Artificial intelligence and machine learning has a history longer than we might realize due to the recent explosion of progress on the field. Levente Szabados, co-founder and senior advisor of Neuron Solutions, and David Taylor co-founder of Promptmaster, provided deep insight into the history of artificial intelligence and of AI audit which can guide companies to successfully transform their businesses to incorporate machine learning models and artificial intelligence.
The evolution of AI in businesses
Levente explained that AI is not a single solution for arising problems, but a collection of still evolving technologies that gain space in waves.
The first of these waves was as early on as the 1950, but with a very limited scope to start out with. Soon after, a bigger wave brought with itself systems that were more widely utilized but each of these systems was tailored to a specific problem, so these were rather rigid.
After this, the spread of machine learning caused the next wave, in which a rigidity was still present as individual machine learning models were created for individual tasks.
In the third wave, machine learning models could solve a broader range of problems including language processing and machine vision, but also in a discriminatory manner. The specified output determined the whole model.
Currently, we are in the fourth wave of machine learning with generative AI solutions. This technology is a leap towards a general purpose technology due to its ability to adapt to different situations, which require some degree of creativity.
The urgency of adopting AIs
Levente revealed that basically there is no field of business left untouched by artificial intelligence. One of the reasons for this is the evolution of language models, for language is universal for human interaction and is omnipresent in businesses. The other big field of utilization is machine vision. These can even be applied in agriculture.
So far, humans could only communicate with machines via programming languages which are various levels of abstraction of machine code. Now however, with the help of large language models and prompts, machines and humans can communicate directly via the English language. This might seem like the apex of these machine learning waves, but even bigger waves might follow. Whichever the case might be, all of these waves are incorporated into the technologies of today and the field we call AI and previous waves are still in use, only less conspicuous.
Companies failing to utilize artificial intelligence will most likely face a significant disadvantage in the market.
Companies and generative AI
Before generative artificial intelligence, companies that wanted to adopt some kind of machine learning model into their operations had to collect specific and relevant local training data for a specific process and then deploy a machine learning model.
Generative models offer a much more flexible option due to the fact that these models are pre-trained providing an existing system for an area, pushing the previous process into the background.
The technology used by companies might change or improve, but the ultimate goal of any business is to increase its enterprise value and one of the best ways of doing so is making their processes as efficient as possible, often through automation. The automation of these processes could be achieved in previous waves, with older technology, but the cost of this transformation and maintenance afterwards would have outweighed the possible gains in smaller businesses until recently. Nowdays automating company processes with AI leads to significant savings not only for large corporations.
The gold rush of generative AI
Levente thinks there are two layers of this technology of artificial intelligence. One layer being a foundational model provided by large corporations such as Microsoft or OpenAI or Amazon which are the building blocks, the base for customizing or utilizing particular tools that build on the foundation models and provide a more specific solution for a task.These niche applications are the second layer.
It is important that foundational model providers not only deliver their generative model, but also describe and encourage different use cases of their product, otherwise risking their product being left underutilized. This also leads these providers to remain small in numbers because their established generative models are used by smaller companies, who specify the particulate problem. In this way, the majority of AI businesses basically provide wrappers for the foundational models that determine the user experience by turning a generalized solution into a problem specific solution. Companies that provide the best user experience will be the winning participants of the market and the easiest areas to deploy these are customer support, marketing, operations management and software development, explained David.
The challenge of incorporating AI
One of the major challenges is for corporate decision makers to gain enough knowledge to assess the gap between the theoretically achievable and practical solutions. This knowledge includes the processes specific to the given corporation, like how those processes work, how they can be automated, an ability to overview the patterns of the company, Levente pointed out.
David explained that the learning process of corporate management is similar to that of individuals and consists of four stages, from which most only made it to the second stage so far. The first stage is knowing about the existence of a technology, but nothing besides. From this stage they need to move towards the second, which is a familiarity with the subject so they can start asking questions and making informed decisions. The anxiety of falling behind or turning business processes upside down is ever present between corporate leaders.
Levente explained that the AI audit is the product of the previous wave and it provides a methodical way that enables companies to share and identify the pivotal points of a project in which the company wanted to apply machine learning models.
Three things are needed to build a successful project. First of all, the technology must be available, which is provided nowdays. The second thing is data capital, which is the accumulation of project specific data which can be used to train a model. Lastly, some kind of pain point must be located. These are the pivotal points which encourage companies in the investment of machine learning models.
The purpose of the AI audit is to find data points that can be tuned into profit, David added. Generative models cut down on the time and cost of developing a custom model, but the methods of AI audit can still be applied even to small businesses and can shed light on the correct specialization of these generative models by understanding jobs of the employees, the work process of the company.
An AI audit that required for example, a 100 hours of work in the previous wave takes only 10 hours in the current one. This not only reduces the time needed, but also the costs of implementing machine learning systems thus making this service and technology available even to middle sized businesses.
The way companies can improve their innovation
Companies and decision makers of companies don’t have to commit to in-depth studying of artificial intelligence and machine learning to help the implementation of these systems into their businesses. The more important kind of learning they can accomplish is experiential learning, in which the mistakes made along the way during experiencing the new technology help to improve usage of those technologies. This requires analyzing and reflecting on past experiences within the given subject.
The task of knowing the exact current technological state of the world is impossible due to the fact that by the time one can assess this technological state, it is already outdated. The best companies can do is to employ people dedicated to learning of innovations, experimenting with these and how these apply to their company will always have an advantage due to a constant state of innovation.
In conclusion, for businesses and individuals, learning by trial and error is one of the surest ways of keeping track of technological progress and learning to adapt to various situations.
The original article is available by this link.
Follow us, for we will have more themes to explore soon!