[cornholio]

As a technologist, I have mixed feelings when I see the fascinating details of life. On one hand, mastering this molecular machine would give us literally God-like powers: we could fabricate, grow or heal anything. We could solve all current problems, we could terraform planets using a few milligrams of DNA and literally redefine what it means to be human.

On the other, I see the human body as a completely unsecured cybernetic system, that can be so easily tricked to pick up any random bit of programming and insert it into it's own code. There is no forethought or design, no rational defense, just good enough systems that have evolved randomly against non-rational adversaries that happened to emerge out of the protein soup surrounding us.

The troubling fact is that mastering the wondrous biomolecular machine necessarily comes with the power to kill every human on the planet. Truly God-like powers.

[ekianjo]

>　power to kill every human on the planet. Truly God-like powers.

You can't really design a perfect virus that will wipe out the human race, because anything you do to affect its properties will also affect its ability to spread. I'd be way more concerned about the destructive power of nuclear weapons, still numerous enough to destroy a very large part of humanity and our vital infrastructures.

[cornholio]

Are you sure you aren't still thinking in the conventional, evolutionary paradigm?

A highly engineered bioweapon could circumvent such problems by separating the infection phase (which could be completely silent and airborne) from the eradication phase. The payload could be triggered deliberately at a later date when a certain secret artificial protein is released in the environment - and then produced in industrial quantities by infected hosts. Or maybe airdropped over areas that should be cleansed.

And that's just scratching the surface of what's conceptually possible. It could trigger specific ethnic characteristics or individuals, it could set up exotic cyber-hybrids like public key decryption in DNA for commands from its command and control. It could create side channels among infected hosts, for example by triggering minute anatomical modifications in the inner ear and the vocal centers, making them able to send and receive ultra- or infrasounds controlled by the mallware.

As a more subtle cyberattack, an infected individual could grow a whole parasitic subsystem that extracts select visual and auditory data and stores them in DNA memory for later broadcast.

[sizzle]

Wow these are novel concepts, do you read a lot of sci fi? How did you come up with these interesting scenarios?

[cornholio]

I'm not really inspired by scifi - but I'm sure some authors have had similar and probably much wilder ideas.

Been thinking for years about the human body as a cybernetic attack surface with no engineered cyberdefense. Most people seem not able to make that mental leap; no, the human body can't behave like a vulnerable Windows 95 machine giving kernel privileges to any ActiveX control it can download, because reasons.

But once you see the biological world like a hacker and DNA like a programming medium, as opposed to a representation of what evolution produced, an endless array of nefarious possibilities become obvious. The rational power of our minds far exceeds what evolution could ever concoct - or defend against.

[telchar]

To be fair, that argument depends on the properties of evolved pathogens. A designed pathogen, in the context of the GP post, would not necessarily be bound by that dilemma.

[mikeash]

What would prevent the development of something like HIV that’s able to spread through the air like the flu?

[cryoshon]

slower replication speed, more difficulty infecting the host, requirement of much greater volume of replication to accomplish the same rate of infection.

think of it this way: every cool feature you add to a living thing has an overhead.

you want your little bacteria to have antibiotic resistance? fine.

but it'll need that much more energy to grow relative to the bacteria which don't have the added burden.

this means that unless there's the selective pressure of antibiotics in the environment, your little antibiotic resistant microbe won't stand a chance -- it's a fraction less efficient under normal conditions, so it's effectively out-competed when in the wild.

as far as massive adaptations like airborne spreading, that's not something that can just mutate overnight. HIV is fragile, so you'd need to engineer an entirely new viral envelope, or, more likely, an entirely new carrier particle that the virus can reconstruct on its own without impacting its infectivity. viruses like the flu have these adaptations by default. but once again, if you decide to turn the flu into HIV, it's going to be at a disadvantage in the wild.

not to say that it is impossible to make virions which are able to out-compete their wild-type cousins when infecting hosts in the real world. far from it. it's just not as easy to make it work as a quick look might find.

[godelski]

What would be the overhead for just something "simple" though? Like making HIV airborne? (Or rather what the lay person perceives as a small change)

It seems like there are plenty of diseases out there where a (apparently) "small" modification could have a dramatic effect on how it spreads. And I'll admit my naivety to the subject and do not know if such small changes are actually small, or the related overhead associated with them.

[cryoshon]

to answer your question, the overhead is very small for a small change. you can add a bit of noncoding DNA to a virus' genome without ruining its ability to compete in the wild. but the survival margins are very thin. on a population scale, natural selection is very harsh. anything that is superfluous given the environment is an inefficiency which eventually results in extinction. of course, between organisms this isn't that frightening because there are different niches, so sometimes a large change can be more viable than a small change even if it's a lot more expensive, provided that the large change lets the organism live in a new niche.

making HIV airborne isn't a simple change, however. it's more like a massive change of niche. it's a change in the transmission modality of the virus -- for comparison, consider the scale of the changes you'd need to make to turn a car into a plane. or maybe a car into a boat.

it's doable, artificially. but the result won't be as good at being a car, plane, or boat as something which was purpose-built for that application and didn't have to carry the features of something intended for a different purpose.

many of the "small" changes that make a disease spread more easily are actually mutations which don't change the ability of the disease to weather external conditions, but rather change the ability of the disease to survive first contact with the host's immune system.

the flu is a great example here. we need a new flu vaccine every year because the flu mutates constantly and drastically. the flu never becomes capable of surviving outside a host for longer than before, though. it just becomes more effective at evading the immune systems of most hosts.

[mikeash]

What about starting with the flu and giving it an HIV-like ability to wreck your immune system? Is the “attack” part too intertwined with everything else to be able to do that sort of mix-and-match operation?

[ASalazarMX]

> You can't really design a perfect virus

Just design several good-enough viruses.

[nabla9]

Interestingly the Gene Regulatory Network (GRN) - the interaction of genes and proteins that control the prosess inside the cell is computationally very similar to recurrent neural network. Gene expression levels is controlled by proteins that are produced by active genes. https://en.wikipedia.org/wiki/Gene_regulatory_network

Harnessing this mechanism directly for neural computation would be grand project.

[daveguy]

I think it's a bit of a stretch to say GRN is "computationally similar" to neural networks. That seems to be a shoehorn of the most popular technology of one field into another field.

Just because the GRN contains feedback loops with multiple influences doesn't mean it's suited to NN computation. The GRN is orders of magnitude more complex than computational NNs and it is orders of magnitude slower than signal transduction of axons.

I agree it will be awesome to be able to harness molecular machinery to "do stuff" for us. The "nanotechnology" in our bodies is way beyond any current manufactured nanotechnology.

We currently have only crude control of the cellular functionality -- think point a wind up toy in the direction we want or modify it by taping a flashlight to the top. We are pretty far from being able to use the manufacturing equipment to make cars instead of windup toys. (Not to mention the manufacturer is already making rockets)

[nabla9]

> That seems to be a shoehorn of the most popular technology of one field into another field.

Recurrent neural network is used to model gene regulatory network. It's not a shoehorn.

See for example:

[1]: Reconstruction of Gene Regulatory Networks from Gene Expression Data Using Decoupled Recurrent Neural Network Model https://link.springer.com/chapter/10.1007/978-4-431-54394-7_...

[2]: Gene regulatory networks inference with recurrent neural network models https://ieeexplore.ieee.org/document/1555844/

[3]: Recurrent Neural Network Based Modeling of Gene Regulatory Network Using Bat Algorithm https://arxiv.org/pdf/1509.03221.pdf

> The GRN is orders of magnitude more complex than computational NNs and it is orders of magnitude slower than signal transduction of axons.

It's possible that we can reduce relevant complexity to the RNN subset that it useful. Feedback loop speeds are slower but they can be below second.

In many search and optimization problems the ability to run say 100 trillion large stochastic RNN's in parallel in a 100 liter tank could be huge. Especially if all you need is glucose and few cheap nutrients to power it.

[daveguy]

The articles you cite start with experimental data about Gene regulatory networks (eg from dna microarray) and then use rnn to characterize or produce the networks known or elucidated experimentally.

None of the sources claim functional equivalence of the GRN by the RNN or vice versa.

From a "big O" computational complexity perspective the gap between what you are describing and the actual case is the gap between P and NP. Just because we can confirm the results of a GRN with an RNN doesn't mean we can produce those results.

Yes, biological computing could harness very powerful parallelism. We are nowhere close to harnessing that power. (See toy manufacturing analogy)