The introduction of ChatGPT in 2022 lead to the development of the now continuing AI hype involving the Large Language Models and other types of AI technologies on the path towards the Generalized Artificial Intelligence or “AGI” for short.
This type of AGI is the pinnacle of artificial intelligence and is hypothesized to revolutionize almost all aspects of modern society as its capabilities to abstract and brain storm towards better ideas is crucial for developing and improving upon common ideas and innovations.
A brief history of modern computation
The introduction of the computer freed up human resources to allow for other types of contributions towards society including higher abstractions, innovative work on building more sustainable solutions and increased speed of innovation as higher computations and memory access became broadly available to the general public.
This allowed for increased availability of highly effective and non-fault computation of numbers which in turn allowed people to focus on other types of ventures including developing more advanced theoretical models and constantly higher mathematical abstractions nearing reality in the simulated world.
Therefore computers are one of the wonders of the world needing the dedicated efforts of human ingenuity to successfully bring forward the individual contributions towards increasing speed, precision and memory.
Introduction of memory and race towards zero
The first introductions of memory allowed computers to register the individual changes and attributions allowing small amounts of data to be stored on individual discs for later usage. This enabled increasingly more complex computations to be included within the computers and discs.
As the memory increased prices for storing data decreased rapidly allowing video and picture storage as well as compressible formats to be developed. This led developers to develop advanced programs capable of doing calculus, engineering and complex analysis. During the same time, the numbers of transistors increased rapidly following what is deemed Moore’s law. This showcased the number of transistors on a microchip leading to a now famous instance of how the evolution of available computing power has changed society and continues a path towards increasing speed.
This has been ongoing and is now reaching limits where individual computers are pushed towards the physical limit posed by the generic physical limitations of the materials involved as the distance between individual transistors becoming so small that quantum effects start to play a role in the exchange of information.
However before reaching this quantum level the available commute power which individuals have access to have increased substantially. This means that individuals are now presented with opportunities of utilizing advanced abstraction software such as large language models, neural networks and similar types of technology.
With this evolution, demand for computing power has surged towards new levels, as users are getting to understand and unlocking the power’s which lies hidden within large artificial intelligence models.
Large language models
With the recent advent of large language models catalyzed by the innovative company openAI it marked the beginning of a new era of computing allowing individuals with a comprehensive algorithmic model enabling advanced abstraction and reasoning in a level never before seen, meaning that each individual have available resources previously thought impossible.
The large language models requires an extensive amount of data and meaning that the individual users activity and attention is quite valuable for companies which can capitalize on the advanced methods used for training and validation of models on all types of different data streams leading to the aggressive harvesting of user data on a variety of apps such as massaging and imaging apps.
The location data and also the people who we interact with on a daily basis is one of the different types of data which is used to train AI and large language models. Which means that the individual users can have access to increasingly large and complex models demanding an exceedingly large amount of energy and water for storing large amounts of data and utilizing large complex models.
This in turns means that individuals and corporations in general benefits the most as they have access to models that are exceptionally advanced without having to pay for usage and subscriptions. The businesses have been built on venture capital meaning that individuals and corporations are drawing advantage on the risky capital currently being pumped into the developments of increasingly advanced models capable of high degrees of reasoning leading to increasingly adverse developments including coding, writing and processing.
One of the recent advancements is deducting facts from supplied or searched information online. This means that the model is able to process new information and utilize it within its currently developed and calibrated models feeding new information into the models.
Beginning of a new era of tech developments
With the advent of the official OpenAI ChatGPT model, it became obvious that a new type of technology race had begun which was the advent of artificial intelligence models aimed at developing a type of AI model which is the AGI models which can be utilized for developing new types of innovations, approaches and methodologies for modifying and utilizing the surrounding environments.
AGI – artificial general intelligence
This type of general artificial intelligence or AGI for short is difficult to obtain and in reality critiques of the approaches towards AGI and the possible dangers thereof argue whether it is possible or an ethical way towards achieving such goals. The technology, is still, to this day, far from reaching the AGI milestones and recently competitions for the most advanced model have been fierce with Google, X, and Claude to post their own models competing against openAI.
In particular, Google’s Gemini models have developed to a reasonable point of interest meaning that they are currently and quietly overtaking OpenAI’s first mover advantage by leveraging its broader access to data from a complete variety suite of services and apps.
This means, that Google will be responsible for developing one of the key players within the new order of technology including the general adoption of AI models in everyday tasks which are performed by office individuals.
Marginal occurrences and hallucinations
However, as the task space and particularly the marginal differences are continuously developing and it’s hard to capture efficiently during the individual evaluations of AI model outputs. The effects of these inconsistencies are hallucinations which sidetracks real people’s hard work and introduces inconsistencies in compliant heavy industries and thus can be a source of significant costs and risks to organizations.
These hallucinations are one of the main sources of errors from these types of computer models and one of the reasons why they are not being rolled out across all infrastructures and models. Instead they are actively used in common conversational type structures between individuals such that everyday pleasantries are more easily created. This means that the general well being of employees can be improved as a consequence.
These hallucinations will become an increasingly hard to overlook issue when the computer models are rolled out across tasks and industries in general society. Mathematically, these abstractions which forms the basis behind AI models, cannot become hallucination free since each of them will have residual approximations when imitating reality, thus they will have problems with providing consistent and accurate results in edge cases across use cases and possible utilizations.
The performance of the individual AI models which are utilized for applications such as heavy compliant rich industries will be disrupted last as the details between making and breaking it is highly ruled by minor inconsistencies. These inconsistencies multiply and can cause organizations to lose money, contracts and materials for the inconsistencies not to mention credibility for performing in accordance with general rules and guidelines set out in society for multiple reasons.
The difference between utilization of computing power across industries including the recreative possible uses such as games and image generation. These sort of use cases utilizes significant amount of computing power and thus resources in particular power usage that graphics cards are requiring when used.
The demand for data centers
As the models become more advanced the individual demands for computing power increases. This creates an increasing demand for graphical cards in configurations which requires complex data center infrastructures. These infrastructures are creating a massive demand for electricity and water since cooling and cleaning of the data centers is a complex task.
Furthermore, the individual space requirements for establishing such new buildings are difficult to handle and many stake holders find it difficult to figure out suitable locations where electricity grids, water availability and general surroundings create an opportunity to establish such data centers.
The solutions to these problems are plentiful meaning that no single problem is solved with a single solution. The general solutions is establishing long term sustainable centers where the infrastructure allows implementation and update of the graphics cards such that the individual users will have access to the recent advances in computing power without having to reinstall new centers from scratch.
Enhancing power usage and longevity
Another big opportunity for the industry lies in developing and contributing to efficient usage of power in the hardware infrastructure as the energy demands are significant in the general society.
Some companies tries and handles the contributions by reinventing the power components within the graphics cards, ensuring that adversarial effects such as corrosion are mitigated efficiently.
When the models are implemented efficiently and the graphics cards are utilized in a manner where individual users are still able to consume and utilize graphics cards to private usage then the supply is still sufficient to meet demand.
However, the prices on memory and graphics cards have increased significantly the previous years where breakthroughs within AI are sufficiently progressed to allow and enable the general public to utilize the available tools.
The Skjærup blueprint 
Leave a comment