Artificial intelligence predicts death: It is not the god of death, but helps humans live longer

Artificial intelligence predicts death: It is not the god of death, but helps humans live longer
Do not go gentle into that good night

Old age should burn and roar at sunset

Rage

Rage

The dying of light

The famous lines by Welsh poet Dylan Thomas are a powerful description of the inevitability of death, and while the sentiment is poetic, the reality is far less simple.

We will all die someday, at a certain time and place, and it may remain a mystery to us until the mystery is revealed, or maybe not. Researchers are now applying artificial intelligence, specifically machine learning and computer vision, to predict when someone will die. The ultimate goal is not to play the role of the Grim Reaper, like a scary sci-fi machine in the Death Universe, but to treat or even prevent chronic diseases and other illnesses.

A recent study on artificial intelligence for precision medicine used an off-the-shelf machine learning platform to analyse 48 chest CT scans. The computer was able to predict which patients would die within five years with 69 per cent accuracy. The article was published in Scientific Reports by a team led by the University of Adelaide.

In our email interview with Singularity Hub, radiologist and PhD student Okedena-Reina said one obvious benefit of using AI in precision medicine is earlier detection of health risks and the potential for intervention. Less obvious, he added, is that it is a foundation for accelerating longevity research. "Currently, most studies of chronic disease and longevity require long periods of follow-up to detect differences between patients with treatment and those without, because disease progression is so slow," he explained. "If we can quantify these changes earlier, not only can we identify disease and intervene more effectively, but we can also detect treatment responses earlier, which could lead to faster and cheaper treatments, he added. "If we can shave one to two years off the time between lab and patient, it would greatly accelerate progress in this field."

Artificial intelligence can also have a human heart

In January, researchers at Imperial College London published results of a study suggesting that artificial intelligence can predict heart failure and death better than human doctors. The study, published in the journal Radiology, involved about 250 patients who had virtual 3D hearts built to simulate their heart function. The AI ​​algorithm then began to figure out which features could serve as the best predictors. The system relied on MRIs, blood tests and other data for analysis.

In the end, the machine did a faster and better job of assessing risk for pulmonary hypertension, about 73 percent versus 60 percent. The researchers say the technology could be used to predict outcomes for other heart conditions in the future. "We want to develop this technology so it can be used across many heart conditions to supplement how doctors interpret the results of medical tests," study co-author Tim Dawes said in a press release. "Our goal is to see if better predictions can guide treatment and help people live longer."

Artificial intelligence is getting smarter

These kinds of applications will get better and better as the machines continue to learn, just like any student in medical school does.

Ochdener-Reina said his team is still working on its ideal data sample but has improved prediction accuracy by 75% to 80% by including information such as age and gender.

"I think there's an upper limit to our accuracy because there's always an element of randomness," he responded, saying that AI will be able to accurately pinpoint an individual's mortality rate, but we can be much more precise than we are now, taking into account each person's risks and strengths, and a model that combines all of these factors could hopefully predict 80% of the risk of near-term death.

Others are more optimistic about how much AI will transform the medical field, and predicting how long people will live is actually one of the simplest applications of machine learning.

Dr Ziad Obermayer told the bureau he needed a unique set of data linked to the electronic records left when people died, but once we had enough people, you could accurately predict whether someone was likely to be alive a month from now, or a year from now.

But AI still has a lot to learn

Experts like Obermeyer and Ochden-Rayner agree that progress will come quickly, but there is still a lot of work to be done. On the one hand, there is a lot of data to mine, but some things are still not straightened out. For example, the images needed to train machines still need to be processed to make them useful.

"Many organizations around the world are now spending millions of dollars on this task as it appears to be the main bottleneck to the success of AI healthcare," Oakden-Rayner said. In an interview with Statistics Canada, Obermeyer said data is fragmented throughout the healthcare system, so connecting information and creating comprehensive databases takes time and money.

He also noted that while there is a lot of excitement about the use of AI in precision medicine, there has been very little activity testing these algorithms in clinical settings, saying, “It’s all well and good to say you have an algorithm that’s good at prediction. Now let’s move them into the real world in a safe, responsible, ethical way and see what happens.”

Besides, preventing deadly diseases is one thing, but using artificial intelligence to prevent fatal accidents? That’s exactly what researchers from the United States and India set out to do when they investigated the disturbingly high death rate when taking selfies.

The team identified 127 deaths caused by selfies over a two-year period. Based on a combination of text, image and location, the machine learned to identify whether a selfie was potentially dangerous. The software collected more than 3,000 annotated selfies on Twitter and achieved an accuracy rate of 73%. "The combination of image-based and location-based features was the most accurate," they report.

What’s next for AI?

A selfie early warning system: One direction we are working on is to have the camera give the user a sense of whether a particular location is dangerous and assign a score to it, said Ponnurangam Kumaraguru, a professor at Indraprastha Institute of Information Technology in Delhi, in a report by Digital Trends.

This discussion begs the question: Do we really want to know when we're going to die? According to an article published earlier this year in the journal Psychology Today, the answer is a resounding "no."

In Germany and Spain, nearly nine in 10 people asked if they wanted to know about their future, including their death, said they would prefer to remain ignorant. But Obermeyer thinks the situation is different, at least when it comes to people with life-threatening illnesses. “What those patients really want is not an objective prediction from their doctors about how long they are going to live,” he told Marketplace public radio.

“Doctors are very reluctant to answer these questions, partly because, you know, you don’t want to be wrong about something so important.” But also partly because patients don’t want to know.

From SingularityHub

From: NetEase Intelligence

<<:  Science and Technology Review: Thoughts on the process of automobile intelligence and its key technologies

>>:  How much do you know about iPhone 6S?

Recommend

Promotion | Introduction to Fanstong Products

The first in a series of popular science articles...

Ogilvy's internal training on advertising creativity

In "Creativity is just incomprehensible? In ...

I am a flasher, not an experienced driver!

I'm a brush machine dealer. Compared with the...

App monetization methods and advertising monetization pain points!

If an app does not have a good commercial monetiz...

Nanny-level tutorial on Tik Tok marketing!

Today I bring you a nanny-level teaching course o...

Optimization strategy for B station video native advertising

As a cultural community and video platform where ...

What are some low-cost ways to acquire customers for your products?

The secret to success on the Internet is to be ab...

Don't buy these 6 kinds of food in the supermarket! You've been cheated

Whole grains may be fake whole grains. Since my c...

6 tips to quickly improve UI design effects

Editor's note: When designing a UI, there are...

21 questions and answers about Android View event mechanism

[[178693]] 1.What are the main coordinate paramet...

iOS A Song of Ice and Fire – Using XPC to pass the sandbox

0x00 Sequence Ice refers to user state, and fire ...

Sina Fuyi advertising display and billing model!

We know that there is a huge difference between S...