AI Tokens: Cutting Through the Hype

Our social understanding of what Artificial Intelligence (AI) is, and what it will become, is constantly changing and evolving. 16 years ago, the Euphoria game engine was developed as the primary game engine for the George Lucas approved (but no longer canonical) Star Wars game (https://screenrant.com/star-wars-force-unleashed-rebel-canon-legends-starkiller/), The Force Unleased. While no real learning algorithms or functions (things we now commonly associate with AI) were implemented into the engine, the advancements of the tech were widely toted as the pinnacle of AI advancement for the time period, as we can see in this video demonstration from that year (https://youtu.be/6o6YVLmOs74). In the footage, we can see the AI Stormtrooper trying to avoid a fatal fall by grabbing onto an object. The experience became even more immersive when the stormtrooper began hanging onto each other, to a human appearing to mimic the behavior of cooperation, or in the most extreme cases, even self-sacrifice. Here we saw the early personification of AI and used it to justify a technological breakthrough.

Similarly, I recall reading an article in Time magazine during the early 2000s that explained the state of AI in warehouse management at the time (“AI” has been used for this purpose in unmanned warehouses for decades already, and without the need to mimic human behavior in the process, machines, when left to themselves, create their own efficient methods which may or may not appear useful to human observers but persist as they improve upon the objective. This means that if an AI-powered warehouse can find an object in 10 mins, that would take a human 20 mins to find in the same space, it does not matter if the warehouse can be efficiently parsed by a human or not, because the purpose of the warehouse is to remove human action anyway.

So how does this scenario relate to the AI that is changing the way we work and live? Well at face value, it doesn’t. In the case of the game The Force Unleashed, AI is a term being used to implement the mimicking of human behavior with respect to body movements and self-preservation. The feelings evoked by seeing this act played out in real-time, and without direct prompting from the user, gave the viewing human feelings that the AI is autonomous (even though it is not).

Today we are again reinventing what AI is. In the case of ChatGPT we aren’t using “AI” in order to maximize real-world space efficiency or access to items within a physical space. We are also not using “AI” to mimic human movements. Instead, we are using AI to mimic human responses to verbal prompts.

This shift in purpose finally does beg the question, is the progress and development of “AI” in fact iterative in the long term? Based on the evidence provided here, we can easily say NO. In fact, the focus of “AI” did not shift to “deep learning” as the core driver until post-2010 (per the Wikipedia timeline here: https://en.wikipedia.org/wiki/Timeline_of_machine_learning).

While the ideas behind each stage of AI development certainly share similar concepts, the application of the technology seems to be constantly shifting and evolving. This raises a very serious question regarding the validity of AI-themed altcoins in cryptocurrency and Web3.

Through the lens of modern AI writer assistants and image creators, the improvement in functions of AI, when compared to only a little more than a decade ago, can hardly be understated. ChatGPT (and similar applications) is proving to be very useful! But what about the limits of current AI? Well, it appears that even in its current form, AI still has some very practical limitations, especially when it comes to discrete mathematics.

This is disappointing when we consider (as Web3 developer DCbuild3r did in a recent Twitter post https://twitter.com/DCbuild3r/status/1636336407647068161?s=20) that the computational shortcomings in this area, do extend to areas of finite mathematics, limiting the usefulness of AI when applied to adjacent fields of emerging technology.

A final concern relates to the possible exclusion of minority opinions by AI. Though this is not a new concern, it is an important one, when we consider that widely held scientific positions have been reversed in the past.

In the end, The usefulness of AI is predicated on the basis of the social consensus providing the correct point of view (in most cases), the AI will not on its own defy overwhelming data that supports a specific conclusion, where humans often due this categorically and intentionally (in order to avoid groupthink or to benefit a disadvantaged group intentionally). While AI may be able to mimic this, If the issue is a direct product of backpropagation (the method used to produce the gradient data needed for the foundation of the AI https://brilliant.org/wiki/backpropagation/), it may be impossible to overcome without a new method entirely.