“Machine Learning”, and how machines have evolved to learn, is not a new topic; in fact, it has been around for many years. Armies of data scientists, equipped with sophisticated algorithms have striven to analyze the ever-increasing volume of data to achieve the best possible outcomes in a continuous race to find the best product, the best fit, and the highest profitability. (I was fortunate to work closely with some incredibly talented data scientists at Expedia and HRS Group, as well as now in my current role at Intellias).
Naturally, these efforts in sophisticated data analysis required guidance, which was typically provided quarterly in the form of “KPIs” (Key Performance Indicators), produced by committees of human beings. in lengthy meetings fueled by countless cups of coffee and, for the younger generation, pallets of Red Bull. Outcomes were predicted, budgets allocated, roadmaps established, and retrospective meetings were set to delve deeply into efforts and results (yes, involving more coffee and more Red Bull).
Fast forward to the 2020’s and “AI” enters the scene. What is the difference now? In simple terms, we are shifting from human-input-driven algorithms to a world where machines determine the rule sets for algorithms themselves — an interpretation of how different data vectors are interconnected, or should be reconnected.
As we move away from the experimental phase and toward everyday AI-use, the question of “How will this tech evolve? And how shall we govern it?” becomes increasingly urgent – it raises a poignant question: “Humanity has ruled and dominated the planet for thousands of years…but (how) can we rule & dominate this new force?”
At this point, we might need to abandon the human notion of maintaining absolute control over forces we might not fully comprehend. Actually this might become an almost Einstein-ish singularity moment for humanity, where it might be an option to relinquish control to the machines, with humanity simply becoming the guiding force by setting principles for them. But how will the future look like? How will A.I. evolve in coming years…?
Given that we try to shape the machine in the form of human-kind’s thinking, the idea came to mind to simply draw parallels with evolution patterns in human history that might serve as guiding principles on how things might progress from here.
And what might be the most pregnant pattern which usually gave humans the biggest jumps ahead? It was the evolution of human warfare, which always seemed to drive technological advancements first more than any other human endeavor (very unfortunate, indeed, given the state of today’s world – but also made clearly evident by the new age of “modern drone warfare“, where AI also now plays a center part, too) . So let me try to outline a possible path based on human history and how some, in my opinion, very interesting parallels could be applied.
In the beginning, there was the Hoplite…
Warfare has ancient roots (yes, the caveman with his big wooden club comes to mind 😉). For the purposes of this storyline, we begin in the Classic Greek period when metallurgical advancements made the somewhat mass production of metal-based weapons feasible, giving rise to a new warrior type: the iconic Hoplite.
This warrior archetype, though slightly varied between city-states like Athens and Sparta, generally consisted of a helmet, shield, body armor, and a weapon such as a javelin or sword. The tactical formation of Hoplites, the phalanx, was a critical element on the battlefield, as it was one of the first disciplined formations that could bring recurring, scalable success through its adaptability for many different military units.
Drawing a parallel to modern technology, today’s AI models—such as GPT-4, Genesis, LaMDA, Bert, and Claude—mirror the Hoplite in their foundational similarities – a warrior here did not differ too much from the next to him in terms of skills and gear. Surely enough above named companies are all advancing towards “multi-modality,” which means integrating diverse data inputs like images, sounds, and videos. But IMHO this is only analogous to the fact that some Hoplites enhanced their effectiveness with better utilization of his pike or adding a sword, but not fundamental changes on how battle was conducted: it was still mostly man-to-man, close quarter combat.
Thus the strategies employed by Hoplite armies were relatively basic, particularly in their interactions with other units. Each warrior was still somewhat on his own when it came to bigger things, especially around adapting to changing battlefield conditions and to specialized tactics—this very much reflects the current state of AI technologies.
We use a “basic” version of what the AI models are capable of (mostly one at a time), very much still in a restrained fashion within a limited framework (Chatbot-LLMs still rule the day!). Thus, we can anticipate significant advancements and refinements in AI applications/operations in the near future.
The Roman Empire: Centurions, Legions and new tactics
One significant challenge in ancient Greek warfare was the leaders’ often misguided “hallucinations” about the feasibility of their strategies. A prime example is King Leonidas and his 300 Spartans, commonly celebrated as heroes. However, there are views that this entire endeavor was led by a somewhat delusional king, convinced that a dramatic stand would be transformative, impervious to any advice to the contrary and leaving his back completely exposed. Similarly, in many battles, phalanxes would disintegrate into chaos when the next steps were unclear.
We see parallels to these “hallucinations” in today’s AI technologies, where models may stubbornly deliver what they deem the correct answer, despite of conflicting realities. The AIs think by all means they HAVE to give an answer, because they are tasked to do so – but they are missing fundamental information below which creates a wrong outcome.
As warfare evolved into more of a science, the Roman Empire perfected it into a continent-conquering force. This era was characterized by meticulous record-keeping and strategic teachings, aiming for predictable and scalable outcomes. A systematic approach was developed to train “officers” (professionals capable of commanding a Legion’s sub-unit) leading to breakthroughs in battlefield tactics. These units could execute effective operations under pressure, whether in attack or retreat.
This brings us to the first potential future direction for AI usage: moving beyond the current model of “different AIs working together” (akin to Hoplites) to a model of a “Centurion” command-style AI. These AIs could lead other AIs in decision-making, similar to Roman Centurions who led about a hundred legionaries, rising through the ranks based on battlefield performance and merit-based votes. Such units would compete with each other on tasks, and the most successful could be promoted to lead more sophisticated “legions” of AIs, while the less effective could be decommissioned or “pruned” through split-second decision-making.
After active duty, Centurion soldiers in Roman times were often reassigned to administrative roles to enhance the overall “framework” of the Empire. Similarly, “operations”-hardened “Centurion”-AIs could take over the optimization of back-office processes based on front-line experience.
Furthermore, specialized units such as light and heavy cavalry, artillery, and mid-range attack archers became integral , each receiving targeted training and doctrine but aligned with each other through command structures for information exchange and rapid tactic/strategy setting. In today’s world, all enterprises consist of specialized units like technology, operations, finance, marketing, sales, and human resources, where specialized AIs are gaining traction.
In tomorrow’s world, we might see “departmental-lead AIs” which automatically interact with other Lead-AIs to create the best possible outcomes. These lead AIs could be constantly changed, pruned, or evolved by AIs voting on each other to build a self-sustaining system.
And thus, we arrive at the second potential future direction for AI: Governance will no longer be held by humans; instead, AIs will self-govern based on leader/follower principles where humans simply set the overall goals of the “Empire” and let the “Centurions” do their job.
The Roman Empire’s method of structured, disciplined military, emphasizing specialization and leadership, endured for nearly two millennia – however, the next evolutionary step changed things once again.
Gods & Generals: The introduction of gun powder and Napoleon Bonaparte
In evolution there always those interesting shifts, which eventually turn out to be game changers, but might not be perceived as such initially. In the evolution of human military strategy & tactics the introduction of the weapons-based gun powder created just such a big disruption. Now, mind you, the Chinese had known gun powder for many hundreds of years before Europe started to use it, but they mostly only used for the peaceful purpose of firecrackers, whilst some 500 years later in Europe this then was the more dramatically used in iron-based guns.
Gun powder then served as that extreme technology enabler, which fundamentally changed the shape of the battlefield almost overnight.
A question then arises: What will the “Gun Powder”-moment of AIs be…? What will be that additional technology-enabler component which allows a complete revamp of applied AI operations and strategy…?
To be completely honest: I simply don’t know… Obviously the promise of Quantum Artificial Intelligence might address this question – but AIs also heavily depend on “connectivity” (to data sources, to other services and, ultimately, to each other), and not just pure raw processing power, thus something additional will need to come into play. Maybe some very advanced kind of Quantum-based Napoleon Bonaparte, who could outfox anyone by superior overall understanding of the battlefield, theater-of-operations and intentions of others.
Such an AI could possess some mind-reading-like capabilities , taking into account the smallest details (Do people sweat in the office today? Are some workers unhappy because their favorite sports team lost yesterday? Is the smell of that particular flower in the flower arrangement in the entrance hall of the company disturbing to some?) to create constant super-micro based outcomes.
Combining different domains: The Second World War
By the time the 2nd World War arrived, a new dimension was added to the pre-dominant areas of “land” and “sea” military branches: it was “air” (via airplanes and the beginnings of “space”via rockets). This led to the principles advanced combined tactics and strategy, where different overall types of military branches started to closely align (especially in the beginnings of WW2 for great success).
How would this translate this to AI though? Here “combined maneuver” of different branches, who previously were rather separated in doctrine, created a superior outcome.
In my opinion this could be the case on how in the future companies, through their AIs, fully automated work witch each other, with the goal to produce optimized split-millisecond timings and reproducible micro-results. Their “combined maneuver” will add a complete new “dimension” to AI use.
Think for a minute on how travel would relate to that (please take into account the principles of “Deviceless Travel” I had written about some months ago): Any smallest step of your journey is computed and nano-second timed, based on any little movement/state-of-body/state-of-mind you are in. AIs from different companies, via the interconnected AIs, communicate constantly directly to produce a myriad of possible outcomes and adjust anything around you for immediate micro-level comfort during any stage of the journey.
Modern Warfare: Anything, Anywhere at Anytime, at Any Detail
First of all: there is NOTHING to be celebrated about how human-kind conducts military operations today… We in essence as human species have not evolved yet to the peaceful societies dreamed up in Startrek-like visions of the future…
Nonetheless, we can see the effects of this change every single day : an almost to the grass-blade documented and surveilled state of human conflict, with thousands of drones covering the skies above, letting escape no one and no detail missed. The myriad of devices seemingly creating an ever switched-on awareness-state of what is going on.
But how to translate that into the future of AI...?
One of the very first science-fiction books I ever read was the, by today’s standards prophetical, science fiction masterpiece “Cyberpunk” from William Gibson.
There was not even an idea about the Internet, but Gibson already wrote about the “Cyberspace” and “Console Cowboys” hacking themselves through the online world – and this being a dark, dystopian version of the future with failed wars, sprawling Megacities and a full-blown-gone-crazy-overboard globalization.
One of the other interesting things was Gibson’s focus on how humans and machines would start to melt into each other, on the one side by direct computer interfaces plugged into human brains and on the other side bio-mechanic chips, where live organic matter would be serving as a sort of a “hardware”.
Now, “Edge AI” (so fully embedded AI into devices), is already a hot topic of discussion today. Whilst “Edge AI” certainly will work already for the previous described sections, its currently still bound to mechanical devices, and us as humans still need to interact indirectly through the help of our body-bound senses, taking information in visual and aural ways.
And this might be where this all then comes together: the “ever switched-on awareness-state of what is going on” might need that final barrier of direct integration into humans (and living beings), where the embedded-AI in our bodies communicates with other embedded AI-devices (in both machines and humans), to create a full new level of awareness.
So this is the part of the story where it now does get a bit scary, right?
Humankind, for millions of years, was always fixed on one topic: domination of others to move up in this world. But all of a sudden our will to dominate and rule others (and yes, that also means via our prime method of warfare) might be simply overruled by microscopic embedded steering of AIs, the good friendly “angel” sitting on our shoulder, taking care of everything, by talking to the other “angels” out there.
In what could that possibly result?
- You would never have to date again. AI’s would decide on who your best dating-partner would be, communicating with each other on the fly, to find that perfect match for you (depending on how much of your true self you could hide from your embedded AI, right?)
- You would never ask any institution for anything. Your loans and mortgages would be granted by passing-by a willing banker in the street, who could feel your “I-need-to-build-a-nest-for-the-family” vibe
- You would not commit any crimes. The AI would stop you and you would get rewarded by lower insurance premiums, cheaper loans (yes, hello Mr. Banker above!), better anti-aging treatments, micro-financing from other embedded AIs who “feel with you” etc etc
But most importantly: we would not ever start any wars again. Because its likely that any AI could tell you that working together for the common good would always produce superior results and would value life above everything. And thus let humanity thrive through peaceful means… and I hope that is the final good conclusion to this article 🙂
(But do rest assured, dear reader, humanity will always find a way to even mess this perfect scenario up, right?)
1 Comments
Pingback: Travel Tech Essentials #146: Evolution |