‘For several years now, scientists have been able to synthesise viruses from scratch,’ stated an article in Wired. ‘These are typically used to develop and evaluate antiviral drugs and vaccines, as well as to expand our understanding of how these micro-organisms behave.’
The pace at which science is progressing while our 24-7 media cycle obsesses about petty activities in Canberra is … astonishing.
And most of it is passing by unnoticed.
Or simply isn’t within the public domain.
Our increasingly cultural and social attention remains focused on what we perceive to be civilisation-altering policies, such as mass migration and tax reform. This is entirely understandable from the public’s perspective.
Although it is a big shift from the late 90s and early 2000s when we used to speak more regularly about the reach of science into the past and future. This was the era of Dolly the Sheep, Jurassic Park, and Bond villains tinkering in their labs. In some ways it was an evolution of the Cold War into the Silicon Suspicion.
Today, ‘science’ only seems to poke its head into bold font for a global pandemic, the declassified White House alien files, when the climate apocalypse mongers need a handout, or where biology collides with AI.
This article is about the latter.
We have reached the point where humanity has asked AI to create new ‘life’ (depending on how you define a virus).
The sheer volume of information that goes into understanding and recreating even fragments of life renders the task perfect for AI.
However, it is one thing for humanity to play God, but how do we feel about outsourcing this act of creation to an algorithm that has no conscience with which to judge its actions?
We are still mulling over the question about whether or not AI could reach a position of sentience, if it is a life of digital expression, and yet here we are possibly becoming grandparents to the next generation of artificial life that skipped over the natural order.
This is not so much a moral question of whether we should create life in this manner, or even a technical question of if we can – evidently that has been answered. My question is, have we properly thought about the unintended consequences? That’s more complicated.
What follows is an ethically complex story with tangible benefits and risks worth addressing.
News of AI creating viruses dramatically broke around the same time Dr Anthony Fauci was shifting awkwardly in front of the cameras, pleading The Fifth to everything, including the colour of his tie. The once-friendly face of Covid has re-emerged in the papers like the exposed bones of a shipwreck leftover from some mysterious ancient cataclysm, half-forgotten by the world. The oh, it was just a bat and a wet market! story has rotted away leaving urgent questions surrounding gain-of-function research, the appropriate use of taxpayer funds, and a shocking lack of transparency from China.
On the balance of probability, many have claimed that Covid was an unintended consequence of humans exploring the edges of science before it was properly understood.
The result of this lack of foresight was a global pandemic, seven million deaths, innumerable injuries, the crash of economic systems, poverty, social discord, a loss of trust across institutions, and a massive financial windfall for pharmaceutical companies and global health authorities.
‘Unimpressed’ doesn’t quite cover public sentiment.
This is the wider context of where ‘science’ is at in the public mind. It is not a comment, in any way, on what goes on in other research facilities.
Indeed, the study involving AI, which was undertaken by Stanford University and the Arc Institute, was aiming to solve real-world problems, such as the rising issue of resistant bacteria.
These pose an increasingly urgent threat in hospitals and the medical world is running out of solutions.
To quote Wired again (as they provided an excellent summary), ‘The results showed that AI-generated viruses were able to rapidly overcome bacterial resistance and establish infection. According to the authors [of the study], this finding demonstrates “a path toward artificial intelligence-generated phage therapies against rapidly evolving bacterial pathogens”.’
The details of this study were published by science.org under the title: Generative design of bacteriophages with genome language models.
AI assisted in the creation of 16 viruses.
From the editor’s summary:
The ability to design complex biological systems with artificial intelligence (AI) has the potential to transform biotechnology, but progress has largely been limited to the scale of individual genes and proteins, with whole-genome design remaining out of reach. King et al. used generative AI models trained on millions of natural genomes to design entire bacteriophages (see the Perspective by Inglesby and Hanke). Experimental tests yielded 16 functional genomes with diverse sequences, structures, and fitness profiles. A cocktail of the generated bacteriophages rapidly overcame bacteria that had evolved resistance to natural bacteriophage. This work lays a foundation for AI-guided design of biological function at the whole-genome scale.
The study goes on to define these genome language models as ‘AI algorithms that have shown promise in designing biological systems’.
Bacteriophages are described as viruses that infect bacteria but are presumed to be incapable of impacting a human being.
In order to achieve this, AI had to be trained. Just as language models are consuming vast libraries of books to learn how to speak, these AI models (Evo1 and Evo2) were fed DNA from ‘millions of genomes from all domains of life’. This was done with restrictions to ensure they were not given access to dangerous data.
Essentially, they gave AI the blueprints of life.
This is the domain of synthetic biology.
The viruses created by AI were not trapped in the digital world as a theoretical possibility, but rather put into Petri dishes with a layer of bacteria to see what they would do. One of the researchers described how the lab broke into applause when it became clear their AI-created virus was eating the bacteria.
Scientifically, it was a remarkable success.
The AI creations ‘showed strong host specificity and diverse fitness profiles, including competitive infection kinetics’ and that these artificially created phages ‘were different from any known natural phages, exhibiting de novo mutations, divergent genes, and regulatory elements, and variable genome lengths’.
Originally, there were 300 genomes synthesised, but only 16 were ultimately successful. Human-only replications of this would have taken a prohibitively long time.
Describing their desire for AI to predict and write genetic code, one of the authors said:
‘In this case, we wanted the model to generate the entire genome end-to-end in a single left-to-right pass. We didn’t add anything.’
These experiments were designed with positive benefits in mind. Of that, there is no question. Their health-related goals are made explicit in the discussion and documentation.
Still, individuals across the industry have warned that governments lack the ability to regulate ahead of the rapidly developing technology and there is nothing stopping AI being used by malicious actors to create the next pandemic.
While some researchers use AI to save lives, others could just as easily use it for the opposite. Instead of feeding safe data, they could train AI on the most dangerous viruses available.
The same problem has been known for many years when it comes to gain-of-function research. Those conversations were never properly answered before Covid struck, and they remain worryingly open from a regulatory perspective.
Even if proper safeguards could be put in place to stop bad actors from dabbling in this Pandora’s Box, there would be no realistic way to prevent a hostile regime from deliberately using AI to create a devastating synthetic virus. The fear is that AI could accelerate an existing concern about human-created life.
This is the eternal tragedy of science and knowledge.
Setting aside the worst-case scenario, there is the other problem of unintended consequences where, perhaps, even a good-intentioned viral creation could go wrong. It certainly wouldn’t be the first time.
Doctors at the John Hopkins Centre for Health Security said: ‘The ability to compose viral genomes using generative AI now exists; the governance to safely steer it does not.’
This AI-assisted technology was foretold and, in the intervening years, no solution to contain it has been found, just as we never worked out what to do about cloning, gene splicing, or gain-of-function.
And we cannot forget that no matter how successful the AI-generated virus is at solving the problem researchers assign to it, evolution is a constant arms race. There will be a natural response. An equal force pushing in the opposite direction operating outside the lab and beyond the view of researchers. These superior AI creations could trigger a reaction from bacteria, just as they became resistant to our sterilisation processes. We have no way to know how the real world will react.
At least when it came to creating nuclear weapons and nuclear energy, two sides of an extraordinary coin, we were competing against ourselves. Competing against nature itself is frightening.
For the moment, my person response to AI meddling in viral creation is: Thanks, but no thanks!
Alexandra Marshall is an independent writer. If you would like to support her work, shout her a coffee over at donor-box.
The post AI gives us 16 new viruses appeared first on The Spectator Australia.
