Artificial Intelligence (AI) has been around for decades, but recent breakthroughs in large language models (LLMs) have reshaped its role in the enterprise and consumer world. According to Bobby Blumofe, EVP and CTO at Akamai, “AI has been around for decades, but I break it down into three distinct apexes. The first apex saw AI as largely irrelevant. The second apex, around 2014–2015, marked the dawn of AI life with the advent of deep learning. The third apex, about two years ago, was the Cambrian explosion, driven by generative AI (GenAI) and large language models.”

As AI continues to evolve, businesses are leveraging it to drive innovation, efficiency, and competitive advantage. However, running LLMs can be expensive and complicated. Blumofe notes, “The big lesson that everyone should recognize is that you can now accomplish much more with much less. Large models can be run, built, trained, and used for inference far more efficiently than before.” This shift in efficiency has significant implications for businesses, as it enables them to deploy AI solutions that were previously cost-prohibitive.

Businesses want real solutions, not chase trends

Despite the hype surrounding AI, businesses are looking for solutions that solve real problems, not just follow trends. Blumofe emphasizes, “Enterprises are looking for AI solutions that solve real business problems, not just follow trends. The LLM is not the answer to all things AI.” Ideal use cases for AI in business include natural language interfaces, summarizing and reformulating information, and image recognition. For instance, a business might use AI to develop a chatbot that helps customers navigate their website or to analyze large datasets to identify trends and patterns.

However, Blumofe cautions, “For most enterprise use cases, a small, specialized, highly capable model is a far better solution. You don’t need—or even want—an ‘Ask Me Anything’ model.” This is because LLMs can be overly complex and may not provide the level of specificity and accuracy that businesses require. Instead, smaller, more specialized models can be trained to perform specific tasks, such as language translation or text summarization, with greater accuracy and efficiency.

Helping make AI more accessible to businesses 

Akamai is helping businesses deploy AI without facing heavy costs and challenges associated with scalable infrastructure. Blumofe explains, “Our goal with our cloud offerings is to provide flexibility through cost-effective, highly scalable, high-performance, and highly distributed infrastructure. We’re focused on protecting all aspects of your application infrastructure and data—including LLMs and the unique types of attacks that they are susceptible to.” By providing a secure and scalable infrastructure, Akamai enables businesses to deploy AI solutions with confidence, knowing that their data and applications are protected.

As AI continues to evolve, open source models are becoming increasingly important for businesses. Blumofe notes, “Open source has helped in a lot of ways. It means that you can own the model that you are using, and that comes with a lot of benefits. You can run it on-prem, on your own device, or in any cloud.” Open source models provide businesses with greater flexibility and control over their AI solutions, enabling them to customize and adapt the models to meet their specific needs.

Open source helping adoption of AI across businesses

Moreover, open source models can help to address concerns around data privacy and security. By using open source models, businesses can ensure that their data is not being shared or used in ways that they do not control. Blumofe emphasizes, “If part of your application needs to recognize images, then convolutional neural networks are your answer. But if you need reliable reasoning, we have symbolic AI, knowledge representation, and a robust corpus of AI focused on planning and reasoning with symbolic representations of knowledge.”

In addition to open source models, hybrid reasoning models are also becoming increasingly popular. These models combine different AI techniques, such as reinforcement learning and chain of thought prompting, to improve the performance of LLMs. Blumofe notes, “Hybrid reasoning makes some sense—You get better answers with one of the multiple techniques. But I think where the term hybrid is actually going to be more relevant is the notion of creating vertically focused solutions using an ensemble of technologies.”

Take away 

With AI continuing to reshape industries, enterprises that leverage Akamai’s solutions will be better positioned to innovate and scale efficiently. As Blumofe succinctly puts it, “AI’s future is now, and Akamai is making it happen.”

By understanding the ideal use cases for AI, leveraging the right infrastructure and expertise, and using open source models and hybrid reasoning techniques, businesses can unlock the full potential of AI and drive innovation forward. With the right approach, AI can help businesses to improve efficiency, reduce costs, and drive revenue growth, ultimately leading to a more competitive and successful organization.

Guest: Steve Winterfeld (LinkedIn)
Company: Akamai
Show: An Eye on AI