AI can be used for the manufacture of toxin ricin, but this can also be obtained from castor beans found in many gardens
American photo archive/Alamy
Artificial intelligence Promises to transform biology, allowing us to develop the best medicines, vaccines and even synthetic organisms, say, for waste use. But some are afraid that this can also be used for darker purposes to create bio -uapones that will not be detected by ordinary methods until it became too late. So, how should we worry?
“Achievements in the field of AI feed breakthroughs in the field of biology and medicine,” says Eric HorvitzThe main researcher at Microsoft. “The new authority is responsible for vigilance.”
His team published a study in which the question is considered whether the proteins can design the same as the proteins, which, as you know, are dangerous, but are quite different so that they are not recognized as dangerous. The team did not reveal what proteins they tried to change the design – parts of the study were withheld – but probably included toxins, such as Ricin, known in the murder of 1978, and botulinum, powerful neurotoxin, better known as Botox.
To make a lot of protein, such as botulinum, you need a recipe – DNA This encodes for this. When biologists want a certain piece of DNA, they usually order it from companies that specialize in creating any desired part.
Due to the fears that potential bioterrorists can order recipes for creating biopones in this way, some DNA synthesis companies voluntarily check the orders to check whether someone is trying to do something dangerous. Squirrels are sequences of amino acids, and screening checks whether the amino acid sequence of any “sequences that cause concern” corresponds, that is, potential bio -apones.
But theoretically it is possible to develop a version of a protein that has another amino acid sequence, but still does the same. Horvits and his colleagues tried to do this with 72 potentially dangerous proteins and showed that screening methods often miss these alternative versions.
This is not as anxious as it seems. Firstly, the team actually did not make processed proteins for obvious reasons. But in a separate study at the beginning of this year, they checked the processed versions of harmless proteins – and mainly discovered that they do not workField
Secondly, while attempts were made by biocorrorism attacks, although very few, there is no reason to think that this is due to the failure system of voluntary scanning. There are already many ways to get around this without resorting to red -haired AI – for example, Ricin can be obtained from castor oil plantsFound in many gardens. This study is the equivalent of a warning that the bank can be robbed by some very difficult Mission is impossibleThe plan of the style, when actually the door of the storage was widely opened.
And the last, but no less important, when state actors are excluded, not a single bioterrorist has never managed to kill anyone, using Bi-apones based on protein. The cult of Aum Shinrikyo in Japan tried to kill people with botulinum, but changed only chemical agents. Letters with Ricin, sent to the White House, did not kill anyone. Based on the amount of the body, weapons and explosives are much more dangerous than bioxins.
Does this mean that we stop worrying about the bioapon developed by AI? Not really. While the Horvitsa’s research looked only at the proteins, these are viruses that pose a great threat – and AI is already used to process everything VirusesField
Last month, the Stanford University team in California revealed the results Their efforts to recalculate the virus What infects the bacterium E. KolyaThe field, as in the case of recycled proteins, the results were not impressive-of 302 developed AI, which were made, only 16 could infect E. KolyaThe field but this is only the beginning.
When he was asked about the viruses developed by AI, James Digganes In the proprietary company DNA Firm Twist Bioscience and a member of the Horvitsa team, it was easier to detect viruses encoding DNA than proteins that cause concern. “Synthesis screening works better for more information, not less. Therefore, on the scale of the genome, this is incredibly informative. ”
But not all DNA production companies perform this screening, and DNA synthesizers become affordable. There are talk about the development of AI tools that refuse to create dangerous viruses or try to find evil intentions, but people have found many ways to circumvent a guarantee, for example, to stop Ice providing instructions for the manufacture of bombsField
To be clear, history assumes that the risk from the “wild” viruses is much higher than the risk of bioterrorism. Despite the fact that the current US administration claims, the data indicate that Sars-No-2 appeared When the bathing virus jumped to other wild animals, and then on the markets in the market, neither the laboratory is involved.
Moreover, potential biorerics could cause incredible damage simply Release of a well -known virus, such as OspaA field with many gaping holes in the efforts to combat the bioapon, there is no need to resort to the cunning of AI in order to get around them.
For all these reasons, the risk of an untied AI developer at any time is probably closer to zero. But this risk will grow like Various technologies continue to advance -And Covid-19 pandemia showed how much chaos can create a new virus, even if it is not particularly fatal. More and more often it will be the reason to worry.
Topics:
