Therapeutic Area

Geoffrey Hinton is one of the creators of Deep Learning, a 2019 winner of the Turing Award, and an engineering fellow at Google. Last week, at the company's I/O developer conference, we discussed his early fascination with the brain, and the possibility that computers could be modeled after its neural structure--an idea long dismissed by other scholars as foolhardy. We also discussed consciousness, his future plans, and whether computers should be taught to dream. The conversation has been lightly edited for length and clarity. Nicholas Thompson: Let's start when you write some of your early, very influential papers. Everybody says, "This is a smart idea, but we're not actually going to be able to design computers this way." Explain why you persisted and why you were so confident that you had found something important. Geoffrey Hinton: It seemed to me there's no other way the brain could work. It has to work by learning the strength of connections. And if you want to make a device do something intelligent, you've got two options: You can program it, or it can learn. And people certainly weren't programmed, so we had to learn. This had to be the right way to go. NT: Explain what neural networks are. GH: You have relatively simple processing elements that are very loosely models of neurons. They have connections coming in, each connection has a weight on it, and that weight can be changed through learning. And what a neuron does is take the activities on the connections times the weights, adds them all up, and then decides whether to send an output.

The purpose of this retrospective study is to measure machine learning models' ability to predict glaucoma drainage device failure based on demographic information and preoperative measurements. The medical records of sixty-two patients were used. Potential predictors included the patient's race, age, sex, preoperative intraocular pressure, preoperative visual acuity, number of intraocular pressure-lowering medications, and number and type of previous ophthalmic surgeries. Failure was defined as final intraocular pressure greater than 18 mm Hg, reduction in intraocular pressure less than 20% from baseline, or need for reoperation unrelated to normal implant maintenance. Five classifiers were compared: logistic regression, artificial neural network, random forest, decision tree, and support vector machine.

One of lung cancer's most lethal attributes is its ability to trick radiologists. Some nodules appear threatening but turn out to be false positives. Others escape notice entirely, and then spiral without symptoms into metastatic disease. On Monday, however, Google unveiled an artificial intelligence system that -- in early testing -- demonstrated a remarkable talent for seeing through lung cancer's disguises. A study published in Nature Medicine reported that the algorithm, trained on 42,000 patient CT scans taken during a National Institutes of Health clinical trial, outperformed six radiologists in determining whether patients had cancer.

Measles, once thought to have been eliminated in the U.S., is popping up in isolated outbreaks as a result of skipped well-child visits and parents' fears that the measles-mumps-rubella (MMR) vaccine is linked to autism. Though some 350 measles cases occurred in 15 states in the first three months of 2019, more than half were in Brooklyn, N.Y., and nearby Rockland County, N.Y., where large religious communities have adopted anti-vaccine positions. Rockland County responded by pulling 6,000 unvaccinated children out of schools and barring them from public places. The county's actions were effective; in just a few months, 17,500 doses of MMR were administered to area children. Yet, wouldn't it have been better to contain the outbreak before it got started?

The condition is the leading cause of cancer-related death in the U.S., and early detection is crucial for both stopping the spread of tumors and improving patient outcomes. As an alternative to chest X-rays, healthcare professionals have recently been using computed tomography (CT) scans to screen for lung cancer. In fact, some scientists argue that CT scans are superior to X-rays for lung cancer detection, and research has shown that low-dose CT (LDCT) in particular has reduced lung cancer deaths by 20%. These errors typically delay the diagnosis of lung cancer until the disease has reached an advanced stage when it becomes too difficult to treat. New research may safeguard against these errors.

While there are anemones in those images, it also seems that the system is recognizing a clownfish.

Researchers at UC Davis and UC San Francisco have found a way to teach a computer to precisely detect one of the hallmarks of Alzheimer's disease in human brain tissue, delivering a proof of concept for a machine-learning approach capable of automating a key component of Alzheimer's research. Amyloid plaques are clumps of protein fragments in the brains of people with Alzheimer's disease that destroy nerve cell connections. Much like the way Facebook recognizes faces based on captured images, the machine learning tool developed by a team of University of California scientists can "see" if a sample of brain tissue has one type of amyloid plaque or another -- and do it very quickly. The findings, published May 15, 2019 in Nature Communications, suggest that machine learning can augment the expertise and analysis of an expert neuropathologist. The tool allows them to analyze thousands of times more data and ask new questions that would not be possible with the limited data processing capabilities of even the most highly trained human experts.

Google AI researchers working with Northwestern Medicine created an AI model capable of detecting lung cancer from screening tests better than human radiologists with an average of eight years experience. When analyzing a single CT scan, the model detected cancer 5% more often on average than a group of six human experts and was 11% more likely to reduce false positives. Humans and AI achieved similar results when radiologists were able to view prior CT scans. When it came to predicting the risk of cancer two years after a screening, the model was able to find cancer 9.5% more often compared to estimated radiologist performance laid out in the National Lung Screening Test (NLST) study. Detailed in research published today in Nature Medicine, the end-to-end deep learning model was used to predict whether a patient has lung cancer, generating a patient lung cancer malignancy risk score and identifying the location of the malignant tissue in the lungs.

CHICAGO --- Deep learning - a form of artificial intelligence - was able to detect malignant lung nodules on low-dose chest computed tomography (LDCT) scans with a performance meeting or exceeding that of expert radiologists, reports a new study from Google and Northwestern Medicine. This deep-learning system provides an automated image evaluation system to enhance the accuracy of early lung cancer diagnosis that could lead to earlier treatment. The deep-learning system was compared against radiologists on LDCTs for patients, some of whom had biopsy confirmed cancer within a year. In most comparisons, the model performed at or better than radiologists. Deep learning is a technique that teaches computers to learn by example.

CHICAGO --- Deep learning - a form of artificial intelligence - was able to detect malignant lung nodules on low-dose chest computed tomography (LDCT) scans with a performance meeting or exceeding that of expert radiologists, reports a new study from Google and Northwestern Medicine. This deep-learning system provides an automated image evaluation system to enhance the accuracy of early lung cancer diagnosis that could lead to earlier treatment. The deep-learning system was compared against radiologists on LDCTs for patients, some of whom had biopsy confirmed cancer within a year. In most comparisons, the model performed at or better than radiologists. Deep learning is a technique that teaches computers to learn by example.