Reflecting on a transformative experience at the Imagination in Action AI Summit in Cambridge on April 18, 2024. Here’s what stood out:
Technology Commoditization: AI technology is now more accessible than ever, enabling businesses to implement out-of-the-box solutions like voice assistants without the need to develop them from scratch. The focus is shifting towards AI as a service, simplifying complex machine learning challenges into text prediction tasks solvable with LLMs.
But remember – as Igor Jablokov (Pryon CEO) provocatively noted, “The Internet died two years ago” emphasizing that SEO can only get you so far, focus on hybrid search engine technology & AI-aware marketing strategy.
Efficiency Gains: Time-to-task has been tremendously reduced. Yet, while it’s possible to buy a car online in minutes, AI chatbots plays a crucial role in helping consumers explore and understand the myriad of options available, saving days of research. AI chatbots are on the rise!
Advancements in AI Techniques: The shift from Retrieval-Augmented Generation (RAG) to Direct Prompt Injection (DPI), especially with Google Gemini 1.5, improves quality of responses, simplify design and as a bonus provides many ways to save on cost of systems running. Moreover, the diminishing returns on LLM fine-tuning, coupled with dropping costs, are reshaping the usage strategies. Never fine-tune your LLM unless there are no other options.
Hardware Innovations: Groq’s development (Dinesh Maheshwari, CTO) of Logic Processing Units (LPUs) showcases a significant leap over traditional GPUs, offering at the very least ten times the performance (and up to 100 times!) with low-latency responses.
Next-Gen Neural Networks: Discussions around Ramin Hasani’s work on liquid networks showcased advances beyond traditional LLMs, pointing towards a future of more dynamic and versatile AI applications.
And we’re proud to help the case with the exploration of biologically inspired algorithms and Natural Neural Networks (NNNs) performed recently by FoundAition.ai.
Now, on the downside, the summit only lightly touched on AI security—a crucial area that warrants more attention, echoing Raad Siraj’s sentiment that “cars need brakes to go faster”, hopefully we’ll see more on that next year. We know well how to use AI, but only learning basic ways to secure AI!
