Project Amber: Alphabet's Depression Biomarker Search

Alphabet’s X, known as the “Moonshot Factory” of the Google parent company, recently released a new blog post concerning Project Amber. This initiative, developed over the past three years, is now being shared as an open-source resource for the mental health research field, with the intention of fostering further advancements. The X project initially aimed to pinpoint a specific biomarker indicative of depression; however, this goal was not achieved (and researchers currently believe that a single biomarker for both depression and anxiety is unlikely to exist). Nevertheless, X anticipates that its work utilizing electroencephalography (EEG) in conjunction with machine learning to identify such a biomarker will prove valuable to others.

The researchers at X had hoped that depression, similar to other medical conditions, might exhibit a distinct biomarker to facilitate more straightforward and objective diagnoses for healthcare professionals. This, in turn, could potentially lead to more effective and consistent treatment approaches. Prior studies utilizing EEG, conducted in laboratory settings with specifically designed games, had shown that individuals experiencing depression often displayed reduced EEG activity when “winning” these games, suggesting a possible avenue for biomarker identification.

While these initial studies appeared promising, the X team recognized the need to refine the EEG collection and interpretation processes to make them more practical for real-world diagnostic environments, such as clinics or public health laboratories, and more accessible to both users and technicians.

A particularly noteworthy aspect of this endeavor, and the accompanying blog post from Alphabet, is that it details a multi-year investigation that ultimately did not yield the expected results—a narrative rarely shared by major technology companies.

In reality, this represents a compelling illustration of a point often overlooked by critics of large tech companies’ strategies—that certain problems cannot be resolved through solutions mirroring those found in software and engineering.

The X team has summarized its key learnings from this extensive research project in three primary points regarding user research, all of which highlight the limitations of relying solely on objective biomarker detection, especially in the context of mental illness. According to the researchers:

X is releasing Amber’s hardware and software as open-source on GitHub, and is also issuing a “patent pledge” guaranteeing that X will not pursue legal action against anyone utilizing the EEG patents associated with Amber through the use of this open-sourced material.

It remains uncertain (though improbable) whether this open-source approach would have been adopted had Amber successfully identified a single biomarker for depression. However, it is possible that the broader community can leverage the team’s advancements in making EEG more accessible outside of specialized facilities to uncover new insights.