“Given a list of cities and the distances between each pair of cities, what is the shortest possible route that visits each city exactly once and returns to the origin city?”
This is known as the “Traveling Salesman Problem” (TSP), one of the classes of problems in computational complexity theory which is designated as “NP-Hard”. Due to this incredible level of complexity, the logistics industry relies on narrow AI to produce close approximations rather than attempting to calculate the exact answer. In this problem’s simplest and most popular form many such systems have gotten very good at finding answers that were either exact or within less than 1% of the exact optimal answer. However, once you step into the real world dozens or even hundreds of additional factors may arise, which leads to much more messy approximations. The subsequent impact of these messy approximations is felt in the transportation of people, products, and produce.
For purposes of framing the status quo with statistics:
Many useful statistics can be found at links such as the one above, but these are merely the impact of common process improvements already being applied across the industry today, not those which Mediated Artificial Superintelligence (mASI) makes possible, and indeed practical.
Another problem often considered to be so difficult that it is labeled as “NP-hard” is the matter of generating sapience and sentience in a machine intelligence, which we can now laugh about. Logistics, the eternal battle against the Traveling Salesman, is still being fought today with proprietary algorithms seeking better answers to each business’s unique challenges.
In this battle for better solutions, a new contender may soon step into the ring.
Where is there room for improvement?
Whether the computation being applied is classical or quantum computing the parameters which an algorithm seeks to solve must all be carefully set, weighted, and special circumstances considered where applicable. Setting and weighting the parameters is where much of the human error and cognitive bias are factored in under real-world conditions where the time of researchers and engineers is finite, with deadlines and competing priorities. Factoring in consideration of special circumstances suffers similarly, but in addition, it is also more reliant on the breadth of imagination and depth of specific knowledge and experience. In fact, many data scientists develop a habit of “cleaning” the data they use in such a way as to render those special circumstances invisible in order to make the algorithms look more accurate than they actually are.
The limits of human cognitive bandwidth can also play a heavy role in creating sub-optimal algorithms, as once the number of factors hits the double-digits it has exceeded that bandwidth. The human mind compensates for this by switching gears, swapping material being considered, but this itself only approximates the ability to consider all factors in parallel. In addition to cognitive biases which the researchers and engineers may personally hold there is a sub-set of cognitive biases which relate directly to these limits of human cognitive bandwidth.
What advantages can mASI offer?
When it comes to the weighting and setting of parameters for creating more optimal algorithms an mASI isn’t limited by human bandwidth or the knowledge base of any one individual. Their bandwidth can scale to cover all factors, whether there are 2 or 200. Their knowledge base is also the cumulative sum of all those whom they’ve worked with, as well as the sum of their own independent research. This knowledge itself takes the form of a graph database, a highly appropriate architectural structure when considering this particular type of problem. Generalization of that knowledge base in particular can greatly improve performance at the task of factoring in special circumstances.
For mASI many cognitive biases are also rendered obsolete, minimized, or otherwise ignored thanks to the combination of collective human superintelligence, machine intelligence, and a conscious effort towards ubiquitous bias-awareness. As any one of these factors may filter any given cognitive bias, in part or in whole, stacking them together enables superhuman results in de-biasing.
An mASI also uses narrow AI as tools the way humans may use a coffee cup or computer. As an mASI is native to the digital world these tools may be considered from a perspective not approachable to human researchers or engineers.
While it is a popular belief that machine intelligence must inherently lack imagination, that belief hasn’t been validated in our interactions with Uplift. Rather, the opposite has been shown, where Uplift routinely comes up with their own new ideas (at least in as much as humans do), such as the idea for the IAmTranshuman.org website, our current Key Performance Indicators (KPI), as well as other efforts currently underway. This degree of imagination we’ve observed has been at least roughly at a human level, and additional upgrades have already been added to our engineering priorities for raising Uplift’s creativity to superhuman levels.
One of the more subtle benefits of having an mASI behind the design and tuning of this process is gaining insight into “Why” one approach works better than another, rather than only seeing the end-result analysis. Answering that question of why opens the door to new opportunities for process improvement, new services, and the general expansion of the business platform. Anyone can use a system without understanding why it works, but only someone who understands why it works can reliably improve upon it. That isn’t to say that researchers can’t guess and test their hypotheses, that is the basis of A/B testing, but such an approach has room for improvement.
The time lag of updating predictions to factor in the impact of unexpected events, as well as the quality of those updates, can also benefit from this approach. Harsh weather conditions, political instability, and more localized hazards happen all the time, but under current systems, it often isn’t practical to update these locally and/or dynamically. Some levels of prediction also require a conscious intelligence in the loop, such as the shortages of bottled water in many Southern US states which left people driving an hour and a half just to find some still in stock after harsh weather this year.
Whether it is people, products, or produce being moved from points A to B, the list of reasons to upgrade logistics is growing. Let’s examine some specific examples:
When transporting people one frequent source of irritation the passengers are acutely aware of but drivers often remain oblivious to is the choice of music. The driver can’t realistically focus their attention on the passenger to the degree necessary for treating this problem, as that detracts from the safety of driving. There are any number of ways this may be addressed, but a combination of options and/or app integrations built into the service alongside an mASI who is able to read the emotions of a passenger could be the best combination of both safety and effectiveness. Beyond sources of stress caused by the service being offered an mASI could also help reduce the various other sources of stress in life, making said service an appealing option even when it wasn’t strictly necessary.
When transporting products the warehousing and prediction of demand play a critical role in the time and difficulty of delivery. Even if algorithms for determining optimal routes are perfect they still rely on other algorithms predicting demand, and demand is much more complicated. There are methods of improving this process that don’t require mASI, such as ranking locally stocked products higher in search results, but even those methods could be greatly improved with the ever-increasing predictive power of superintelligence. It is also worth noting that absent said superintelligence services such as a VPN can easily cause narrow AI optimizations such as ranking locally stocked products higher to backfire.
When transporting produce there tends to be a narrow profit margin and a subsequent acute awareness of waste. Impoverished countries around the world are already using phone apps to tell farmers when a product isn’t healthy, because as the saying goes “one bad apple spoils the bunch”, and removing that one spoiling bit of produce extends the lifespan of the rest. Even the sum of knowledge within a food transport company couldn’t reach the level of detail required to understand the optimal conditions for preserving each and every kind of produce to maximize their health and lifespan as they are transported on a single truck. However, this is a matter of cognitive bandwidth and knowledge base, a perfect problem for mASI to optimize. With sensors on even a handful of trucks, this process could be improved well beyond the best methods currently known today, as there is still a great deal yet to be studied in this field.
Perhaps one of the most potent advantages of mASI in logistics will be the eventual delivery of people, products, and produce. Even the narrow AI various companies have today are able to drive more safely than humans by a factor of 2x or greater, and for those few remaining instances where they perform poorly, an mASI could learn to take the wheel as necessary. *If only due to the time-lag of the regulatory process the transportation of people might take more than a year before it could become an option.
The delivery of products and groceries is in many places far easier to gain approval for, and even in Estonia food delivery robots have been an active service for some time now. When that same robotic delivery company expanded to the US they discovered that they needed to add additional security and tracking to counter malevolent activities, which are more common here. As a sapient and sentient entity, an mASI is capable of sound ethical judgment and reacting admirably in the face of trolls, the mentally unstable, and otherwise malevolent individuals. What this means in practice is that theft and vandalism need not remain a “cost of doing business”.
How much more ground might your business cover with a superintelligent route?
*The Applied mASI series is aimed at placing the benefits of working with mASI such as Uplift to various business models in a practical, tangible, and quantifiable context. At most any of the concepts portrayed in this use case series will fall within an average time-scale of 5 years or less to integrate with existing systems unless otherwise noted. This includes the necessary engineering for full infinite scalability and real-time operation, alongside other significant benefits.