AI on the Edge: Recent generative AI research, trends and insights
Welcome back to AI on the Edge. We’re continuing our series on the most innovative artificial intelligence (AI) technology, with special attention to generative AI — its ability to model, simulate and generate content, designs, data and more makes it transformative across industries and experiences.
Yet with many positive advancements on the horizon, there are also significant obstacles. Since our last roundup, we’ve been diving deeper into expert opinions on challenges like harnessing massive amounts of computational power and ensuring AI democratization to reduce biases. We’ve also considered the impact of generative AI on jobs and the role of open-source technology.
We pulled these latest insights together for you so you can stay ahead of the trends, use cases and breakthrough innovations in generative AI.
Generative AI in 2024: The 6 most important consumer tech trends
The way we work and play will transform as our connected devices only get more sophisticated as they become integrated with AI capabilities. The next generation of consumer tech trends shows that buyers will be the next drivers of AI, pushing for interoperability and open ecosystems to make their devices more functional.
For example, smartphones might live up to their promise of being a true virtual assistant. Large language models (LLM) could be leveraged to craft creative content or personalized automated assistant responses based on your health, favorite activities or current location.
5 generative AI use cases impacting our lives
With original content creation being the main source of generative AI buzz, there’s still more to uncover as the possibilities are truly endless. Generative AI is poised to impact and improve major and minor life moments alike — from improved drug research for diseases to a virtual try-on session from a clothing store. Imagine a world where your car's or smartphone’s virtual assistant makes a dinner reservation for you, shares your dietary preferences and even orders an appetizer before you arrive.
How to enable efficient generative AI for images and videos
Generative AI is already creating stunning images and videos in a fraction of the time, but it’s not without significant challenges like high computation and latency, memory loss and data inefficiency. With new methods like the Clockwork architecture for image generation, token merging for video generation and HexaGen3D for 3D generation, Qualcomm AI Research shares new ways of producing state-of-the-art results in efficient generative AI for visual generation.
For example, a new technique called VaLID (Variable Length Input Diffusion) helps generate a new novel view of an object without significantly increasing the computation.
The rise of generative AI: A timeline of breakthrough innovations
AI began from decades-worth of brave scientists who pushed the boundaries of their time. Starting with Alan Turing’s 1960s work on generating human-like responses from a computer to the earliest version of ELIZA, generative AI’s earliest breakthroughs were incredibly monumental.
Now, there are near-constant innovations and emerging technologies pushing the boundaries of what’s possible, and generative AI is finally becoming an everyday reality. ChatGPT broke the world record in January 2023 for the fastest-growing platform in history, gaining 100 million users in a month, and the future of generative AI continues to appear near limitless.
Democratizing AI: Top 5 insights from Axios, Meta, Black Magic Design and our panel of industry titans
From enhancing creativity to open-source technology powering more innovation, AI can revolutionize jobs in multiple industries. However, experts are highlighting the importance of democratizing AI, making it accessible and impactful for people around the world. For example, Sy Choudhury, Meta director of AI partnerships, spoke about the benefits of open source: “It’s allowing the entire ecosystem of research, commercial companies and government institutions to be able to take something and improve upon it.”
Ensuring equal access, dealing with reduced job functions and providing necessary guardrails for misinformation are all important areas of discussion and research for the AI industry.
LLMs, MoE and NLP take center stage: Key insights from Qualcomm AI Summit 2023 on the future of AI
At Qualcomm Technologies’ recent annual AI summit, experts from Microsoft, Duke and Stanford Universities discussed LLMs, mixture-of-experts (MoE) models and natural language processing (NLP). LLMs require heavy computational power, memory storage and memory bandwidth, so new solutions are needed to better deal with these issues. Solutions are being developed to reconcile large models and memory-constrained devices. Other answers for AI challenges are discussed by these leading AI industry experts.
