Surfing the Singularity - "The Coming Wave" (a book report)
Mustapha Suleyman knows a thing or two about AI. Originally co-founder of DeepMind, a company and IP eventually acquired by Google, Mr. Suleyman is now CEO of AI at Microsoft. In this latest "Surfing the Singularity" blog installment, we'll review his recent book "The Coming Wave". Hang ten!
Go Where You Wanna Go
As a game, Go is notorious for its huge array of potential moves, exponentially more complex than chess for example, where computer models beat the best chess player way back in 1997. In 2016, DeepMind's model AlphaGo beat the best Go player in world after being trained the better part of a year with reinforced machine learning on a data set of human Go games and computer-vs-computer play. The following year, DeepMind's AlphaZero exceeded that performance in just a few days of training computation without ever being shown a single Go game, just having been described the rules of the game.[1]
<div style="text-align: center;">Alas, born at the wrong time.</div>
In his Bill Gates-recommended [2] book "The Coming Wave", Mr. Suleyman's dystopian thesis is this: that the combination of AI, synthetic biology, and a host of other general purpose technologies such as robotics and additive manufacturing will combine into a major technological wave which will wash over the human race and alter it in unprecedented ways. Much as in past waves - the harnessing of fire, the wheel, the printing press, the combustion engine - each set off dramatic and often cataclysmic societal change the likes of which was certainly not obvious or expected by the "engineers" which developed the tooling. Call it the "rule of unintended consequences". The author supposes there have been about two dozen such waves over human history, and as expected in these times, the rate of arrival of transformational technologies is accelerating.
Take the printing press. Originally in 1440 there is but one device, the lab prototype. Fifty years later there are 1,000 printing presses in Europe. From producing just a few thousand hand-copied manuscripts a year, the bookmakers now produced millions. Demand for books soars, cost per unit drops, adoption deepens. What was the impact of this new information proliferation in the society? As Suleyman writes: "Gutenberg just wanted to make money printing Bibles. Yet his press catalyzed the Scientific Revolution and the Reformation, and so became the greatest threat to the Catholic Church since its establishment." And in spite of the efforts of certain Byzantine lords to control the press, proliferation of the technology was and is the default, driven by FOMO, at least.
Straight Outta Coruscant
The mass-scale rollout of AI is already underway, hand-in-hand with surveillance devices at the edge, high speed networking, nearly bottomless storage, and high performance computing on demand to make sense of it all. All so "The Algorithm" can feed you tailored news (and ads) with your morning coffee. And more, much more. Large Language Models (LLMs), trained on the corpus of human written and other creative output can now generate helpful suggestions in a variety of useful contexts (such as blog writing). And as I wrote about in my last blog [3], LLMs are useful code assistants too, although here current state of the art is about a 50% success rate on senior-level software development tasks. So yes, there is room to grow, but in line with the acceleration of the rate of change, we expect that gap to be closed in short time.
What then? A whole host of human-centric but generally rule-oriented tasks - think: back-office work in the finance and insurance industry - will become fair game for AI augmentation, meaning, human replacement. We see the rise of autonomous vehicles - think: bus and cab drivers, mail and package delivery, pilots. Air traffic controllers. Call centers. Medical radiology readers. Not one of these applications requires a super "artificial general intelligence" (AGI), simply a good model tailored to a specific task set, aka "artificial capable intelligence" (ACI). This is nearly all line-of-sight to market.
What then is not here now? The author spends a good amount of time discussing the rise and impact of artificial and synthetic biology, CRISPR gene hacking and the like. Not being personally equipped to analyze such biotechnologies, I'm simply going to leave that one to the reader, suggest its some heady stuff, but otherwise stay in the domain of the electro-mechanical. But even with this scope limitation, what is the wider wave?
<p>
</p>
Consider the rise of the bots, farm bots. GPS-guided autonomous tractors (already a thing).[4] These robots don’t look like C3PO tending the moisture reapers on Tatooine - these robots look like farm equipment and are painted John Deere green. Amazon and Walmart distribution warehouses are already highly automated - combined with autonomous vehicles and AI-driven back-office work, how many employees do we think Amazon will need in 2 years? In 5 years? They currently employ 1.5 million people and reduced headcount 5% in the last 2 years while growing revenues over 20%.[5]<p></p>
Mind the Gap
And while your local Joe the Plumber [6] may continue to have a job visiting homes for some time to come, the use of single-task robotic automation in construction, especially new commercial construction, and property maintenance is on the rise. Concrete and paint-spraying robots. High rise window washers. Roofers, and general laborers to move material around a job site - bots - flying, floating, swimming, walking, drilling, boring bots. And more factory bots too, using traditional and additive techniques, to print their own parts. Bots that make bots. And an unemployment office at the Department of Labor run by AI.[7]
What about joining up with Uncle Sam, see the world, serve your country? Drone warfare in Ukraine has shown the folly of the massing of expensively equipped troops, and in the Red Sea the risks associated with large and high priced floating projections of power. Hypersonic weapons, beyond the capabilities of a human in-the-loop system to thwart. The result is asymmetric bot-on-bot warfare, beyond the battlefield, beyond borders. What are we to do with legions of technically unemployed, if they are not even useful for cannon fodder? And what are we to do with the State, if it cannot provide a system which benefits the population, which can keep it protected from proliferating technological threats? With information, wealth, and power centralized in the hands of a self-selected few, is it pitchforks and torches to the barricades then?[8]
While its clear and no surprise the coming wave will benefit those with technical and financial authority, there is a chance of a boomerang effect which will result in forces in the opposite direction. The tooling, including the availability of sophisticated AI models and the means to run them, is being democratized. While increasing in capability, the cost of military-grade drones has decreased orders of magnitude in the last decade.[9] Rabble-rousing AI deepfakes proliferate. As Mr. Suleyman says, "anyone motivated to sow instability now has an easier time of it", not just state actors, agents, or oligarchs, but anyone with a few thousand dollars and an axe to grind. And considering the examples from recent 21st century past, if a rogue actor were to leverage the technology for nefarious purposes (think: 9/11 and the Patriot Act), there would surely be immediate call by the population for protection, likely but perhaps not exclusively by the State, backed by pervasive security surveillance. And this time, the means to fully execute on that wish exists.
The China Syndrome
<p style="text-align: center;">”The system works! That’s not the problem!”</p> <div style="text-align: left;">It is a coming wave of contradictions and competing forces, and it sounds disruptive and quite unpleasant to say the least, perhaps even a human catastrophe. And besides avoiding the topic of bioengineering, we also haven’t yet discussed what happens when we actually do get to superhuman generalized AI - we’re still talking here about relatively dumb AI with human actors in charge, in theory.</div> <p></p>
The author Mr. Suleyman concludes that the containment of this new technology - this artificial intelligence backed by autonomous mobility - a containment which has rarely if ever been possible (nukes being maybe the sole exception), must be done successfully, and urgently. Its a good sentiment, albeit one which may be too optimistic, even blindly. Can the march of this autonomous AI "progress" with its obvious and as yet to be seen additional consequences be stopped? I would argue, and the author would likely in the final analysis have to admit, that it cannot.
What to do about it? Maybe we should give serious thought to the existential question of what it actually means to be human.[10] Or, alternatively, as Timothy Leary said...[11]
Until next time. - andy
References & Amusements
[1] "The Coming Wave", Mustapha Suleyman, Crown Pub., 2023
[2] Bill Gates blog, https://www.gatesnotes.com/holiday-books-2024
[3] "Surfing the Singularity: Super Grover!", https://surfthesing.blogspot.com/2024/12/surfing-singularity-super-grover.html
[4] "John Deere Robot Planter", https://www.cnet.com/tech/john-deere-robot-planter-the-future-of-farming-looks-like-fewer-chemicals/
[5] https://www.statista.com/statistics/234488/number-of-amazon-employees/ and https://www.statista.com/statistics/266282/annual-net-revenue-of-amazoncom/
[6] https://www.nytimes.com/2023/08/28/us/politics/samuel-wurzelbacher-joe-the-plumber-dead.html
[7] https://www.dol.gov/agencies/oasam/centers-offices/ocio/ai-inventory
[8] https://www.stlouisfed.org/community-development-research/the-state-of-us-wealth-inequality
[9] https://www.technologyreview.com/2023/01/30/1067348/mass-market-military-drones-have-changed-the-way-wars-are-fought/
[10] https://www.organism.earth/library/document/unapologetically-human
[11] https://www.youtube.com/watch?v=IPSzTBP5PAU