Deep Learning Holds Promise for Much Longer Lives

By Russ Banham


The longest living human on record was Jeanne Calment, a French woman who died in 1997 at the age of 122. That’s a pretty sizable leap from today’s average life expectancy at birth in 2018 — 70 for males and 74 for females on a worldwide basis. But it’s peanuts compared to what may lie ahead.

Today, the possibility exists for people to achieve what scientists like Alex Zhavoronkov call “extreme longevity,” an age where people who pop off at 100 are mourned for dying too young.

We can thank (or blame) artificial intelligence for such long lifespans. As Zhavoronkov writes in “The Ageless Generation,” his new best-selling book, “In the not-too-distant future, medical science will possess the technology to slow and even reverse the aging process itself.”

Into the Deep

Zhavoronkov, who is the founder and CEO of Insilico Medicine, an AI and bioinformatics company focused exclusively on aging and age-related diseases, is keen to share the ways people can use emerging technologies to squeeze out quite a few extra years.

“By leveraging deep neural networks, we’re able to discover novel molecules targeting specific diseases and develop them into compounds for life-extending drugs,” he said. All of this, “at a vastly faster rate than is achievable in the current clinical trials process.”

A deep neural network is a form of deep learning, a subset of AI, that employs a multilayered system of artificial neurons to glean insights from massive amounts of data. The difference between artificial neurons and the ones we carry around in our brains is the skyrocketing speed by which the former acquire knowledge.

The data in this context relates to the aging process and the ability of different molecules to prevent or halt disease. By training the deep neural network to predict a particular person’s biological age based on genetic, environmental, and other relevant data, the specific reasons why the individual will likely live to be 80 years of age, but not much longer, are discerned. Once understood, the same tools are used to ferret out “correctives,” Zhavoronkov said.

These correctives are unique molecular compounds that offer a powerful opportunity to inhibit the manifestation of terminal diseases like cancer, Parkinson’s, or Alzheimer’s. By nipping such diseases in the bud, lifespans are extended. “Ideally, we want to prevent disease before it manifests,” Zhavoronkov said. “Several cancers have been cured in animals genetically similar to humans, substantially increasing the lifespans of worms, fruit flies, and mice. The problem is these compounds don’t translate well to humans.”

Deep neural networks narrow these odds. “We use massive repositories of what are called the `omics’ data, for genomics, proteomics, transcriptomics and metabolomics data,” he explained. “These biological molecules translate the structure, function, and dynamics of an organism, what we refer to as `gene expression’— a snapshot of all the genes.”

By training the deep neural networks to recognize the underlying signaling pathways in a disease, Zhavoronkov said, researchers can quickly identify the relevant biomarkers. From there, they can generate new molecular compounds with the properties that combat disease.

To get a sense of what he means by a “massive” repository of omics data, the total number of small molecules alone is estimated to be between 10⁶⁰ and 10²⁰⁰. For humans alone to study and test this many molecules as potential drug compounds would take an eternity — the equivalent of looking for a needle in a giant haystack. But AI, in this case deep neural learning, is really good at finding the needles.

“By mining the biological data, we hope to progress pretty quickly toward developing a chemical compound to hit the target and prevent the disease from manifesting,” he said. “Regenerative medicine is in a far more advanced state than most people realize. The pieces are coming together.”

Scratching the Surface

Although Zhavoronkov declined to estimate how long someone born today can expect to live, he’s confident we’ve not yet reached our biological lifespan limit.

2018 study published in the Science journal suggests that a fixed limit to the human lifespan has yet to be ascertained. The study indicates that people at age 100 have the same 50 percent chance of dying within a year as people between the ages of 105 and 109; they also share the opportunity to live another 1.5 years. In other words, the risk of death at extreme ages seems to plateau, giving hope that we’ve yet to reach an expiration date.

Some scientists like Aubrey De Grey, chief science officer at the SENS Research Foundation, project that human lifespans will extend into the hundreds and possibly in excess of 1,000 years. The thinking is that by the time someone born today makes it to 100 years, the state of AI and medical science will have advanced to the point where the person can conceivably live to the age of 150. Fifty years later, the same progression occurs, lifting the individual’s chances of making it to 200. And so on.

“I don’t like to speculate on how long we can hope to live, but I will say that human beings have a way of setting seemingly impossible goals like landing on the moon and then pulling it off,” Zhavoronkov said. “In that regard, I would aspire to live young as long as possible.”

A Dry Well?

Living to a ripe old age is not without complications. For one thing, the world could become as crowded as a Los Angeles highway at rush hour. Long-term care is another dilemma. In the past, it was common for children to take care of their aging parents. However, that was when the children were in their 50s and 60s and their parents were in their 80s and 90s — not 120-plus.

Another sobering consideration is how society will support masses of people living only half their lifespan by the time they retire. An analysis by the International Monetary Fund indicated that if people live a mere three years longer than expected, pension-related costs in both advanced and emerging economics could increase by 50 percent.

“Science is allowing us to live longer, but not necessarily with an improved quality of life,” said Robert Hartwig, a professor of finance at the University of South Carolina’s Darla Moore School of Business.

In Hartwig’s class on health and life insurance, he describes longevity risk as one of the greatest global challenges yet to be addressed in meaningful ways. “If you look at the age people generally retire today through the lens of their savings, pension, and Social Security income, and consider the possibility they may live decades and decades longer, you realize the well will soon be dry,” he said. “There are simply not enough economic resources available on the planet to allocate to hordes of people living to 120 or 150.”

Hartwig concludes the class by asking his Gen Z students when they expect to retire; the majority respond around the age of 55. “Working much longer than one anticipates is needed to offset a disaster in the making,” Hartwig warned.

But is it fair to expect a construction worker, truck driver, or other laborer to work well in their 80s and 90s? “Society needs to create job opportunities for individuals who historically would have transitioned out of the workplace and into retirement,” said Hartwig. “This could entail retirement at current expectations, followed by a period of reeducation and a less toilsome occupation.”

To incentivize people to work beyond their retirement years, Zhavoronkov suggested changing the age of eligibility to receive pensions and unpenalized access to 401k savings and Social Security benefits. “Although people will object,” he said, “in time, the new age of retirement will become accepted as the norm.”

In other words, the time to prepare for a very long life is now.