Over the past few years we have seen innovative and frenetic advances that are driving the field of Artificial Intelligence (AI) with great dynamism, surpassing our expectations and projecting a horizon of transformation and efficiency in a multitude of industrial and technological processes.
As we approach the threshold of 2024, this effect is being even more exacerbated with the pervasiveness of generative AI and its transformative potential that are evolving the way we relate to the digital world.
A direct effect of this change is the automation and efficiency of the most routine tasks related to content generation, resulting in significant savings in time and resources. As an example, it can be used to write emails, generate code snippets, assist in the creation of content of all kinds and even in technical matters such as advice on the creation of data models and code generation.
Within the area of Data Analytics, although it is true that the impact of Generative AI is in its early stages, it is clear that the main providers of Business Intelligence tools are developing their roadmap based on it. Tools associated with Google (Looker), Microsoft (Power BI) and Amazon (QuickSight) began to launch preview features based on Generative AI that not only help the data analyst in the development of reports in a more efficient way, but also help end users to acquire independence when exploring the underlying datasets according to specific needs. Prominent examples of this would be Duet AI in Looker, Copilot in PowerBI or the recent announcement of GenBI in Amazon QuickSight underpinned by Q functionality and supported by AWS Bedrock LLM’s. It is expected that by 2024 we will be able to see the strengths and weaknesses of these functionalities and an establishment of them after their massive use.
The next steps of Generative AI are overcoming conventional limitations. This technological disruption is undergoing a democratization driven by the combination of massively trained models, cloud computing infrastructure and open source philosophy, making it easily accessible to professionals worldwide.
Recently we have seen great advances in tools, algorithms and generative language models specialized in simulating a “mode” of expression, be it language, visuals or sounds, for example. However, the trend is towards “multi-modal” generative artificial intelligence. The integration of all these dimensions is going to be the most common form of presentation of upcoming Generative AI products.
The concept of pervasive AI will be further developed over the next year with the development of autonomous Generative AI.
“Autonomous Agents” such as AutoGPT (developed by OpenAI in 2023) are applications that effectively operate autonomously, generating and responding iteratively to prompts or instructions until a final goal is reached. Ideally, this type of functionality should proliferate in tasks such as content creation, data management or analysis where a certain degree of autonomy provides great agility in development.
In contrast, other variants are designed for cooperation with humans in the resolution of tasks and are proving to be tremendously effective, such as “Agent AI”, which provide great capacity for personalization.
One of the most important challenges in the coming years will be the evaluation of the effectiveness and performance of these products based on LLMs (large language models).
How effectively does it respond to diverse instructions, does it generate coherent, accurate and contextually relevant outputs, or are there frequent hallucinations that generate unwanted or incorrect responses?
Metrics are currently used for measuring performance on specific tasks such as translation, synthesis, text generation, such as BLEU, Rouge, perplexity or diversity. However, research is also being done on the evaluation of LLMs from other models in an autonomous way, which is currently a developing area of research and one in which there is great interest from the scientific community.
There are potential risks associated with the use of generative artificial intelligence for malicious purposes, such as the generation of fake news or deep fakes. These issues raise concerns about privacy and the spread of misinformation.
One of the main architects of the great impact of Generative AI during 2023 has been Sam Altman (founder of Open AI) who highlights that “If artificial intelligence goes wrong, it can go very wrong”. This phrase hints at the great relevance that the regulation of this technology by governments and institutions should have, which in turn should allow it to be accessible to the general public so that we can benefit from it at the same time.
>>Download the full document Tech4biz Trend 2024<<
Image: Unsplash | Julien Tromeur
Data Scientist in Keepler Data Tech: "Live full, die empty" defines my state. This becomes my lifestyle taking me out of my comfort zone and driving my voracious learning attitude about different aspects of Data Science. I love learning by teaching and am always open to new challenges that push me further my comprehension."
Data Analyst en Keepler. "I have a PhD in theoretical physics with a tendency towards generalist knowledge. This concern for learning has led me to work in different areas related to data, thus achieving a complete vision of the projects in which I work. Currently I am very interested in developing products in the cloud to streamline and improve decision making."
0 Comments