Artificial intelligence reveals surprising patterns in Earth’s mass biological extinction

The idea that mass extinctions allow many new species to evolve is a central concept in evolution, but a new study using artificial intelligence to study fossils has shown that this is rarely true and there must be another explanation, scientists write in a new study published in the scientific journal Eurekalert.

Charles Darwin’s landmark work On the Origin of Species ends with a beautiful summary of his theory of evolution: “In this notion that life with its several faculties was originally breathed into several forms or into one, and that, while this planet cycles through the unchanging law of gravity, infinite forms, the most beautiful and most amazing, have developed and evolved from such a simple beginning.” In fact, scientists now know that most species that ever existed are extinct.

In general, throughout Earth’s history, species extinctions have been roughly balanced by the emergence of new ones, with a few major temporary imbalances, which scientists call mass extinctions. Scientists have long believed that mass extinctions create productive periods of species evolution, or “radiations”-a pattern called “creative destruction.”

A new study by scientists at the Earth Life Science Institute (ELSI) at Tokyo Institute of Technology examined the joint recurrence of fossil species using machine learning and found that radiations and extinctions are rarely linked, and thus mass extinctions probably rarely cause radiations on a comparable scale.

Constructive destruction is central to classical concepts of evolution. It seems obvious that there are periods when many species suddenly disappear and many new species suddenly appear.

However, radiations comparable in magnitude to mass extinctions, which are called mass extinctions in this study, have received much less analysis than extinctions. This study compares the effects of both extinctions and radiation during the period for which fossils are available, the so-called Phanerozoic Eon. The Phanerozoic (from the Greek for “apparent life”) represents the most recent ~550-million-year period of the total ~4.5 billion years of Earth history and is important to paleontologists: before this period, most organisms that existed were microbes that could not easily form fossils, so the prior evolutionary history is difficult to trace.

The new study suggests that creative destruction is not a good description of how species emerged or died out in the Phanerozoic, and suggests that many of the most prominent periods of evolutionary radiation occurred when life entered new evolutionary and ecological realms, such as during the Cambrian explosion of animal diversity and the Carboniferous expansion of forest biomes. Whether this is true for the previous ~3 billion years, when microbes dominated, is unknown, since the paucity of recorded information on such ancient diversity did not allow for a similar analysis.

Paleontologists have identified several of the most important mass extinction events in the Phanerozoic fossil record. These include basically five major mass extinctions, such as the mass extinction at the end of the Permian, when it is estimated that more than 70% of species became extinct.

Biologists have suggested that we may now be entering a period of “sixth mass extinction,” which they believe is mainly caused by human activities, including hunting and land-use changes caused by agricultural development.

A well-known example of previous “Big Five” mass extinctions is the Cretaceous-Tertiary Extinction (usually abbreviated as “K-T,” using the German spelling Cretaceous), which appears to have been caused by a meteorite impact on Earth ~65 million years ago that wiped out the non-African dinosaurs.

By observing the fossil record, scientists have concluded that mass extinctions lead to the emergence of particularly productive life forms. For example, it is commonly believed that the mass extinction of dinosaurs during K-T created a gap that allowed organisms such as mammals to repopulate and “radiate,” enabling the evolution of all sorts of new mammal species, ultimately laying the foundation for the emergence of humans.

In other words, if the event of “creative destruction” had not occurred perhaps we would not be having this discussion.

The new research began with a casual conversation in the Agora, a large common room where ELSI scientists and visitors often dine and strike up new conversations. The two authors of the paper, evolutionary biologist Jennifer Hoyal Cuthill (now a researcher at the University of Essex in the UK) and physicist/expert in machine learning Nicholas Guttenberg (now a research associate at Cross Labs working with GoodAI in the Czech Republic), who were postdoctoral fellows at ELSI when the paper began, were discussing whether machine learning could be used to visualize and understand fossils. During their visit to ELSI, shortly before the COVID-19 pandemic began limiting international travel, they worked feverishly to expand their analysis to examine the correlation between extinction and radiation events. These discussions allowed them to relate their new data to the breadth of existing understandings of mass extinctions and radiation.

They quickly found that the evolutionary patterns revealed by machine learning differed from traditional interpretations in key respects.

The team used a novel application of machine learning to examine the temporal co-occurrence of species in the Phanerozoic fossil record, examining more than one million records in a massive database of nearly two hundred thousand species.

Lead author of the study Dr. Hoyal Cuthill said: “Some of the most challenging aspects of understanding the history of life are the enormous time scales and number of species. New machine learning applications can help us by allowing us to visualize this information in a human-readable form. This means we can, so to speak, hold half a billion years of evolution in the palms of our hands and gain new insights from what we see.”

Using their objective methods, they found that the “big five” mass extinctions previously identified by paleontologists were assigned by machine learning methods to 5% significant disturbances in which extinctions outpaced radiation or vice versa, as well as seven additional mass extinctions, two combined mass extinction-radiation and fifteen mass radiations. Surprisingly, in contrast to previous descriptions emphasizing the importance of radiation after extinction, this work showed that the most comparable mass radiations and extinctions only rarely coincided in time, refuting the idea of a causal relationship between them.

Dr. Nicholas Guttenberg, co-author of the paper, said: “The ecosystem is dynamic; you don’t have to break off an existing piece for something new to emerge.”

The team went on to find that radiation can actually cause major changes in existing ecosystems, an idea the authors call “destructive creation.” They found that, on average in the Phanerozoic Eon, the species that made up the ecosystem at any given time had almost all disappeared after 19 million years. But when mass extinctions or radiations occur, the rate of species change is much higher.

This provides a new perspective on how the modern “Sixth Extinction” occurs. The Quaternary period, which began 2.5 million years ago, witnessed repeated climatic upheavals, including abrupt glacial shifts when high-latitude areas of the Earth were covered by ice. This means that the current “Sixth Extinction” is destroying biodiversity that has already been disrupted, and the authors suggest that it will take at least 8 million years for it to return to its long-term average of 19 million years.

Dr. Hoyal Cuthill comments that “every extinction that happens before our eyes destroys a species that may have existed for millions of years before that point, making it difficult for the normal process of ‘new species’ to emerge to replace what has been lost.”

Subscribe
Notify of
guest

0 Comments
Inline Feedbacks
View all comments
0
Would love your thoughts, please comment.x
()
x