Big Tech Cares About Profits, Not Curing Cancer
When critics point out the immense human, environmental, and financial costs of the ongoing AI investment boom, one of the stock answers of Big Tech has been to promise that the costs are irrelevant because the benefits of AI will be almost mythological in scale. One particularly prominent version of this argument is that AI will help find a cure for cancer, a claim that Dario Amodei and Sam Altman (over and over) have made.
Alphabet, the parent company for Google, has been particularly aggressive in pushing that narrative. Alphabet President Ruth Porat said last fall that “[w]e should be able to cure cancer in our lifetime” thanks to AI. Demis Hassabis, the co-founder of DeepMind, the tech giant’s most accomplished AI lab, has gone even further, suggesting in a 2024 60 Minutes interview that “we can cure all disease with the help of AI.”
The emptiness of that rhetoric, always suspected, is now quite clear.
Two weeks ago, Google shut down AlphaFold, DeepMind’s AI-powered protein structure prediction tool. AlphaFold represented perhaps the most persuasive data point in favor of the narrative that AI could lead to scientific breakthroughs, including in cancer research. The model’s ability to accurately predict protein structures helped Hassabis become a Nobel laureate. A week after shuttering AlphaFold, Google removed Hassabis from his role directing DeepMind, using the time-worn tactic of giving Hassabis a face-saving “promotion” to a powerless role.
The *Financial Times* [shed light](https://www.ft.com/content/1453e9c2-4922-482f-8720-0bafd7e07df7?syn-25a6b1a6=1) on Google’s reasons for these moves:
While Hassabis is widely admired for his scientific leadership, several people familiar with Google’s thinking saidsenior executives had become frustrated by what they saw as his lesser focus on the commercial demandsof the company’s AI business.
His decision to make AlphaFold . . . freely available became a source of tension inside Google, according to a person familiar with the matter.The project generated little commercial return despite the scale of its investment.
In other words, AlphaFold was too much of a scientific endeavor for Google’s liking. So Google shut it down and kicked its key backer upstairs to a role where he couldn’t hold back the tech giant’s profits.
If someone had consciously set out to demonstrate the disingenuousness of Big Tech’s grandiose AI promises, they could hardly have scripted it better.
AlphaFold’s protein-prediction capabilities arguably represent the most significant scientific achievement that can be attributed directly to advances in artificial intelligence.
Proteins consist of chains of molecules called amino acids. Many amino acid chains “fold” in characteristic ways to form the structure of specific proteins. Misfolded proteins directly cause certain diseases and are believed to play a role in many others, including Alzheimer’s, Parkinson’s, and certain cancers. Consequently, knowing the physical structure that results from protein folding is important for many areas of drug discovery and development. But scientists have long struggled to predict a protein’s three-dimensional structure even if they know the protein’s amino acid sequence.
DeepMind designed AlphaFold to help resolve this prediction problem. The first version of AlphaFold won a prominent protein-folding prediction competition in 2018. Its successor, AlphaFold 2, blew away other entries at the same competition two years later, helping Hassabis become one of three laureates for the 2024 Nobel Prize in Chemistry. The last iteration of the model, AlphaFold 3, was able to predict more complicated structures and interactions involving proteins, DNA, RNA, and other molecules.
One would think that Google would have bent over backwards to boost AlphaFold. In addition to providing a rare example of a concrete AI-driven scientific achievement, AlphaFold offered an at least vaguely plausible path to the AI industry following through on its rhetoric about curing cancer through AI.
The company decided to instead kill the project because (1) AlphaFold’s breakthrough was not generalizable and, relatedly, (2) Google was struggling to profit from it.
AlphaFold is an amazing tool, but it is still just a tool. While it represents a major technical breakthrough for the specific tasks it was designed to perform, it is of little-to-no use in unrelated (or, indeed, even in closely related) fields of science. The additional capabilities of successive generations of AlphaFold came not from AlphaFold improving itself, but from its (human) developers imbuing it with new capabilities by designing more sophisticated models with improved architectures and expanded training data sets. So while AlphaFold’s capabilities are impressive, they are also quite narrow.
On that note—AlphaFold did not “solve” the protein folding problem, as some news sources (including some science-focused outlets) suggested. There are a significant number of proteins whose structure AlphaFold cannot reliably predict. Even for protein structures it predicts with high accuracy, AlphaFold cannot explain why or how the proteins fold the way they do. (AlphaFold’s inability to generalize outside of the tasks it was trained to perform very much applies to large language models, as Gary Marcus frequently points out.)
Which leads to the second point. Alphabet is spending $200 billion in 2026 alone on new AI infrastructure. Because Hassabis insisted on making AlphaFold’s code and database publicly available, there was little prospect that the project would help Google recoup its enormous AI investments. Had Hassabis agreed to make the invention proprietary, Google would enjoy a commercial monopoly on its use. Because he didn’t, the technology’s commercial potential was limited. That, it seems, sealed the fates of AlphaFold and Hassabis.
In viewing cancer through the lens of profit maximization, Google is simply following in the footsteps of the rest of Corporate America. Myriad Genetics, a corporation based in Salt Lake City, infamously filed patents on a number of human genes that it discovered through novel testing techniques, including two genes linked to breast cancer, so that it could retain a monopoly on the market for selling tests for the gene. The Supreme Court thankfully ended that particular practice by ruling in 2013 that human genes could not be patented. Cancer drugs can, however, still be proprietary, a fact that drug companies have used to set the price for many patented cancer treatments at extortionate levels.
Against that backdrop, Google’s actions are not surprising. But they certainly are galling. Google, like the rest of the AI industry, had gone to great lengths to convince policymakers and the public that we all should get behind Big Tech’s vision of an AI future. It then sidelined Hassabis because (God forbid) he dared to take some modest steps that might have allowed others to have some small share in that future.
At least the charade can now end. Big Tech does not, in fact, care about curing cancer — unless it happens to boost profits. Let’s all remember that the next time a tech executive attempts to justify the AI industry’s actions by touting all of the scientific wonders AI will supposedly help humanity achieve.
Matt Scherer is a fellow at Open Markets Institute, where his research and advocacy focus on developing policy responses to the eventual bursting of the AI bubble. His Hard Reset pieces focus on highlighting the risks posed by the AI bubble and pushing back against the hype that is inflating it. The opinions expressed here are solely his own.