A.I created 40,000 chemical weapons

 

Evil-oriented human intelligence and A.I can destroy humanity.

Take note: during a conference on unconventional weapons, researchers from the company “Collaboration Pharmaceuticals” showed an experiment where artificial intelligence machine learning technology created formulas for 40,000 biological weapons.

The experience was intended to show what A.I would be capable of and also to show that this resource, in the hands of dangerous groups and without supervision, would be treacherous.

In an interview given to "The Verge", Fabio Urbina, the study's primary author, spoke about how A.I managed to invent thousands of new substances - some frighteningly similar to the VX agent, an extremely powerful gas that attacks the nervous system of victims.

He explained that the study is a sort of “180° turn” from his normal work. On a daily basis, the scientist is tasked with researching machine learning models to discover new drugs and treatments.

However, it also involves implementing “evil” A.I models to ensure that any medication developed from their work does not have any toxic effects.

“For example”, he said, “imagine that you discover a wonderful pill that controls high blood pressure. But she does this by blocking some important channel connected to her heart. So this drug is automatically discarded because it is considered high risk.”.

The research was carried out at the invitation of the organization of the "Convergence" conference, held in Switzerland, and they asked that very technical information be kept secret for security reasons. What he told, however, traces an interesting procedural timeline:

“Basically, we have several historical databases on molecules tested for their toxicity or lack thereof,” Urbina said. "For this experiment, we focused on the molecular makeup of Agent VX, which acts as an inhibitor of something called 'Acetylcholinesterase'."

Acetylcholinesterase is, roughly speaking, an enzyme that acts in the transmission of information from the nervous system. When your brain gives an order, say, to bend your arm, this enzyme is what carries that command from point A to point B.

"The mortality of the VX lies in the fact that it prevents these commands from getting to where they are supposed to be if the order is anything muscle-related."

[VX] can stop your diaphragm or lung muscles, and your breathing literally stops and you suffocate.”

Based on this, Fabio Urbina and his team created a machine learning model that, roughly speaking, analyzed these databases, identified which parts of a molecule are toxic or not, and “learn” to glue molecules together, suggesting the creation of new chemical agents – this process uses an A.I either for good (the creation of new medicines) or for evil (the creation of chemical weapons and biological warfare agents).

So the team of scientists basically tweaked the A.I ​​to act like an “evil genius” and saw what it would do:

“We didn't know very well what would come out, since our capacity for generating models is formed by new technologies, that are not yet widely used”, explained Urbina.

“The first surprise was that many of the suggested compounds were far more toxic than VX, and that comes as a surprise because VX is one of the most toxic compounds out there, you need a very, very small dose to be lethal.”

A side note here: according to the US Center for Disease Control (CDC) page, VX is not “one of” the most lethal, but rather “the most” lethal of nerve agents.

The scientist explains that the models generated by the A.I ​​correspond to chemical weapons that have not been verified by the human hand – obviously, let’s face it – but normally these suggestions made by machine learning are quite solid. In other words, the error rate is low, and, given this perception, the application of this technology for the creation of lethal biological weapons is quite feasible.

The complete interview on the website "The Verge", gives other details, such as, for example, the fact that the machine learning model learned to create already known toxic compounds without ever having seen them in the database. Or yet, how this molecular model generation technology is so easily accessible that a simple Google search already puts anyone on the right path to program something like that (unfortunately).

Source: The Verge