This is an amazing, comprehensive survey of what amounts to a fifth industrial revolution, moving in tandem with the developments in creating AGI. There are existential risks associated with both and this field needs someone like Nick Bostrom to produce a non-technical risk assessment comprehensible to the general public. Great job!
Totally agree with your Bostrom suggestion. That said, I'm ever less convinced that reason and facts are up to the job of educating the public. It's like trying to teach a small child not to touch the stove. You can tell them a million times and try to patiently explain, but until they actually touch the hot stove themselves a lot of what you're saying will go in one ear and out the other. Once they touch the stove though, understanding then mushrooms. Most people are not that great at understanding things in the abstract, they need to see it to believe it.
So for example, if some biologist were to produce a horse with four heads, and that image goes viral across all forms of media, then the public will start paying attention. And by then...
I suspect that figuring out how genes behave will prove a far more daunting task than bringing down the costs of DNA printing. There are simply so many layers on top of the genome... People often talk about a "second genetic code," but to me it always felt like there was a "2nd, 3rd, 4th and nth genetic code"
I suppose that's where the advantage of fully synthetic cells lies. By starting from a "simplistic" (for biological standards) foundation that is extremely well-known and predictable, we can then build on top of that step by step in a similarly predicted fashion and tailor our synthetic organism to do exactly what we want. But to get all the wonders of life using this approach would be a lot like building them from scratch...
We'll probably always want to highjack naturally occuring biological systems in one way or another. They're just too powerful for us not to
The use AI models to predict the effects of large-scale genome engineering has interesting implications. Considering that AI models are increasingly afflicted by the "black boxes" problem as they grow larger, meaning that they may deliver accurate predictions but we may never know why, I suspect we may come create biological machines that we can't really understand. All we'll know is that our models told us that genome X would produce phenotype Y, and it works
Thanks for the comment. And yeah, I agree. Prediction is the largest hurdle to easy-access, synthetic genomes. Synthesis, I suspect, will be basically solved in 2-3 years.
Without prediction, we'll just keep building E. coli, yeast, and other genomes with **slight** modifications over wildtype. Prediction enables us to build truly de novo organisms.
Wow, an impressive article indeed. This is why I live on Substack. While much of the article goes over my head, I think I get the overall bottom line that very big things are happening in the field of biology. I did a lot of reading on CRISPR a couple of years ago, and this article persuades me that I need to return to such reading.
Imho, discussion of regulation is largely meaningless because the people we need to be most worried about will not be submitting themselves to any form of regulation. So many of the articles I read on emerging technologies like AI and genetic engineering seem to assume that Western elites will remain in control of such technologies and are in a position to manage them via governance schemes cooked up at academic conferences etc. Just not buying it...
Here's who is most likely going to determine the future of these fields, the Chinese Communist Party. China is, for example, four time as populous as America, has much more talent to draw upon, and need not worry too much about push back from the public. If the people at the top in China decide they want to move in some direction, they can just do it.
Any power given to the good people is also given to the bad people, and as the scale of power grows the bad people get the edge.
This is an amazing, comprehensive survey of what amounts to a fifth industrial revolution, moving in tandem with the developments in creating AGI. There are existential risks associated with both and this field needs someone like Nick Bostrom to produce a non-technical risk assessment comprehensible to the general public. Great job!
Totally agree with your Bostrom suggestion. That said, I'm ever less convinced that reason and facts are up to the job of educating the public. It's like trying to teach a small child not to touch the stove. You can tell them a million times and try to patiently explain, but until they actually touch the hot stove themselves a lot of what you're saying will go in one ear and out the other. Once they touch the stove though, understanding then mushrooms. Most people are not that great at understanding things in the abstract, they need to see it to believe it.
So for example, if some biologist were to produce a horse with four heads, and that image goes viral across all forms of media, then the public will start paying attention. And by then...
It will be too late.
Extremely interesting overview!
I suspect that figuring out how genes behave will prove a far more daunting task than bringing down the costs of DNA printing. There are simply so many layers on top of the genome... People often talk about a "second genetic code," but to me it always felt like there was a "2nd, 3rd, 4th and nth genetic code"
I suppose that's where the advantage of fully synthetic cells lies. By starting from a "simplistic" (for biological standards) foundation that is extremely well-known and predictable, we can then build on top of that step by step in a similarly predicted fashion and tailor our synthetic organism to do exactly what we want. But to get all the wonders of life using this approach would be a lot like building them from scratch...
We'll probably always want to highjack naturally occuring biological systems in one way or another. They're just too powerful for us not to
The use AI models to predict the effects of large-scale genome engineering has interesting implications. Considering that AI models are increasingly afflicted by the "black boxes" problem as they grow larger, meaning that they may deliver accurate predictions but we may never know why, I suspect we may come create biological machines that we can't really understand. All we'll know is that our models told us that genome X would produce phenotype Y, and it works
Thanks for the comment. And yeah, I agree. Prediction is the largest hurdle to easy-access, synthetic genomes. Synthesis, I suspect, will be basically solved in 2-3 years.
Without prediction, we'll just keep building E. coli, yeast, and other genomes with **slight** modifications over wildtype. Prediction enables us to build truly de novo organisms.
Wow, an impressive article indeed. This is why I live on Substack. While much of the article goes over my head, I think I get the overall bottom line that very big things are happening in the field of biology. I did a lot of reading on CRISPR a couple of years ago, and this article persuades me that I need to return to such reading.
Imho, discussion of regulation is largely meaningless because the people we need to be most worried about will not be submitting themselves to any form of regulation. So many of the articles I read on emerging technologies like AI and genetic engineering seem to assume that Western elites will remain in control of such technologies and are in a position to manage them via governance schemes cooked up at academic conferences etc. Just not buying it...
Here's who is most likely going to determine the future of these fields, the Chinese Communist Party. China is, for example, four time as populous as America, has much more talent to draw upon, and need not worry too much about push back from the public. If the people at the top in China decide they want to move in some direction, they can just do it.
Any power given to the good people is also given to the bad people, and as the scale of power grows the bad people get the edge.