More Gemini Features Coming to Google TV
Google TV integrates Gemini AI for image/video generation, enhances Google Photos, and adds a dedicated YouTube Shorts feed to the homepage.
Read moreNowadays, it seems that algorithms are involved in every aspect, identifying dangerous abnormalities. Just this month, Google released a series of news, with a study showing that its artificial intelligence system is better at accurately detecting breast cancer in breast X-ray examinations than doctors.
However, for many healthcare professionals, despite the obvious ability of algorithms to process data, the subtle, judgment-based skills of nurses and doctors are not so easily digitized.
Dr. Adewole Adamson, a dermatologist and assistant professor at Dell Medical School, said in an interview with Edge magazine: "There is a perception in society that finding more cancer is always better, but that is not always the case." "Our goal is to find more cancer that actually leads to death." But the problem is that "there is no gold standard for cancer."
As research has found, you can show the same early lesions to a group of doctors and get completely different answers as to whether they can be diagnosed as cancer. Even if they do agree on the appearance of the lesion - and their diagnosis is correct - they cannot know if the cancer poses a threat to someone's life. Incorrect judgments can lead to overdiagnosis.
Once you label it as cancer, it triggers a series of medical interventions that can be painful, expensive, and life-changing. In the case of breast cancer, this may mean radiation therapy, chemotherapy, breast tissue removal (lumpectomy), or complete removal of one or both breasts (mastectomy). These are not hasty decisions.
Adamson said that this complex diagnosis was not given due attention in Google's research. First, the company's researchers trained their image algorithm to determine whether it was cancer. However, the lack of a fixed standard for cancer diagnosis, especially for early-stage cancer, raises questions about whether such training data provides a good baseline. Secondly, Google's algorithm can only produce binary results: yes, it is cancer, or no. As Adamson said in a recent paper, medical diagnostics require room for uncertainty, a third option that represents the gray area of diagnosis, prolonging the diagnostic process rather than ending the debate.
When asked about these issues, Google's team told Verge that their algorithm reduces the false positive rate (events that occur when something is incorrectly identified as cancer), reducing the threat of overdiagnosis. They also emphasized that they will investigate the kind of non-binary analysis advocated by Adamson in the future.
"This is exactly the next step in our research with partners," said a Google Health spokesperson. "We hope to explore workflow considerations, user interface considerations, and more in many other areas."
However, for Adamson, these challenges are bigger than a paper. He said that overdiagnosis "is a problem for many different cancers; prostate cancer, melanoma, breast cancer, thyroid cancer. If AI systems get better at detecting smaller and smaller lesions, you will create many false patients with 'diseases'."
When artificial intelligence is combined with medicine, overdiagnosis is a challenge, but for some doctors, the root of the problem is deeper. Artificial intelligence can replace entire medical job categories, such as radiology, but this idea is not found in papers or algorithms, but comes from blind confidence in artificial intelligence.
In 2016, artificial intelligence pioneer Geoffrey Hinton (one of the three "fathers of artificial intelligence" who won the 2018 Turing Award) said, "People should stop training radiologists now. It is obvious that deep learning will do better than radiologists in five years." In 2017, Andrew Ng, co-founder of Google Brain, repeated this point when commenting on an algorithm for detecting pneumonia with X-rays: "Should radiologists worry about their jobs?"
In recent years, this rhetoric has subsided, but for real radiologists, these voices always sound a bit misleading and insulting. While algorithms can certainly identify specific pathological features in medical images, this is far from the same as wearing a white coat and "observing, smelling, asking, and cutting" in the ward.
The core of the problem is that radiologists do more than just look at images, said radiologist and health technology consultant Hugh Harvey. "This is a complete misunderstanding of the work of radiologists," he said. "This work is more like reading a novel and trying to write a summary of it."
In a blog post in 2018, Harvey pointed out that medical diagnosis involves receiving patients, collecting data in various ways (from fluoroscopy, ultrasound, biopsies, etc.), correlating it with other parts of the diagnosis, and engaging in various ancillary tasks such as teaching, training, and auditing the work of others. Harvey said, "Artificial intelligence cannot replace the work of radiologists in any sense." "Artificial intelligence can only find things that humans find difficult to find and present them to radiologists for their opinions."
The root of the overconfidence in the world of artificial intelligence lies not in animosity towards radiologists, but in the inherent affinity of artificial intelligence itself. So far, machine vision has proven to be the most powerful form of deep learning and the dominant style of artificial intelligence. It is a test of image recognition that opened the current AI boom in 2012, and from autonomous vehicles to facial recognition, deep learning vision algorithms support its most powerful applications.
Because of this, AI researchers have achieved a lot by applying relatively standard vision algorithms to medical datasets. This produces many "firsts" because AI learns to identify feature X in data Y and leaves people with the impression of rapid technological progress. Doctors say that the most useful tools in applications like these - those that simply mark data features for doctors to verify - are the most useful. But those who try to diagnose more complex cases themselves may not be able to handle potential medical challenges. Especially when many of the created algorithms have not yet been integrated into clinical environments.
As Harvey said, "Deep learning is being treated as a hammer, and tech companies are also looking for nails, but a hammer is not suitable for all nails."
If there is a consistent theme in the overlap between artificial intelligence and medicine, it is that medical issues are not as simple as they initially appear.
Healthcare journalist Mary Chris Jaklevic pointed out in a recent article that much of the misinformation here stems from the "machines vs. doctors" narrative found in many artificial intelligence research and subsequent reports. Such narratives are both catchy and eye-catching, attracting readers' interest in the present and shaping their understanding of the future. But it is also one-dimensional, reducing the complexity of medical diagnosis to a few numbers and simple images.
Nevertheless, most experts involved in this work - whether programmers or doctors - remain cautiously optimistic about the potential of artificial intelligence in the field of healthcare. As Adamson pointed out, it is precisely the scalability of artificial intelligence that makes it so powerful and gives it so much hope, while also requiring caution.
He pointed out that once the algorithm has been thoroughly reviewed, it can be deployed quickly and easily almost anywhere in the world. But if these checks are rushed, then adverse effects such as overdiagnosis will quickly increase.
"I don't think artificial intelligence should be thrown in the trash, quite the opposite," Adamson said. "If designed properly, it has the potential to do good. What concerns me is not the technology of artificial intelligence, but how we will apply it."
Google TV integrates Gemini AI for image/video generation, enhances Google Photos, and adds a dedicated YouTube Shorts feed to the homepage.
Read moreAlphabet Q1 paid subscribers reached 350M, fueled by YouTube & Google One. YouTube ad revenue slightly missed at $9.88B amid Premium growth. Total Q1 revenue hit $109.9B, beating forecasts.
Read moreAnthropic releases Claude connectors for Adobe, Blender, and Ableton, enabling AI workflow automation and app control. It also sponsors the Blender Development Foundation.
Read moreApple introduces an App Store model splitting 12-month plans into monthly installments. On iOS 26.4+, it lowers upfront costs without reducing total price, adding billing tracking.
Read moreGoogle launches Aadhaar digital IDs in Wallet for India, enabling secure QR/NFC verification. A phased partner rollout targets key sectors, while passport-based credentials expand globally.
Read moreYouTube tests "Ask YouTube," an AI tool blending text and video for complex queries. Initially for US Premium users, it supports interactive follow-ups, expanding Google's AI search.
Read moreGoogle grants US DoD classified AI access, joining OpenAI and xAI in defense deals amid Anthropic's dispute. Contractual safeguards against surveillance and lethal weapons remain unenforceable.
Read moreMistral AI launches Workflows, a Temporal-powered orchestration platform for secure, Python-driven enterprise AI automation featuring human-in-the-loop approvals.
Read moreCustom subliminal audio maker

AI Social Media Manager

AI Search Visibility

Workout and Calorie Tracker
Stock research platform

AI interview assistant

Scientific figure generator

AI courses with real code