Making Science, the digital acceleration company, launches Trust Generative AI, a Generative Artificial Intelligence platform that solves two main problems on systems such as OpenAI: the difficulty of accessing non-public company data and the absence of updated information in generative models.
The launch comes as generative artificial intelligence, based on large language models such as GPT and Bard, aims to transform the way people work, taking companies to a higher level of automation and efficiency. A recent MIT study in which different professionals were asked to use Chat GPT to write reports, press releases and other business documents achieved savings of around 50% in report writing and improve report quality.
However, large language models such as GPT-3 cannot access non-public company information and do not have recent information stored. The GPT-4 language model was trained with data up to September 2021 and didn’t update information after that date. “The availability of large language models and generative artificial intelligence opens up a world of possibilities unthinkable just two years ago. In this sense, Trust Generative AI is a solution that creates a bridge between private and recent company data and large models generated with public information. This will undoubtedly facilitate the adoption of Generative AI,” explains José Antonio Martínez Aguilar, CEO of Making Science.
The platform (developed on Google Cloud Platform and OpenAI) optimises both data cycles and the training of artificial intelligence models. It validates the answers obtained using models developed on Google Cloud Platform on private customer data and more recent data that does not exist in models such as GPT-3 and GPT-4.
Therefore, companies maintain complete control over their data and can generate less generic content, more aligned with their policies, and style guides optimised with up-to-date information.
An example of the efficiency of Generative AI is how Trust Generative AI generated 100,000 product descriptions for the Ventis.it website in minutes and at a cost more than 10,000 times lower than traditional means of generation. The generated descriptions were more than 95% aligned with other descriptions generated by the editorial department and were produced using prompts optimised by Trust Generative AI.
Trust Generative AI can also generate all types of content, such as call centre conversations, FAQs or emails, images and other entities, connecting to services such as Google Cloud’s Vertex Generative AI, DALL-E2 and Stable Diffusion.
In an environment of rapid technological change and uncertainty about the use of technology, it is crucial to develop tools that help build trust in the use of technology. Trust Generative AI aims to play this role.