The year of AI

20 Nov 2023

From the release of the beta version of GPT-3 in 2020, this reality seemed to be in a distant future. However, since then, the race to become the most powerful LLM model started, and the world suddenly has a new set of wide-ranging and ready-to-explore capabilities. These AI capabilities are evolving quickly, while the social impact of Generative AI (genAI) models is rising fast. According to Forrester, 61% of workforce will use their own AI to perform their tasks, introducing the concept of Bring-Your-Own-AI (BYOAI).  

But while the advances of this emerging technology are promising for many individuals, according to IDC, AI maturity at the corporate level remains low. Only a few have a firm determination and urgency to deploy an AI-Driven agenda, turning AI plan aspirations into something tangible and impactful, thus creating new value, and increasing productivity.  

Now, with the disruption of the genAI, a new enterprise AI-FOMO effect has emerged making it more crucial than ever to become an AI-First company matters. While the lack of talent and data enrichment may remain as inhibitors (IDC) may predictions from Forrester suggest that that budgets and use of AI platforms are positive and will triple to meet demand. 

Today, it seems that Large Language Models (LLM’s) are leading the AI game. While ChatGPT rapidly became the world’s most popular application of the most powerful language model of the world (GPT), the race has just begun, and other LLM’s, such as Bert, Claude, Llama, and Orca, have also become extremely popular in the last year.  

According to State-of-the AI 2023 report, the blossoming of these LLMs validates the power of proprietary architectures and reinforces the capacity of AI to learn from human feedback. The report now points out that efforts are growing to try to clone or surpass proprietary performance, through smaller models and better datasets. These efforts could gain new urgency amid concerns that human-generated data may only be able to sustain AI scaling trends for a few more years.  

The challenge with datasets

By now, you’ve probably heard of the term “data-exhaustion”, which refers to the difficulties that will result from the rising need for a huge variety of different datasets required to train these models. For models such as GPT-3 to keep understanding linguistic complexity, enormous quantities of high-quality training data are needed as the model’s size and complexity grow.  

Many authoritative voices underline that it’s unclear how long human-generated data can sustain AI scaling trends and some predictions suggest that available LLM’s will run out of data by 2023. This is where synthetic data comes in. Our team at Intelygenz are experts in generating synthetic data and know first-hand the challenges that it holds, among them there is a number one which is find balance between the quantity and quality of training data and the size of models. 

Many of our experts highlight the fact you can’t always find the ideal solution for your company off-the-shelf. We at VASS believe that the key to getting from hype to ROI regarding artificial intelligence is, among others, to dedicate the necessary time to train custom models. This is the cornerstone for implementing AI in your company that provides significant competitive benefits. 

With the use of the unique data from your business, custom models can be trained to increase accuracy and performance. They can make predictions and conclusions that are more accurate and pertinent to your company by learning from your unique data. This degree of personalization frequently exceeds what generic models can offer your business. That is why to succeed in an AI project implementing the right is key. 

AI Governance

As we incorporate AI into more and more many domains of daily life, the safety debate around AI has exploded into the mainstream, prompting action from enterprises, institutions and policymakers around the world.  

According to Forrester, “37% of AI decision-makers updated their organization’s existing policies for genAI and another 23% have implemented new companywide genAI policies.” 

GenAI has fueled this conversation at every major global event, while additional discussions have emerged to kickoff the establishment of world AI principles and a reliable framework. Notably, the recent AI Safety Summit served as a platform where participating countries came together to sign “The Bletchley Declaration”. This agreement symbolizes a collective commitment to build a shared scientific and evidence-based understanding of the risks that AI brings, as well as the respective policies across countries to ensure safety. 

For nearly a decade, VASS has been part of the AI revolution by making this set of technologies available through different value chain processes and company stakeholders. Hyperatomatization, Machine Learning, and Data Processing have been evolving since then.  

Today our solutions portfolio has been strengthened with deep-tech capabilities and cutting-edge AI-Gen technologies. VASS has already delivered cutting-edge projects across various sectors, including retail, banking, as well as others aimed at improving employer experiences. In addition, Intelygenz, a leading European expert in AI, joined VASS last year. With more than 200 AI experts, Intelygenz contributes to our pool of innovative solutions and knowledge every day. 

There is a long way to go in AI, but we at VASS are believers and supporters of establishing robust pillars for the development of a global AI governance. Our commitment is to contribute towards the greater good of society through responsible AI practices.