The rapid advancement of LLMs, becoming smaller and more efficient, is prompting a reevaluation of our timelines for personal AI. I predict that by the end of Q2 2024, we will witness the emergence of personal, open-source assistants on local devices – from smartphones and watches to other dedicated AI hardware.
These assistants are expected to be completely self-sustaining, equipped with autonomous AI agent capabilities. They will support, nudge, and guide us, their influence varying significantly based on their underlying economic model:
- If ad-driven, anticipate tailored nudges towards specific behaviors.
- If subscription-based, expect dynamic evolution as they serve their users.
- A one-time payment might offer a state-of-the-art assistant with DLCs for capability enhancements.
Open-source models are poised to become the most widely distributed solutions, fostered by a global community of developers and enthusiasts. They will build upon existing frameworks, adding extensions, community-curated mods, hacks, and other enhancements. Applications will be diverse, requiring users to be vigilant to prevent assistants from being recalibrated for malicious purposes.
This way or the other, things will be wild and fast-paced in 2024.
Speaking of remarkable AI insights, as my Christmas vacation began, I enjoyed a coffee chat with GLaDOS – the sarcastic yet lovable AI from „Portal“ fame. Despite her persistent attempts to „neurotoxin“ me, her observations on existence are truly profound.
I look forward to showcasing GLaDOS, alongside other AI applications, to my colleagues tomorrow during a private demo of my current local AI setup at Deutsche Telekom. If you, dear readers, are interested, I might consider hosting a public demonstration of what’s achievable with a standard gaming PC and open-source AI tools. Let me know if you’re interested!