It's Time for a Data Reality Check

AI Gets Real: Infused with Trust and Explainability

 We can still hear echoes of the initial thunder over ChatGPT.

Bloomberg Intelligence  estimates  that  Generative AI could be a $1.3 Trillion market by 2032, growing at a CAGR of 42%. In 2023 alone, nearly $50 billion was raised by Generative AI and AI-related startups, per  Crunchbase  data. But the honeymoon phase may be ending. The flood of investment dollars that poured in at the outset has settled into a steady but rationalized flow. Now, business must get past uncertainties, fears, and skepticism — and making these big, ambitious concepts work in real-world situations.

ChatGPT, and large language models like it, are primarily geared toward increasing productivity. They can spin impressive yarns but lack the depth for real-world applications and the ability to be adequately meaningful in a specific domain or customer context. And, with widespread examples of data hallucinations as well as risks of data poisoning and fraud, they also lack the trustworthiness needed to drive real adoption by business users and leaders.

To truly harness GenAI, applications need to be “grounded” with specific context and data. This might involve pointing the AI to relevant and unique sources, validating, and providing visibility to data lineage, providing additional clarity around desired outcomes, establishing protections to safeguard data, or even building custom LLMs trained on specialized data. The goal: accurate, explainable results you can trust. The trustworthiness of outcomes driven by data depends on the ensuring the quality of the data itself (garbage-in, garbage-out), the quality of the processes that manipulate and interpret this data, and of course the quality of the infrastructure that stores, secures, and supervises this data.

Edge and IoT: Surviving the Data Avalanche

The lines continue to blur between the physical and digital worlds. 

Forbes estimates that the number of IoT devices has more than doubled in recent years, from 10.3 billion in 2018 to 25 billion in 2025. A convergence that massively expands our understanding of our surroundings and paves the way for truly immersive experiences. That’s the good news.

The challenge is that this data avalanche makes it harder for true insights and value to come through, let alone manage. Much of that data is generated at the edge, where challenges with connectivity, security and scalability become obstacles towards value.

One solution lies in edge AI: the coupling of edge computing with the recent innovations in AI effectiveness. 

The deployment of AI algorithms and applications in edge devices allow for filtering, processing, and refining of data closer to the source, rather than in a private data center or cloud computing facility. This approach can increase edge intelligence across a diversity of inputs, increase availability and reliability, improve real-time response through reduced delays and system overloads, reduce costs of networking, and increase security and privacy of the data. At the same time, cloud computing can support and supplement such edge deployments of AI by running the AI model during training and retraining, managing the latest versions of the AI model and application, and processing more complex requests.

From Cloud-First to Cloud-Smart

For years, the cloud was touted as the be-all and end-all for companies’ IT needs. But that cloud-only or cloud-first approach has run into its own wall of reality. 

IDC data indicates that 70 to 80 percent of companies are repatriating at least some of their data back from the public cloud. This includes both large enterprises that made wholesale migrations to the cloud without sufficient preparation, as well as cloud-native startups that have achieved scale.

Said differently, the future of IT infrastructure lies in the balancing of workloads between the public cloud (even multiple public clouds), on-premise and co-location environments. Behind this Great Rebalancing are concerns about data transfer fees, security and privacy, data sovereignty and country-specific data residency laws, and performance considerations.

The need of the era is to be cloud smart. 

Where data sits requires a thoughtful approach and depends on a number of factors: the nature of the applications, the contents of the data, the profile of its users, and the requirements and constraints of the geographies involved, among other things. IDC data indicates that more than 50% of investments in GenAI projects in the near term are being allocated to digital infrastructure — which makes sense, since a hybrid cloud foundation supports robust data management, and in turn enables more effective data use through powerful AI applications.

Cloud-like consumption will also be a part of that story with a continuing drumbeat towards "everything-as-a-service". Tech leaders who embrace as-a-service (aaS) models in 2024 stand to gain significant advantages in terms of agility, efficiency, security, and cost optimization. Not to mention competitive differentiation in an increasingly digital world.

Overcoming the Carbon Footprint of AI

According to the International Energy Agency (IEA) and 8 BillionTrees, data centers today generate nearly 4% of global greenhouse gas (GHG) emissions — more than the GHG emissions from the aviation industry. Driven by AI, the Synergy Research Group reports that hyperscale data center capacity alone will almost triple in the next six years. 

Gartner believes that AI may consume more power than the human workforce by 2025, offsetting carbon zero gains. With the increasing adoption of AI and growth in complexity of machine learning models, the consumption of data, power and compute resources will only grow.

In the face of all this, the good news is that AI can also be part of the solution. Sustainable AI practices can create a real dent and drive improvements in efficiency, even overcoming the footprint of AI. These include hardware optimized to reduce energy consumption, federated learning, and energy efficient coding. Gartner indicates that this impact could even be between 5 and 10 percent reduction in carbon dioxide emissions.

In 2025, companies will increasingly wrestle with the balance between doubling down AI and digital investments, and stepping up their focus on minimizing their environmental impacts.

The choice of storage and infrastructure hardware providers is crucial for companies planning their digital transformation investments: the right choice can determine the footprint of these companies for years to come. In addition, there will be additional data center investments in renewable energy such as solar panels and wind turbines, cooling techniques such as free cooling, and managing server utilization during off-peak hours.

The road ahead for 2025 may look daunting, with this swirl of intersecting and often competing forces creating a fog of uncertainty. Even so, the possibilities outweigh the challenges, especially when bold investments are grounded in reality.