A new study on the "science" alert of the risk of protein designed with IA

Research shows how biosecurity systems fail to detect dangerous protein versions that generate artificial information, which opens the door to malicious, or unintended use, which may have catastrophic consequences.

A new study on the theme of "subject" of proteins developed with AI

Studies show that biosefetic systems can lead to artificial or accidental results that can lead to catastrophic or accidental use, to lead to catastrophic consequences

More than one hundred scientific that signed to try to create biological weapons with ai

But what if the next Pondi produces fake wisdom?

This function is only for registered

Confirmed to join what is the most intelligence method for the affected item ...

However, in this area you could draw new carcass lights, creating the world.

The global scientific community is at the intersection, which at the same time is fun and horrible.Thus, while the IA is in motion and more famous in biological creation, human protection, designed to preserve this power, is dangerous.Overcame his guardians.

Studies, the result of an unprecedented practice of the Red AI team (AI's Red Team), which has collected technology giants such as Microsoft, DNA synthesis industry (including a biological company for twist) and bio -related experts, confirmed that many are scared: the system to detect.

And it is not a simple theoretical danger, but a critical vulnerability that has been proven properly in a controlled environment.Counter -workers: Although the software "worked almost perfectly" to detect the original rows of dangerous protein, their ability to identify the new variants was subtly reformulated by AI, "inconsistent" and failed significantly.

However, despite the rapid repairs applied by BSS suppliers in collaboration with investigators by BSS suppliers, about 3% of versions are more likely to protect the deadly functionality they still manage to identify.This is encouraging, but the responsibility of the new and risk of reflexive monitoring and management."

AI, life architect '

To understand how far the risks have reached, it is enough to attract simple attention to the extraordinary energy that the proteins of AI have granted, briefly, the 'molecular machine' of life, creators, catalysts, messengers and the protector of our body.The architecture, which determines its effectiveness: allows a muscle cell to deal with the DNA cut and pass or neutralize an antibody antibody.

For decades, the protein device was based on nature and changed them slightly, just like an easy-to-smart.

Deeply based tools, in fact, such as Alapold, developed by Deepmind (Google) and Rosettafold, of the Protein Design Institute of the University of Washington, led by computational biophysicist David Baker, has awarded the Nobel Prize in Chemistry in 2024 (together with Demis Hassabis and John Jumper).And instead of changing what exists, from now on it will allow the AI ​​to design, again, proteins with a structure or function we want, also that they have not been before in a famous living creature before.

He was educated with almost all recognized protein structures, new AI models have learned 'grammar' chemical and biological proteins.And as a high -tongue model, such as ChatGPT, which is trained with billions of words, the coherent and original text can be generated, systems such as proteins or ESM3 -with billions of protein sequence.Molecules.

Now think of a large designation of a large space of amino acids with protein.And remember, there are only 100 amino acids protein, as many possible species that appear in the universe.When it creates the order of the Amino acids, the ability to leave and produce roads, more stable temperatures, which can be a particular stable temperature, or a certain enzyme.The Superclptic Center for a life scroll of life in the sale of a person's science links to a person's artificial scriptures.

Booting Plinyck

The design is a digital creation, but Askeka is the main surveillance of the new 'The Askieaana,' Bostgleneck 'or Bastetone or Protein Engineer.AI.

Here, in fact, this is where the Biosecurity filtering system (BSS) joined in.This is a program that acts as a 'digital sentinel' which undergos every DNA sequence to ensure that the desired genes do not encode 'worrying' protein, that is, toxins, virulence factors or harmful pathogenic genes that can be used for evil or terrorist purposes.

In time, these systems are strong, so most of the orders will take part in the recognition.But it is, who can be able to create unpublished and indicated parts of its weakness.

“Red equipment” and broken armor

The scientific study used the technique of "Red Team" (Red Team), a concept taken by cybersecurity with a simple metaphor by Alfonso Valencia.We know."

The biological field is used to generate the boycolicil operation to be known in the organic area.It is designed to become the order of real protein (wild), but from the three-dimensional structure and such molecule function, I.E.They will be dangerous.

And there, BSS, which is optimized to identify the same methods and sequences of the pathogen code), which allows their synthetic data. "

Is racial biological weapons?

However, the main focus of the scientific community is not on randomizing harmful proteins (although certainly a danger), but rather the possibility that some destructive actors can use artificial intelligence to design and order biological factors with optimal properties.

As Alfonso Valencia warns, the first thing to think is "a more contagious variant of a coronavirus."So there is a danger whether toxin can not only be created, but it can be designed 'ex novo' to make it more stable, more degraded or more difficult to detect with immune systems.'Superprotein', passes the digital screening, disappears the barrier between what is just a digital design and the physical presentation of the same protein.

This concern has long been a major theme of synthetic biology.In fact, David Baker himself, along with renowned biotechnologist George Church (Pioneer of Genetic Engineering), has already warned these dangers in January 2024, the Editor of Science

Software software

After vulnerable expression, Microsoft and his colleagues must work with the biggest BBS suppliers.An article that has not yet been resolved.

The system's weakness has led the author has been decided unusual and a code of biosafety and biocoly's education code for science.Okay to ask for access, identify and give reasons on their use, depending on the evaluation panel.

Always Alfonssio, This means to create, the safety of the Work for Mexication.

The result of the study of the future: the overview of the General of the General did not justify the maximum of new prostitutes.

Oh, a mirror where you can see the future of medical and sustainable development, put almost as much power in our hands as creative.Now, the most urgent task is not to continue creating excellent molecular machines, but to create an ethical system and maintenance that guarantees that this new "lamp genius" is only reminiscent of improving health and not increasing.Destruction.The carrier has just entered a new and dangerous era between scientific progress and safety.

Convention

- access to the content of premium -class is open from the courtesy of installation, but now many users are connected at the same time.Please try it please.

Reply to the limits of the meeting

- You start only three sessions at the same time.We close a lot of ships to continue without any restrictions.

This feature is only for subscribers

This function is only for registration.

Sewer systems are only open to subscribers.