Reflection on AI, you ready to deploy it?

1. "Harnessing AI's Power: The Responsibilities We Can’t Ignore"

"With great power comes great responsibility."

This well-known Voltaire attributed quote popularised by Spider-Man more recently, captures the essence of the discussion around the evolving landscape of artificial intelligence (AI). While AI advancements offer immense opportunities, you now know it also brings significant ethical, environmental, and practical responsibilities that mustn't be overlooked. At the last Gartner Symposium in November 24 everyone was asking “how do we scale this? How do we demonstrate value and communicate the fact it’s really costly?”. People are able to recognise the real power in AI but are still not quick enough or rigorous enough to grasp the difficulty in deploying it.

As AI technology (traditional machine learning and the emergence of transformer models) grows increasingly powerful and accessible, organisations need to approach its development and deployment with careful thought and responsibility. Too many people fail to recognise that it’s much harder to deploy, scale and deliver AI investments and deliver real value confidently than they realise.

2. "Composite AI: Revolutionising Industries, One Solution at a Time"

The realm of AI is experiencing a profound transformation, underpinned by the evolution of composite AI and the rapid progress of large language models (LLMs). These developments herald a promising future for AI, albeit one that demands vigilance regarding ethical, environmental, and practical concerns. And remember LLMs are only relevant in about 1 in 5 use cases, the rest should be addressed with classic machine learning approaches.

Composite AI marks a shift in how AI is leveraged across various industries. By integrating diverse AI techniques such as machine learning, natural language processing, and knowledge graphs, composite AI enables the creation of more adaptable and scalable solutions. This integration not only enhances predictions and decision-making but also automates complex tasks, allowing businesses to fully exploit AI's potential in various settings. Once you learn how to combine different methodologies, your organisation can mitigate the risk of failure linked to relying on a single technology. Thus, we’re seeing composite AI poised to become a cornerstone of future AI architectures, helping organisations overcome the limitations that have traditionally constrained AI models.

3. "Beyond GPT-4: The Expanding Frontier of Large Language Models"

In tandem, the field of LLMs has witnessed breathless advancements, especially with models surpassing the once-standard capabilities of GPT-4. OpenAI’s ChatGPT is just one model, too few people are familiar with equally capable alternatives like Anthropic’s Claude, Perplexity, Mistral, Meta’s Llama and Google’s Gemini. Examine the graph-like capabilities on offer with Notebook LM. Innovations such as extended input context lengths and multimodal functionalities have expanded LLM applications to include text, images, audio, and video. This expansion heralds a new era of multimodal and prompt-driven applications, making the development of interactive solutions more feasible. Nonetheless, the challenge lies in ensuring that these generative AI tools are reliable and seamlessly integrated into organisational workflows, necessitating a sceptical and judicious approach when evaluating vendor claims. The excellent Simon Williamson’s blog (https://simonwillison.net/2024/Dec/31/llms-in-2024/#synthetic-training-data-works-great) succinctly summarises the issue here :

“LLMs are power-user tools—they’re chainsaws disguised as kitchen knives. They look deceptively simple to use—how hard can it be to type messages to a chatbot?—but in reality you need a huge depth of both understanding and experience to make the most of them and avoid their many pitfalls.”

4. "Balancing Act: AI’s Growing Accessibility and Environmental Concerns"

The reduction in costs and increased accessibility of LLMs have democratised their use, bringing sophisticated AI capabilities to a broader audience. However, this is not without its challenges. The environmental impact and energy demands associated with AI infrastructure development remain growing concerns. As AI continues to spread, the construction of data centres and the need for energy-efficient solutions become critical issues - we don't have the generation capacity to keep up and by the way we're consuming far too much fresh water for cooling in the process. Companies such as Nvidia are responding with the development of smaller edge devices like Jetson and mobile phone manufacturers are embedding powerful CPU cores in their phones to run small AI models, but the balance between technological advancement and environmental sustainability remains a delicate one. As energy costs in Europe grow and demand for fresh water follows suit the environmental question is going to be an increasing problem you'll need to address.

5. "Strategic AI Integration: Navigating Risks and Opportunities"

Simultaneously, AI engineering and knowledge graphs are gaining significant traction. AI engineering offers the framework necessary for deploying AI solutions at scale, helping enterprises transition from isolated projects to comprehensive AI portfolios. Techniques like DataOps and ModelOps are instrumental in ensuring smooth, repeatable deployments.

Knowledge graphs like Neo4j provide a visually intuitive means of representing complex relationships within both physical and digital realms. This capability complements the predictive nature of generative AI, offering logic and explainable reasoning essential in applications where transparency and reliability are paramount. Check them out and become familiar with them because they offer a huge opportunity in a number of business use case scenarios.

While the potential of AI is unmistakable, it is crucial for organisations to adopt a strategic outlook in its adoption, considering non-technical aspects such as governance, risk management, and ethical implications. Navigating data privacy laws, ensuring data accuracy and completeness, and managing inherent biases in AI training data are vital steps in realising AI’s full advantages.

Ultimately, exploring AI's transformative power is as intricate as it is promising. By carefully integrating composite AI techniques and adopting advancements in LLMs with caution and responsibility, you and your organisations will be in at the forefront of this evolving technological landscape. However, they must remain cognisant of the substantial responsibilities that accompany such power. Indeed, "With great power comes great responsibility," a principle that holds true as AI continues to shape your executive boards, shareholders and citizens expectations alike.

For further information check out our latest research: https://www.gartner.com/en/articles/hype-cycle-for-artificial-intelligence#:~:text=GenAI%20has%20passed%20the%20Peak,standardized%20processes%20to%20aid%20implementation.