“Generative AI and its possibilities”. (3/5). In a world where technology is constantly evolving, the world of companies is undergoing an unprecedented revolution. Major companies such as Meta, Google and AWS are even being overtaken by the lightning speed of generative AI, mainly thanks to the launch of ChatGPT. What does this mean for the company world? Is this really a change to come? Yes, without a doubt
Winds of change blow across the company technology landscape
The development of generative AI is the result of the convergence of three crucial factors. Firstly, it is based on the creation of a new AI model called “Transformers”, which emerged from Google’s research work. This model outperforms its predecessor, recurrent neural networks, thanks to its notable capacity for parallel processing. What’s more, it takes advantage of the ability to crawl and reorganize the entire Internet, enabling unprecedented access to human knowledge on a massive scale. Finally, this evolution is greatly enhanced by the advent of specialized processors, such as GPUs, which are ideally suited to the calculations required for AI. The company OpenAI has capitalized on these technological advances to offer a conversational agent for the general public: ChatGPT.
The potential of generative AI is vast: it can generate new content, such as images, text, 3D models and music. What’s more, they enable unprecedented understanding and organization of complex textual data, radically revolutionizing the way we interact with systems and documents.
How is Generative AI revolutionizing the company?
The emergence of this technology may lead companies to rethink the very structure of their information systems. Traditional office tools, customer service and the way people interact with applications will all be transformed thanks to generative AI. What’s more, thanks to ChatGPT, access to data relevant to the company is greatly facilitated and can be shared more widely, optimizing information release.
ChatGPT is already in daily use in many companies for summarization and reformulation tasks. However, the issue of confidentiality and the preservation of company data remains a major one, highlighting the need to raise employee awareness on this subject.
Although ChatGPT is only aware of information prior to September 2021, it can be combined with a search engine to take into account more recent information, or directly with company data, thanks to the RAG (Retrieval Augmented Generation) principle. However, this involves sending the company’s data to OpenAI, which again raises the question of confidentiality.
There are, however, limits to AI today. Hallucinations” or inaccurate answers, legal risk on training data, learning biases and ethics are among a few. In particular, significant computational resources are required for training, which entails significant costs and delays ($1.2m for GPT 3 training).
Towards open AI models with personalized learning for companies
In parallel with the success of ChatGPT, there is a move towards open AI models with user licenses compatible with commercial use, providing both greater transparency on how the algorithms work and customization possibilities for each company. It is becoming possible, at reasonable cost, to train models to integrate specific vocabulary or information from a particular field.
It’s already clear that generative AI will become a major competitive advantage for companies that operate efficiently with their text data. Even if the technical solutions for integrating generative AI into companies are not yet fully mature, it would be wise to identify certain actions to implement now, given the power and speed of the coming technological wave.
In conclusion, generative AI is a revolution that is dramatically changing the landscape of company technologies. More than just a tool, it is a new way of conceiving interaction with data, facilitating access to information, optimizing exchanges and customer services, while offering new possibilities in terms of automation and content generation.
Pioneering companies such as Generali and Orange have already begun to operate this potential for their calls for tender, or to adapt application interfaces for their maintenance technicians. This sends out a strong signal: generative AI is not only a technology of the future, it’s also now practically available.
It is therefore essential for companies to adapt and start considering AI training models (prompt engineering to learn how to interact with an LLM), to make the most of this technological revolution. This will nevertheless require remaining vigilant in the face of challenges, including the crucial issues of data confidentiality and bias inherent in AI.
Marc BORDESSOULE – Expert Data Science JEMS