This is part IV of a “long-running” series of articles that I publish when “science” catches up to coming trends, in particular, we indulge in “Avanta-Garde” science in these articles. Reading parts I, II and III are not obligatory, they provide a more profound grasp of the subjects and their rapid progression. Part II, in particular, comes highly recommended.
One of the main aspects of this series has been the use of cutting-edge technology in science, medicine, and biology. In the last few years, many researchers were already achieving unthinkable results back in 2018, creating something akin to a big dataset of proteins and what's even more impressive is their ability to manipulate these proteins, generating novel variations, all without relying on AI. Understanding the intricate mechanics of protein functionality has long been a major goal in Biology.
Uncovering the hidden secrets of proteins makeup - the combination of peptides and amino acids to create a protein, protein structure - their shapes which influence their function as much as the amino acids that make them, protein interaction - how proteins interact with each other, elicit signals and create cascade effects, are all important steps paramount to solving many of life’s mysteries and especially diseases. In 2020 Google’s AI company DeepMind launched the AI-based software “AlphaFold”, aimed at predicting how proteins fold.
Understanding these pivotal steps leads to the insight of understanding how proteins interact with each other, and how they bind (glue) together. In the era preceding AI integration, achieving success in these endeavors entailed a lengthy timeline spanning several months, with a success rate hovering at a mere 1%. The advent of AI, however, has heralded a transformative shift. Success rates have surged to an impressive 50% or more, accomplished within significantly reduced timeframes.
Grasping the precise molecular regions where an antibody engages with a specific section of another protein or pathogenic agent, leads the designed product to effectively inhibit deleterious processes. This understanding serves as the cornerstone for the design of novel drugs and therapeutic interventions. Such is IgG4 structure and how it can be used to treat many inflammatory diseases and certain…hidden pathogens. (this is called foreshadowing)
Our brains are one of the most energy-efficient supercomputers in nature, it takes an unbelievable about of hardware to equate what our brains do both consciously, and especially unconsciously, therefore does computer chips built-in human brain tissue sound too futuristic ? To further blur the lines between science fiction and reality, and perhaps evoke a certain level of skepticism from you, how do you react to robots made of living tissue ? Here comes the Xenobots.
What’s fascinating about Bongard’s xenobots is that they can be made from normal cells taken from frog embryos – no genetic tweaks required.
Although scientists already knew these cells could move on their own, in this case they’re being used as materials to generate predictable, robot-like behaviours, such as herding particles around a Petri dish, cooperating like sheepdogs and even birthing balls of other cells that might be regarded as xenobot babies.
So if you’re going to make a xenobot, where do you start? Well, the Vermont team starts in a virtual Petri dish, on a computer, where an artificial intelligence (AI) program ‘evolves’ bunches of frog cells, based on their shape, to perform whatever task it is the scientists are interested in.
“It creates a population of virtual xenobots, deletes the ones that do a poor job and makes randomly modified copies of the survivors,” explains Bongard.
The scientists tell the AI how many rounds of this artificial selection process to complete and in just a few seconds, they have their design.
As an example, that design might be a ball of cells with a hole in the middle, like a pouch, which works well for transporting objects.
It’s the AI-based design process that’s the “real masterpiece” of the team’s approach, according to Dr Falk Tauber, a bio-inspired technology expert based at the University of Freiberg in Germany.
Without the virtual Petri dish, he notes, testing hundreds of different cell configurations could take weeks or even months using real cells.
Since the launch of ChatGPT 3.5, its accession towards viral status, and more recently its significantly more capable model GPT4, the entire field AI field went into an exponential growth phase. Now you have training data and models to improve AlphaFold capabilities. You have different LLM (Large Language Models) capable of doing Chemistry, being able to automatically create and test chemical compounds, including novel ones, with speed and accuracy unthinkable to human chemists, achieving 2 weeks of work in significantly less time.
Amid this myriad of innovations, one aspect of the application of LLMs to different fields of human activity that most fail to grasp is the fact the entire field is in its infancy, right now the models could be considered “inefficient”, energy-intensive, computationally “costly” endeavor needing massive clusters of GPUs specifically built for AI/deep learning research, a breakthrough in any part of this equation leads to renewed exponential growth. To your consideration, the only reason ChatGPT 4 doesn’t posses a multimodal (meaning it can interpret data from any source, text, image, video, audio) is only because OpenAI can’t acquire GPUs as fast as the business grows.
Many scientists were already sounding the alarms in early 2022 about the potential dual use of AI tools to develop new chemicals, bioagents, and possibly pathogens, at that point the use of such tools towards these goals was science fiction-esque and there was a significant knowledge gap, being broadly aware of the subjects was a necessity to engage in this endeavor, and I brought forth the point that the gap on using machine learning and AI to generate novel bioagents was narrowing. This is where LLMs come in. LLMs such as ChatGPT and now many other open source models such as ChemGPT can easily bridge the knowledge gap, as we live in an Artificial Intelligence arms race the field advances at a fast pace, this theoretical gap will become non-existent within a short period.
There are two perspectives on this matter, where LLMs can bridge this gap. The first is akin to the “Safeguarding the Future” course at MIT where they tasked non-scientist students to attempt to use LLM chatbots such as ChatGPT and prompt the AI to aid them in causing a pandemic. In one out the AI suggested four potential pandemic pathogens, explained how they can be generated from synthetic DNA using reverse genetics, supplied the names of DNA synthesis companies unlikely to screen the order, and identified protocol and how to troubleshoot them. The machine helped a non-scientist to outsource the creation of a pathogen with pandemic potential.
The second perspective is actively either training your own models or jailbreaking current models to effectively help you uncover and create chimeric pathogens, given the large knowledge body LLMs can achieve and how they can contextualize information, at the current level you already can do some rather creative designs. Which makes the next a real problem.
Benchtop DNA printers are coming soon—and biosecurity experts are worried
Report calls for better safeguards to prevent bioterrorists from weaponizing new technology
Biologists who have been obtaining DNA sequences online from companies will soon have a more convenient option: benchtop machines that can print all the DNA they need. But this technology brings with it new risks by circumventing how synthetic biology companies now screen for would-be bioterrorists. A report released yesterday by a Washington, D.C., think tank urges companies and governments to revamp existing screening to prevent someone with malign motives from making a toxin or pathogen.
The current screening system, which is voluntary, “could be upended by benchtop DNA synthesis,” says report co-author Jaime Yassif, vice president for global biological policy and programs at the Nuclear Threat Initiative. “Governments, industry, and the broader scientific community need to put stronger safeguards in place to ensure this technology is not exploited by malicious actors and that it doesn’t lead to a catastrophic accident,” she says.
The ability to synthesize DNA has been around since the early 1980s. The technology has become a central component of genetic research and is used to develop novel pharmaceuticals, agricultural products, and biofuels. Synthetic DNA sequences are available online from roughly 100 companies, which print the DNA and ship it to their customers.
Advances in DNA synthesis technology will heighten those concerns, says the report, by offering any lab the chance to buy a benchtop DNA printer that can make DNA on demand. Over the next 2 to 5 years, the report notes, the length of stretches of DNA that can be synthesized with these machines will likely increase from about 200 base pairs today to as many as 7000 base pairs, the size of the smallest viruses.
Back in 2021 and again when I began publishing my Substack on the subject matter of DNA printers, I received feedback on how the endeavor of creating pathogens using such methods was extremely hard, not reliable, or unfeasible to be done, because it took too much knowledge, too much “tech” for a bad actor to exploit these tools. Here we are.
All propositions for regulation of DNA printers are laughable at best, impossible to implement at worse, or do you really think nobody will be able to jailbreak a machine that has to be connected to the internet to print DNA, if people are still able to jailbreak ChatGPT no matter how much OpenAI patches any flaw found. You can also be creative, and circumvent the (optional) screening and order DNA that won’t be flagged, and within a short period, you can have your end goal in hands.
The only “real” hurdle up to this point is cost, a new, cutting-edge DNA printer costs roughly between 50.000 to 150.000 USD, plus material, and lab costs. Training an LLM model is tricky, but hardware wise you can do it using a powerful gaming laptop (2500 USD upwards) it does take more time and a lot of tuning, or if funds are available the normal GPU cluster (1.500 USD each) or just ordering a bunch of AI-tailored GPU (6.000 USD each), other creative methods are available and more are created every other month now, and you can also rent GPU time to train your models (also expensive).
On the other side of the spectrum, we have low-tech but advanced science.
One of many illegal labs being run around the globe with a lot of investment behind, the lab discovered in California alone had hundreds of thousands of dollars in genetically engineered mice alone, with a lot of powerful laboratory equipment. With the subjects discussed here, how easy it is becoming to produce and narrow knowledge gaps, and illegal laboratories running and creating God knows what, you may now understand why there is such a massive push to develop protocols and guidelines against “Disease X”, the US Department of State recently created the Bureau of Global Heath Security and Diplomacy and among its tasks is focusing on creating new frameworks to deal with unknown potential threads, referred to as Disease X.
Do you recall my advice on you, my reader, employing the simple analytical framework of linguistic and frequency analysis (how phrases and terms are used, and how many times they are repeated in the media) ? CEPI to support development of self-amplifying mRNA vaccine technology for use against Disease X.
Disease X vaccine is backed by multiple countries
Globally, nations are rallying to face the potential threat of Disease X, with commitments totaling $1.5 billion (£1.15 billion) to fuel the creation of a Disease X vaccine. Alongside the UK government’s pledge of £160 million ($210 million), contributions have also come from the United States, Japan, Germany, Australia, and Norway, along with investments from private entities like the Gates Foundation and the Wellcome Trust to combat Disease X.
Disease X is all the rage these days. There are multiple perspectives to all of this, a play to bring heavy regulation towards both AI technology, and synthetic biology, the common Military/Defense, and academic grant mills, but a huge part of this is in response to the changes and advancements the pandemic brought. Before “quitting” my job in January 2020 I wrote an extensive report, sent to all clients at the time forecasting the next 2 years with precision, and underlying major global changes up towards 2025, and a point I consistently and repeatedly brought up was how the pandemic would bring an accelerate era of synthetic biology.
Either by deliberate or by accident the next pandemic will be man-made. And the world will need a faster response than with Covid, a more scalable platform not prone to massive drawbacks such as the mRNA one, a novel immunization platform will be needed, either by vaccines or drugs, most likely developed by using AI systems designed for this specific purpose. I know this last comment will rustle some feathers here, but I didn’t survive the war, explosions, being shot, stabbed, brain damaged, and other injuries, recovered, to die at the hands of a nerd. At a personal level, I will judge the technology be the judge of my personal choice, but at a global level, it will be needed. Of course, the WHO treaty and the entire pandemic response can of worms is left to be debated, since make no mistake, whatever comes will be used as an excuse for global authoritarian control. I pray it is a synthetic whimper, not a bang.
SARS-CoV-2 and its pitiful “containment” attempts such as the novel Adenovirus, and mRNA-based vaccines brought massive immunological changes, perceptible both short and long-term, but the long-term changes are the ones that I have focused on for the most part of 2023, these changes can be tracked, analyzed, and weaponized. AI here merely changes the velocity of this analysis, but it can be done by a small team too. I didn’t choose to endeavor in forecast out of personal trait (or habit at this point), but attempting to mitigate whatever may come. Piggybacking on these changes with any number of pathogens would lead to abysmal levels of both disruption and potential destruction.
I usually end up there either with cryptic messages or hinting for the attentive reader to search my substack for the information I hint, further building what could be described as a hidden (hehe) puzzle. But I will be straightforward this time. Acquaintances better positioned (especially financially) have already executed many of the steps described in this article.
Everything you just read here is not a matter of if. Just a matter of will and when.
Given we are now living in the initial stages of a global war, where hybrid, unrestricted warfare is the norm, drastic changes must be made. From my perspective, both outside and inside, we can only fight fire with fire (AI vs AI).
I truly hope this doesn’t ruin your Sunday, or bring you bad vibes.
If you chose to become a supporter, thank you, and thank you for reading this work.\
Hopefully this article conveys the message as I thought in my head. Some points here will make more sense in the near future. And I may just expand on this next month. Lots of caveats to many aspects of this.
Such as using migrants as precision guided "bioweapons".
"A new, cutting-edge DNA printer costs roughly between 50.000 to 150.000 USD."
To put this cost in perspective for readers, our lower-tech microscopes run $400,000 to $900,000. Batches of scientists from many countries, whose backgrounds and motivations are unknown, cycle in and out of labs and have access to all of the equipment and data. It's a trust-based system.