Google just beat humans at spotting breast cancer — but it won’t replace them
BUSINESS · TECH · SCIENCE
January 4, 20203 min read, 609 words

Google just beat humans at spotting breast cancer — but it won’t replace themGoogle just beat humans at spotting breast cancer — but it won’t replace them

 in 
Published: January 4, 2020  |  3 min read, 609 words
A new study reveals Google AI is often better at detecting breast cancer than actual doctors Google is developing artificial intelligence to help doctors identify breast cancer, according to a research paper published in Nature today. The model, which scans X-ray images known as...
Google just beat humans at spotting breast cancer — but it won’t replace them Read more

Scores for this article.

Percentage of critic and public trust in this article.
Factual Error1
img-trusted
N/A
critic score
critic reviews: 1
img-trusted
67%
public score
public reviews: 3

CRITIC REVIEWS

Well Sourced
January 7, 2020
An interesting and investigative look at ways that advanced technology and medical science can be brought together in newly productive endeavors. Good sourcing, regrettably though, only the Abstract of the study is available rather than the full version. Nevertheless, the Abstract plus the other linked items, do provide additional context and useful information. The Author presents the referenced subject matter evenly without editorializing. Fine quotes and commentary. In a general sense, very helpful information for those following medical advances and even for those with only a passing interest. Well worth a read. Four stars!
January 7, 2020
Is this helpful?

PUBLIC REVIEWS

Factual Error
January 8, 2020
There is one very big mistake in this article, breast cancer is the second most common CANCER cause of death in women, NOT the second leading cause of death in women period. It may be a typo because the citation they provide has the correct description, but claims like that do hype up the fear of breast cancer and obviously spread misinformation. The description of the actual research seems sound and does highlight the positive side to AI. It also brings up that the AI system is not perfect either and currently both actual humans and an AI algorithm may be the strongest option.
January 8, 2020
Is this helpful?
Balanced
January 6, 2020
Although the headline is a bit more celebratory than the content of the article suggests, the piece is balanced and provides all the various sources and context necessary to understand the importance of the story.
January 6, 2020
Is this helpful?
Credible
January 6, 2020
I didn't see major shortcomings in this as a general-audience story. But there are MANY non-trivial flaws: (1) The story cites the NYT (not any peer-reviewed study) in saying current testing misses 20% of cancers but makes no mention of the net cost of those misses. Anybody halfway close to the field knows that there is concern about over-aggressive responses and complications of surgery from attacking cancers that are less likely to present actual risk. It leaves to the reader's imagination the consequences of positive and negative errors. (2) The article uncritically quotes Google's PR that this is intended to *supplement* existing readings. This both shows the link to Google PR and also raises the question of how the technology would be put into practice. Over-promising by autonomous driving advocates makes me wonder whether the next step down this slippery slope is to push for Google-proprietary tech replacing current practice, (3) I've separately seen comments that Google's failure to publish training parameters should not be cause for concern, but AI modeling is rife with over-parameterization and tweaking that cause models to look great against the test and validation data sets, but much less useful in the wild. Scientific practice has long called for independent validation of findings and Google's blasé response is disquieting. (4) The article makes no mention of Google's eventual role in health care, other than to note that Google has had privacy problems with earlier forays into health care. The story does not mention Google's funding for this as coming from a non-profit foundation, and presumably Google intends to “monetize” this or similar efforts. Again, for general readers who mostly take away, “AI techniques might be helpful some day in mammogram evaluation,” this article works OK. It's only that the intended audience (technology fans) may over-interpret this as promising a free lunch—a better world thanks to Google's research who need to read between the lines.
January 6, 2020
Is this helpful?