, Baidu CEO Robin Li said that “people would be disappointed that the technology doesn’t meet the high expectations generated through the initial excitement” and “only about 1% of companies will endure when the AI hype fades.”
This time, though, the prediction is a cocktail of good and bad. Experts from Oxylabs’ AI/ML Advisory Board — including Adi Andrei, Ali Chaudhry, and Oxylabs CEO Julius Černiauskas — predict waning enthusiasm and increased scrutiny, but also say there will be exciting advancements for AI in 2025.
The Law Of Diminishing (AI) Returns
In spite of the hype surrounding generative AI and large language models, the trio of Andrei, Chaudhry and Černiauskas caution that the technology may not deliver on its promises at the scale many earlier hoped. Chaudhry, a research fellow at University College London and founder of ResearchPal, believes AI is on course to face diminishing returns.
“Scaling laws are showing their limits. Some AI labs are already hinting that expanding model size and training data won’t yield the breakthroughs we’ve seen in the past,” he said in a press release.
Chaudhry also predicted tighter regulations aimed at controlling AI’s societal impact, particularly regarding gen AI. “2025 will be pivotal for AI safety,” he said, adding that “we’ll see both technical and policy efforts to mitigate its risks.”
Meanwhile, Černiauskas noted that the conversation is shifting toward responsible and sustainable AI practices. “Servers powering AI development strain the environment, and there’s growing concern about transparency in how AI operates,” he said. “AI companies must address these challenges to maintain public trust.”
Andrei, AI/ML expert and cofounder of Technosophics, was even much bolder in his prediction, noting that the Gen AI bubble is right on the verge of bursting. “The influx of money pumped into Gen AI without clear ROI has inflated expectations to unsustainable levels,” Andrei explained.
He cited American billionaire Tom Siebel, the founder and CEO of C3.ai, who has been quoted saying “the market is overvaluing AI” and that “there’s absolutely a bubble,” as an example of the sentiment by many CEOs and experts across Silicon Valley. Andrei also noted the growing resistance to gen AI among professionals and the public alike, which could further deflate the hype.
“There is a rising movement of ordinary people from diverse professions, such as writers, artists, computer scientists, engineers, and philosophers, who found common ground against the gen AI paradigm. This has raised awareness within the general population of the irreconcilable issues posed by technology and the fact that it is being forced onto people by billionaires and their organizations,” he concluded.
More AI Regulations
Although Oxylabs’ predictions come with dark clouds over AI, companies haven’t stopped adopting the technology. This accelerated AI adoption is a major driver of the push for more regulations to guide how AI models are built and used, especially as concerns about bias, privacy and misinformation continue to grow.
“AI development has outpaced our ability to fully understand its implications,” noted Černiauskas. “2025 will be a year where AI safety becomes a central focus for governments and organizations.” The EU’s AI Act, which is the world’s first comprehensive piece of legislation on AI, provides a legal framework on which many governments are looking to build regulatory AI frameworks.
Green AI, which prioritizes energy-efficient computing, is another growing trend. With AI models requiring vast compute and energy resources, reducing their environmental footprint is now an ethical necessity.
Good News On The AI Horizon
While the challenges around compute power, ROI and bias still loom large for AI, Chaudhry said that multimodal models will take center stage in 2025, particularly text-to-video models that produce longer, more realistic videos. “We’ll see significant advancements in how these models adhere to the laws of physics, improving their quality and applicability,” he noted.
Another area of promise is automated machine learning which, according to Černiauskas, could democratize machine learning and AI for everyday people. “With AutoML, professionals across diverse industries can create AI tools specific to their needs. This has the potential to accelerate AI adoption and open up new possibilities.”
There’s no doubt that there’s real utility value in AI systems and applications. But the greater conversation around AI in 2025 will likely be defined by balance — safely getting the best out of AI, without overselling its promise.
