In September, the annual AI Horizons summit was held in Pittsburghās Bakery Square. The summit was hosted by the AI Strike Teamā a coalition of tech CEOs, energy executives, and policy makersā and was a forum for those groups to rub shoulders, check out developing technology, and to address public criticism over the AI buildout.
At $643 a ticket, the audience for that response was unclear. I was able to attend on behalf of MWA, and here is what I learned:
Big tech sees AI as the solution to AI
One of the first panels I attended was about the publicās mistrust of AI. The elephant in the room was OpenAIās recent āHugging Faceā incident. During an internal security test, OpenAI’s agents slipped past controls meant to keep them off the internet and compromised the systems of Hugging Face, another AI company. This AI-breakout sparked fierce debate about how AI research should be regulated among developers, politicians, and regular people alike.
During the panel, questions over how AI could regain regular peopleās trust drew predictably rosy answers. Panelists explained how they use AI to track health and wellness; how AI could one day be used to manage the electrical grid and hospitals; how AI could potentially level the playing field, in their view. Leigh-Ann Russell, Global Head of Engineering at BNY said, āThereās a lot of disadvantaged communities who havenāt had education around AI, but actually this is the greatest opportunity for them to level up.ā Russel did not address how industrial activities like AI data centers, powerplants, transmission lines, often target those same communities and further burden them with increased pollution, noise, utility costs, and lost tax revenue.
The panelists conceded that AI models grow more capable by the day, and that safety and control over them is no longer a hypothetical concern. Zico Kolter of Gray Swan AI said monitoring these systems will demand far more attention in the years ahead. Asked about safeguards against AI going rogue, Kolter pointed to his company’s enforcement models, AI agents whose job is to make sure other AI agents stay within the bounds of their tasks and training. For now, the watchers and the watched are made by the same company.
Environmental impacts were hardly mentioned.
Throughout the conference, AIās software side, including agents, programs, and chatbots, was often discussed separately from the data centers, power plants, and pipelines needed to support it. That meant plenty of discussion about how quickly AI can grow, but much less about the combined environmental impacts of that growth.
As tech companies look to expand AI into more parts of everyday life, energy companies see data centers as a major new source of demand. Both stand to benefit from a rapid buildout.
One of the few checks on that pace is environmental review through permitting. Jaclyn Presnal, VP of Power Innovation & New Energy Ventures at Williams Companies, named āthe lack of permitting reformā as a primary industry challenge.
On another panel, EQT CEO Toby Rice said, āWhen this AI boom came out a couple years ago, I got really excited,ā and argued that natural gas is well suited to power large data centers. He also pointed to the Homer City project, saying it will generate $120 million in royalties for property owners who lease to EQT. The broader EQT agreement has been valued at $15 billion.
That theme ran throughout the conference. The economic case for building quickly was front and center. The environmental consequences, and the communities that will live with them, were far less visible.
Watch the full program here: