In 2023, Jasper and Lutz predicted that closed AI models, like those from OpenAI, would dominate over open models due to their superior performance and heavy financial backing. This largely held true, but we see that open models are catching up fast. Major players like Meta have open-sourced models largely to cut infrastructure costs, with examples like the LLaMA model alongside contributions from the Mistral team. We believe models are becoming commoditized and more prevalent across use cases and industries.

Next, our duo anticipated multimodal AI, integrating text, voice, and visual inputs, would transform human-computer interaction. While chat-based interfaces, such as those used by ChatGPT, have indeed become the primary mode of interaction, progress has been slower than expected. The integration of multimodal AI remains challenging, even though notable advancements from Google and promising developments from newcomers like Microsoft’s CoPilot indicate growing adoption.

Jasper and Lutz’s prediction of significant growth in MLOps (Machine Learning Operations) tools has come to fruition, with increased uptake of platforms like Databricks as organizations strive for effective machine learning deployment and management. The necessity of MLOps is underscored by the inherent limitations of LLMs (large language models), which require robust orchestration to ensure reliability and accuracy. This space will remain dynamic and evolving, with constant innovation and competition among tools.

Garnering less public excitement was the uptake of generative AI in Music. While public excitement around AI-generated music has diminished, AI tools are quietly revolutionizing music production, especially among indie artists and podcast jingle producers. The reduction in production costs has enabled a broader range of creators to produce high-quality content quickly, and major music labels and producers are experimenting widely with these tools.

The expected surge in AI startups has happened, though their success hinges on effectively integrating AI to solve specific problems. Many startups often add AI layers to traditional products without solving any fundamental issues, thus struggling to stand out.

For AI hardware, media and specialized companies were expected to make strides, potentially challenging Nvidia’s dominance. However, Nvidia continues to dominate the AI hardware market, bolstered by new product releases like the B200 GPU, which offers significantly improved performance and energy efficiency. While the prediction of an emerging hardware competition is valid, Nvidia remains the clear leader with its comprehensive ecosystem and developer support.

Contrary to fears voiced in some quarters, AI continues to thrive with sustained investment and development, avoiding an anticipated AI winter (a period of reduced interest and funding in AI). Interest and innovation in AI remain strong. That much is unambiguous.

Current efforts focusing on improving AI's specialized capabilities rather than achieving human-like general intelligence mean that Artificial General Intelligence remains a distant goal. While the AGI conversations persist, often driven by high-profile figures like Elon Musk, practical advancements are minimal.

Tune in to The Edge, our very own podcast breaking down the latest AI developments, to explore these insights and more: