Three years after ChatGPT freaked everybody out, the AI crowd has landed on the same page: less hype, more homework. 

World Summit AI 2025 made one thing clear: we’re no longer in the potential phase. We’re in the impact phase. Children’s lives are being saved through AI-powered cancer research. Patients are getting matched to clinical trials. Businesses are making actual money. 

Same page, finally

The ninth annual summit had most everyone agreeing on three things: 

• Being excited about the role AI is already playing in solving real-world problems.
• Remaining vigilant against the potentially negative – if not catastrophic – aspects of pushing through innovation without careful consideration. Guardrails, please! Don’t pop the bubble by overstretching the overall economic system, please!
• Knowing the nitty-gritty of what’s required to apply AI innovation successfully.

This shared mentality marks a notable shift: earlier, industry discussions mainly centred on the difficulties of creating real impact with AI. Now, the focus is more on sharing measurable results. 

Or as the event website framed it, “Under this year’s theme, ‘Back to the Future: It’s About Time’, we explored how AI can help us reclaim time for creativity, connection, and progress, while guarding against the risks that threaten our future.”

What’s happening with AI today is a call to action to scale our humanity.

Diversity over tech bros

While some are racing to scale GenAI to achieve Artificial General Intelligence (AGI), others are scaling their efforts to address real-world problems, sensible regulation, and keeping it all people-centred. It’s this latter group that was particularly well represented at this event. And if this group had a rallying cry it would be: “What’s happening with AI today is a call to action to scale our humanity.” 

Hyperscalers like Google and Amazon were also present, but they are being less vocal about AGI, focusing more on down-to-earth approaches that achieve short-term ROI. After all, haven’t we sowed enough money already? Isn’t it time for some reaping?

Checklist for AI success

For instance, Rahul Pathak, the VP for Data and AI Go-To-Market strategy at AWS, gave the talk ‘Unlocking an AI mindset‘ in which he distilled conversations with 500+ customers into a now rather familiar framework for successful AI implementation: start with the problem, not the tech; connect your own data, because that’s where differentiation lives; and set guardrails, so discipline becomes freedom. Oh, and just do it – or get left behind.

And by following these rules, organizations can enjoy a 3.7x return on every dollar spent, with returns typically landing within 13 months. For instance, AstraZeneca freed up their scientists for more valuable work by saving 43,000 hours in biomarker analysis. How’s that for ROI?

When AI saves lives

Joe Depa is EY’s Global Chief Innovation Officer. In his presentation ‘AI for Good: delivering impact with data, trust, and value‘, he shared a personal story that highlighted the potential of AI. In 2022, his team raised funds to support childhood cancer research – a cause close to his heart. Working with Dr. Soheil Meshinchi at Fred Hutch Cancer Center, they helped create 1,200 genetic sequencing profiles for acute myeloid leukemia, forming one of the largest pediatric cancer databases in the world.

The impact? Within three years, they identified 100 potential treatment targets, launched two clinical trials, and saw children who otherwise wouldn’t be alive today enter remission. Additionally, the database was made publicly available to researchers worldwide.

Depa then set a challenge: “Everyone in this room has experienced something – cancer, mental health challenges, disabilities, aging parents. Whatever you’re passionate about, apply AI to help. Work with a nonprofit. Make the world better.”

New opportunities for cancer patients

Depa’s story is far from an isolated example. At City of Hope National Medical Center – which bills itself as the birthplace of biotech (not entirely without reason, given its role in the development of synthetic human insulin) – they are already using AI systems that reliably summarize hundreds of pages of patient documents in minutes, allowing oncologists to focus on care rather than paperwork. They’ve also streamlined the clinical trial process.

In her presentation ‘Harnessing AI for multimodal data analytics: Practical applications in health, science, and research‘, the hospital’s Chief Research Technology Officer (and former Microsoft Chief AI & Precision Health Officer) Dr. Helia Mohammadi noted that these are not just experiments but “already demonstrating real-world impact by reducing delays in care and opening new opportunities for patients who previously had limited options.”

Her key message: “AI is not about adopting technology for its own sake. It’s about solving high-impact problems responsibly, measuring outcomes, and ensuring equity in care. The healthcare industry is at a pivotal moment. With the right governance, collaboration, and ethical guardrails, AI can redefine how research is conducted, how care is delivered, and how hope is restored.”

“What you see is not always 100% truthful.” — Rob Petrosino, FBI

So much for the wow. Now for the yikes

Rob Petrosino is Strategic Engagement Advisor for AI at the FBI. In his presentation, ‘Digital deception: How AI is rewriting reality‘, he made clear that the deepfake future has already arrived. 

A 2024 survey showed that 51% of people could not distinguish between deepfakes and real content. Meanwhile, 15% of US teens have admitted to seeing deepfakes of classmates. “And I can assure you those deepfakes are not PG-rated. And that this number is only about teens who admitted seeing them,” says Petrosino. Then came another number: “31 teens between ages 13-17 committed suicide in the US from 2021 to 2024 related to deepfake content.”

Meanwhile, Google searches for ‘AI girlfriend’ are up 2,400%. And with them: AI psychosis. In just six months, according to Petrosino, 25 documented cases emerged of people interacting with AI, having core beliefs supported or shattered, and experiencing complete mental breaks. 

“We’re past the crisis stage. You need to stop yourself and realize that what you see is not always 100% truthful,” advises Petrosino. “More likely than not, you’ve consumed this media. I just hope it did not work to change any of your opinions.”

“When people rise, empires always fall.” — Karen Hao

The empire problem

The journalist and author of the New York Times bestseller Empire of AI, Karen Hao, gave an equally urgent keynote, ‘How Silicon Valley is reshaping the world‘. Hao argues for a complete rethink of how we develop and embrace AI. Companies like OpenAI are already resembling empires of old, according to Hao, consolidating power, concentrating wealth and threatening democracy itself.

There are simply too many costs associated with generative models like ChatGPT and the quest for AGI. These approaches have very little to do with potential societal benefits related to education, healthcare, and climate change – the very arguments of Silicon Valley to pursue this path.

“We’re throwing trillions of dollars at compute, fresh water, and energy – resources desperately needed elsewhere,” Hao says. “And this money will never come back.”

Her prescription was clear: Stop scaling. Shift away from colossal general-purpose models and reinvest in small, task-specific AI systems that tackle challenges well suited to AI’s computational strengths, such as DeepMind’s AlphaFold, the work behind the 2024 Nobel Prize in Chemistry.

Hao also offers a get-out clause: “Empires are made to feel inevitable, but history has shown us that when people rise, empires always fall.”

“It sees the purchase but not the longing, the scroll but not the emptiness.” — Jason Snyder

Give friction to the flattening

Futurist Jason Snyder reinforced Hao’s critique more philosophically. In his presentation, ‘Time worth wasting: Preserving meaning in a predictive world‘, he shared his more personal approach.

“AI trains on everything we do but never on why we do it. It sees the purchase but not the longing, the scroll but not the emptiness. It can simulate knowledge but not wisdom, replicate tone but not tenderness. In a world optimized for efficiency, something deeply human is slipping through the cracks: meaning itself.”

He noted that while figures like Leopold Aschenbrenner predict we’ll reach AGI by 2027, these people are “mistaking faster math for deeper intelligence… In fact, it is only working to flatten our human experience. We need all those blips of humanity, those quirks, those tension points – or as someone described it: that friction – because that’s where we get meaning, that’s where our stories grow and evolve and become more meaningful.”

While he is still a tech-optimist, Snyder does offer some solid non-tech advice: “Stay deliberately human. Teach children boredom. Make something that doesn’t scale. Let time feel slow – slowness is reverence, not inefficiency. Question everything AI tells you. Be the devil’s advocate. Every time you choose friction, you choose to matter.”

So, what’s your call to action?

In short: nobody knows what’s coming. So, get involved – and stay freaking human. Embrace the blips.