Deep Dive: Dario Amodei's Machines of Loving Grace
The CEO of Anthropic Expects Advanced AI as soon as 2027
Dario Amodei, the co-founder and CEO of the AI company Anthropic just wrote “The Machines of Loving Grace,” a paper anticipating powerful AI as soon as 2027 - what he calls “country of geniuses in a datacenter”.
He anticipates radical technological progress in medicine and biology in the wake of developing such a formidable AI:
Thus, it’s my guess that powerful AI could at least 10x the rate of these discoveries, giving us the next 50-100 years of biological progress in 5-10 years.14 Why not 100x? Perhaps it is possible, but here both serial dependence and experiment times become important: getting 100 years of progress in 1 year requires a lot of things to go right the first time, including animal experiments and things like designing microscopes or expensive lab facilities. I’m actually open to the (perhaps absurd-sounding) idea that we could get 1000 years of progress in 5-10 years, but very skeptical that we can get 100 years in 1 year. Another way to put it is I think there’s an unavoidable constant delay: experiments and hardware design have a certain “latency” and need to be iterated upon a certain “irreducible” number of times in order to learn things that can’t be deduced logically. But massive parallelism may be possible on top of that15.
You may notice some similarities and differences to what I wrote in “OpenAI's Strawberry/o1 - The Edge of the Intelligence Explosion,” in which I explained how the US Federal government - and for now, only the US Federal government - could begin a period of runaway scientific and technological progress simply by leveraging the o1 AI model using, in turn, other advanced AI programs from other companies like Google DeepMind’s AlphaFold and GNoME.
Probably the greatest difference is my expectation that the above work with o1, if possible and if executed, could lead to the most-important aspect of the fabled Singularity - a time when technology is not only changing faster than we anticipate, but faster than we can anticipate.
Amodei clarifies that he does not expect the kind of instantaneous changes often associated with the Singularity in the first year or even few years after achieving advanced AI.
His arguments largely make sense… but let me offer a caveat.
We don’t need advanced cities or starships materializing out of nowhere from nanotech or other science-fiction outcomes to reach the pace of a Singularity.
The definition of technology changing faster than we can understand isn’t as hard to reach as we might imagine.
I believe that for most of us, the leading edge of technology has already passed this “event horizon.”
Most ordinary people don’t know how fast technology is moving on the leading edge and would be shocked by how rapidly it’s moving even now.
If you’re involved in one of these fields or keep track of one or more in your work or hobbies, you’re likely not lost - but how many people know about even basic things like all the major breakthroughs in AI?
Or gene therapy curing sickle cell disease? Or AI flying US f-16s? Or the combination of generative design, 3D printing and CNCs designing and constructing advanced parts for cars, planes and spaceships within a day?
You know… Basic things.
The future isn’t closer than we think.
It’s already here, just waiting for us to notice.
