2024: The year Generative Artificial Intelligence triggers the transformation of unstructured Data into Action.
Alberto García Arrieta
Leading Data & AI in Spain and Portugal at Accenture | Driving business reinvention through Data Transformation and Artificial Intelligence
In 2023 we witnessed how AI has revolutionized the way we interact with data allowing, finally, to a technology that will be human by design. Here are six innovative highlights from 2023 that I believe will continue in 2024 :
Applications powered by AI.
As already noted by Accenture's Technology Vision 2023 when the year was just starting, generative AI's potential in augmenting human capabilities is huge.
Throughout the year, we've seen the emergence of numerous applications designed to augment our productivity and creativity. Yet, widespread adoption, particularly in mainstream applications and enterprise systems, remains limited. This lag in adoption does not help technological maturity as they are two facets of the same coin.
Looking ahead to 2024, we can expect a surge in Generative AI features across a broad spectrum of Enterprise SaaS systems, extending beyond the likes of Microsoft’s Copilots or Google’s Duets. One of the challenges will be to smoothly transition our work processes to accommodate and leverage this expanding realm of possibilities.
Currently, the personalization level of these applications is elementary compared to the full capabilities of AI. However, the strategic use of intelligent coaching, tailored to each employee using curated data, can be a strategic differentiator in these early stages.
Let's also think beyond mere efficiency. Imagine creating an AI personality that not only captures your customers' attention but also serves and sells your products and services with the expertise of your best sales representative. If you haven't started developing such a customer-focused AI coach in your industry, it's highly probable that your competitors are already on it. Developing this unique AI personality for customer service could become a distinct channel, and the sooner you start, the better.
This trend of integrating advanced technologies is often first seen in the video game industry. We are about to experience a new generation of video games that don't just entertain but also pave the way for real-life coaching experiences. These games will also provide new training datasets for our data-thirsty Large Language Models (LLMs). When pondering over rumors of GTA VI for 2024, I recalled reading about NPCs (Non-Playable Characters) powered by Generative AI in the book "AI 2041". It's a sign that what was once fiction is becoming a reality.
AI models' accelerated evolution.
In 2022, Artificial Intelligence was not as advanced as it is today, with the emergence of increasingly powerful models each month throughout 2023.
I am have left out a few models such us Falcon, Bloom or Mistral but it is impossible to mention all of them. I believe we have all appreciated the growth of open models this year. Several of my colleagues suggest exploring open-source models as the initial option for various applications, prior to considering their commercial counterparts. I hope we also start to see more and more transparency in the data used for base training ensuring conformity with intellectual property rights.
While new modalities start their stabilization process and we learn their true usages, I anticipate the emergence of not just smaller but also sector-specific, industrial, and geographic initiatives aimed at enhancing model performance, potentially through the fine-tuning of models on specialized datasets.
The art of asking the right questions.
Parallel to the the fever for new artificial intelligence models we learned to use existing models to create solutions through prompting. But it is not as simple as finding the right words… with fantastic resources like GitHub - microsoft/promptbase: All things prompt engineering we continued to hone prompt engineering techniques.
Recommended by LinkedIn
Prompts would never work right at scale without frameworks for NLP applications, now adapted to LLM models. These frameworks have evolved tremendously during 2023 enabling, not just inference and serving, but easing deployment and LLMs usage.
In the future, the focus of AI frameworks is expected to pivot towards benchmarking and testing applications, including synthetic data, a critical area for continued innovation at scale.
Unlocking dormant enterprise data.
We've progressed beyond using just structured data, now harnessing unstructured data through Vector and Graph databases within RAG architectures. Recent months have seen us integrating advanced protection techniques into our corporate applications to safeguard against hallucinations and cyberattacks.
In 2024, our focus will shift more towards leveraging unstructured data in corporate settings to bypass current model limitations and scale AI effectively. Prioritizing data transparency is essential, ensuring our customers and employees understand the sources of information used in our AI applications to build trust.
Usage of cloud Software and data platforms to unlock innovation.
Major technology companies and data platforms specialized players have made significant strides in creating products, services, and platforms to support generative AI applications.
In the coming year, we can anticipate further improvements, particularly in how Data Management adapts to evolving data paradigms. The opportunity to leverage all these services for effective Data Management and operational readiness of Artificial Intelligence based applications is immense.
However, we should not underestimate complexity of relying our applications on this new generation of services. Best practices include ensuring avoidance of strange unproved usages of services, managing the potential for cost overruns, and enhancing the resilience of our systems to prepare for potential service disruptions due to ongoing changes in the underlying platforms.
Risks that deserve our attention.
The journey of AI in 2023 has been a balancing act between embracing innovation and mitigating risks. We've observed the AI community tirelessly working towards developing more sustainable models and hardware. Governments worldwide have stepped up, crafting laws and policies to regulate AI's expansive reach. Security firms have evolved their tactics to shield against AI-powered threats, while private entities are diligently constructing ethical frameworks.
To harness AI's full potential, we must scale its usage. Rethinking our enterprise architecture, with these technological advancements at the forefront acting as a new set of cognitive systems that will compliment operational, engagement, informational and corporate sets of systems. Our approach should include democratizing AI access, ensuring data quality and security, and fostering a reinvented user and employee experience. Integral to this process are robust testing and benchmarking layers that uphold ethical principles and human rights.
Looking ahead.
The coming year promises to be a landmark period in our ongoing journey to leverage data effectively. With advancements in applications, models, architectures, and risk mitigation strategies, 2024 is here to set new benchmarks in the AI landscape.