Summary of Key Points
A team from Stanford University used the AI tool Evo 2 to create 16 new viruses that do not exist in nature, using known phages as templates. These viruses are capable of killing only bacteria. While this technology could help humanity combat superbugs, it also exposes gaps in global biosecurity regulations. It is highly unlikely that ordinary people will be able to replicate this process. In the future, this technology could potentially be extended to designing more complex forms of life (such as bacteria), but it is not expected to trigger a “biochemical crisis” in the short term.
Detailed Explanation
The “Three-Step Magic” of AI-Designed Viruses
AI does not create viruses out of thin air; rather, it acts like a “advanced tailor”:
- Step 1: Selecting a Template: They chose the φX174 phage, one of the smallest viruses known, with only 5,386 base pairs, which is well-understood by scientists, as the base material for their designs.
- Step 2: Providing Instructions: They input combinations of A/T/C/G (the building blocks of DNA) as design specifications, and the AI generated 700,000 potential gene sequences.
- Step 3: Testing and Selection: From these 700,000 sequences, they selected 302 for further testing, resulting in 16 new viruses. The similarity of these viruses to naturally occurring ones ranges from 93.2% to 98.8%, which is greater than the difference between a cat and a tiger (95.6%), indicating that they are essentially new species.
A Critical Difference: Previous attempts to modify existing genes focused on making small changes, whereas this approach creates completely new, replicable systems, akin to designing entirely new clothing styles rather than altering existing ones.
A Lifesaver Against Superbugs
Antibiotic-resistant bacteria pose a major global health threat:
- They directly cause over 1 million deaths each year, more than the combined toll from malaria and AIDS. In China alone, 145,000 deaths per year are attributed to antibiotic resistance.
- Pharmaceutical companies are reluctant to invest in developing new antibiotics due to the long development cycle and low profits, leading to a vicious cycle where more resistant bacteria emerge, making existing drugs less effective, requiring higher doses, and thus further fueling resistance.
AI-designed phages that target bacteria represent a breakthrough:
- Phages only attack specific bacteria and do not harm human cells. However, natural phages are scarce, and bacteria can easily develop resistance to them.
- AI can design multiple phages simultaneously, which can be used in combination (a “cocktail therapy”) to make it difficult for bacteria to develop immunity. Experiments have shown that the 16 newly designed phages effectively eliminated E. coli strains resistant to the original phage. This technology could potentially be used against superbugs such as MRSA (methicillin-resistant Staphylococcus aureus) and drug-resistant tuberculosis.
Is a “Biochemical Crisis” Near?
- No immediate threat: The success rate of AI-designed viruses is very low, at only 5%, and even for experts using this technology, the success rate is less than 8%.
- The complexity of the genetic material increases significantly; for example, the genome of the COVID-19 virus is six times more complex than those used in these experiments, potentially increasing the difficulty by several orders of magnitude.
- Data Safety: AI algorithms are trained to exclude pathogens that infect humans or animals, so the viruses created in this study only kill bacteria and are harmless to humans.
Long-Term Risks to Be Aware Of:
- The barrier to creating harmful viruses with AI has lowered: Malicious actors can use existing knowledge of pathogens to develop dangerous organisms without access to advanced laboratories.
- Lack of Regulation: There are no universal standards for DNA synthesis, leaving a significant security gap in global biosecurity.
Is It Possible for Ordinary People to Create Viruses?
Don’t be misled by claims of equal opportunities with AI; there are several major obstacles:
- Technical Barriers: The AI generates base sequences, but synthesizing and culturing the viruses requires specialized knowledge in molecular biology and experimental skills.
- Cost: Creating a virus like φX174 costs tens of thousands to hundreds of thousands of dollars; larger viruses (like those related to COVID-19) could cost millions.
- Skill Gap: Experts in AI, virology, and laboratory techniques are in short supply worldwide.
- Laboratory Requirements: Specialized BSL-2 (Biosecurity Level 2) laboratories are necessary for this process.
Unless you are a “genius with access to resources,” creating viruses using AI is practically impossible.
The Next Step for This Technology
AI-designed viruses are just the beginning. The Evo 2 model can already handle more complex genetic sequences:
- Human mitochondrial DNA (16,000 base pairs), yeast chromosomes (330,000 base pairs), and simplified bacterial genomes (580,000 base pairs—100 times larger than those used in this study) have all been modeled, though they have not yet been tested in laboratories.
- In the future, it might be possible to design bacteria that can detect toxins in water, “safe microorganisms” that can only survive in laboratory conditions, or engineered bacteria that can absorb carbon dioxide.
However, the main challenge remains: the current accuracy of AI-generated genomes is only 70%. Raising this rate to 100% is essential to avoid creating unexpected and potentially dangerous organisms.
In Conclusion
AI-designed viruses are a promising tool in the fight against superbugs, but effective regulation is necessary. Technology itself is neutral; it is human intentions and existing security gaps that pose real threats. While a biochemical crisis is unlikely in the short term, we must strengthen our biosecurity measures over the long term.