We are living in the most consequential decade in artificial intelligence’s history. We don’t know exactly what the future will look like, and we don’t know exactly what innovations it will bring, but we know one thing for certain: AI will be shaped by the people who have access to the most computing power and data.
And that makes this a critical time for higher education. For academe to retain the best talent, we must have access to a scale of computing that has long been unimaginable in universities. In May the Stanford computer scientist Fei-Fei Li warned that academe is “falling off the cliff pretty fast” in terms of its resource relationship to Big Tech: Stanford’s natural-language lab had 64 graphics processing units, Stanford’s shared faculty computing cluster had 800, while Meta was aiming to have 350,000 by the end of 2024.
Today, those with the most access are in Silicon Valley or outside the United States, where computing power lies with private business or with foreign governments. But here in New York State, we are starting to level the playing field for researchers at our top universities. Other institutions should follow our lead if they don’t want to be left behind.
This past spring, New York’s governor, Kathy Hochul, and the state legislature established Empire AI, a first-of-its-kind consortium comprising New York’s top public and private academic institutions, as well as philanthropic partners. Empire AI was created to address the growing imbalance in computing power between industry and academe and to orient the future of AI development toward the public good. Through the consortium, our state’s top colleges are being given shared access to a facility housing an amount of computing power that was previously unthinkable. Alpha, the first phase of Empire AI, is already the 243rd most powerful computer in the world according to the TOP500 list, the industry-standard benchmark for supercomputers. Still, this enormous computing power represents only a fraction of the total power Empire AI will eventually wield.
To understand the importance of this access, you first must understand that the imbalance between academe and industry is a relatively new development. For decades, the best minds were attracted to academe to work freely on open-ended long-term research with students and peer faculty and to educate the next generation of intellectuals. But over the past decade, the story has changed.
The largest tech companies, like Google, Apple, and Meta, have an advantage over academe because of their outsize access to computing power and data. This causes an AI brain drain, where many talented researchers are opting to go into industry rather than higher education. Computing power has become cost-prohibitive, and academic researchers have been priced out of the market. If industry alone has all the access, AI development will be largely driven by competitive market forces and focused on commercial interests.
As academics, we may not develop solutions that are immediately lucrative, but our discoveries are vital for the development of technology in the long term. Research and development takes time, especially when it aims to take on the biggest issues facing the world today. That’s not to say that market-oriented impulses won’t lead to people-oriented AI solutions. They already have. But the world won’t be able to experience the full breadth of possibilities AI has to offer until academics can compete too.
Providing researchers access to world-class computing is one of the best recruiting tools academe has to offer. It is essential that the best minds continue to see research universities as a destination better than or equal to the opportunities being offered by Big Tech. In the next decade, we need to be advancing research but also training the next generation of researchers.
If states are looking for a way to bring their top researchers into the fold, Empire AI is the model to follow. It brings together New York’s public universities, its top private colleges — Columbia, Cornell, NYU, and Rensselaer Polytechnic Institute — and the Flatiron Institute, a private research institution, to concentrate the state’s academic strength in one powerful consortium. What’s more, alongside a $275-million investment from the State, $125 million was contributed by the founding institutions and other private partners, including the Simons Foundation and the philanthropist Tom Secunda, making Empire AI a truly collaborative effort. Other states have the academic talent to create similar consortia, and it’s critical that they do so. If they don’t, their researchers will be left behind, allowing big-tech companies to control AI’s future.
Building AI consortia at the state level also supports laudable federal efforts like the National Artificial Intelligence Research Resource (NAIRR), which aims to provide academic researchers across the country with the resources they need to perform high-level AI research. As it stands, a combination of state and federal funds is needed to meet the massive demand across the board for computing resources. By combining state power with private philanthropic power, Empire AI serves to augment the strength of NAIRR and other federal efforts. The more statewide consortia we have across the country, the stronger our collective research will be.
Unfettered by the demands of the market, academics have a special kind of freedom: one that allows them to not focus solely on making a profit. Instead, we can look at the most pressing issues and work to solve them. No matter how long it takes.
