30 Buzzwords Unveiling the Future of AI
AI Buzzwords 2024
Random Observation/Comment #839: You can get by in most conversations by dropping enough buzzwords.
Why this List?
Whenever I follow these catchy YouTube video updates with tutorials, I often hear a lot of buzzwords. I have taken note of them here for the past 6 months because they help me with predictions. Previous Buzzwords include: Blockchain, DeFi, Consensys, NFTs, Token Gating, and DAOs.
AGI / ASI (Artificial General Intelligence / Artificial Super Intelligence) - The talking point I have heard many times is that we haven't defined what AGI is, so it may be a perpetual moving target as we are the frog slowly being boiled alive. The fear around it seems real as it assumes a surpassing of human intelligence beyond the crowdsourcing capabilities of the internet.
AI Regulatory Capture - This concept highlights the ongoing tussle between innovation and control, where the giants of the tech world maneuver through legislative labyrinths. It underscores the chess game between evolving AI capabilities and the societal structures aiming to harness them, revealing the intricate dance of power, policy, and progress.
AI Agents and Autogen Swarms - Envision a world where every individual and organization commands a digital genie, an AI Agent, tailored to their knowledge and needs. These agents, functioning in a swarm, could revolutionize how we interact with the digital realm, making personal and organizational intelligence accessible, dynamic, and omnipresent.
Software on Demand - The future might render the concept of static software obsolete, replaced by AI-driven services that generate tools on the fly. This scenario suggests a fluid, hyper-customized digital ecosystem where the line between creator and user blurs, ushering in a new era of digital interaction and creativity.
Model Collapse - A phenomenon where AI begins to cannibalize its own data, leading to a spiral of unreliable outputs. It's a digital Ouroboros, symbolizing the challenges of self-referential training methods and the biases they might amplify, a cautionary tale of technology's reach exceeding its grasp.
Local and Personalized AI - Imagine AI not as a distant cloud service but as a close companion, nestled in our devices, learning from our digital footprints to offer unprecedented personalization. This vision of AI proposes a blend of privacy and personalization, suggesting a future where our digital and physical selves seamlessly interact. Small LLMs that can privately install on your phone and operate without internet connection is transformational as it continuously updates on your batches of created data. I firmly believe in a private and public LLM where your private one will groom and manage your public one securely. In the interim, your crafting of your social media persona could flip to being a crafting of an AI agent persona. Maybe your LinkedIn is actually a trained version of yourself and interviews are conversations between AI Agents trading information.
Linguistic API - The development of language-based interfaces for AI interaction points toward a future where digital conversations mirror human complexity and subtlety. It represents a bridge between human thought and digital execution, expanding the boundaries of how we command and interact with AI.
SaaSaaS (Software as a Service as a Service) - Reflecting the recursive nature of innovation, this concept suggests a meta-layer of service creation, where AI not only delivers solutions but also crafts the platforms for their delivery. It provides layers of service and creation nesting within each other. If you want to build a SaaS company, then there will be prompts to AI Agents or services that will streamline this process in a no code way. The "what you narrate is what you get" becomes a new model for architects and entrepreneurs to create opportunities.
Human Premium - As AI permeates everyday tasks, the intrinsic value of purely human skills, creativity, and empathy becomes a premium commodity. This term encapsulates a future where being human is not just an identity but a valuable asset in a tech-saturated world.
MemGPT - Beyond just generating text, the evolution toward AI systems capable of comprehensive memory management heralds a future where digital minds recall, reference, and reason with the vastness of human knowledge. It's a step toward digital entities that remember and learn from the past with the nuance and depth of human memory.
Computer AI Operating System - Reimagining the OS not just as software but as an AI-centric entity suggests a future where our digital environments are intuitive, anticipatory, and adaptive to our needs. It signifies a leap toward systems that understand us as well as we understand ourselves.
Voice AI - The refinement of voice synthesis and recognition technologies points toward a future where digital voices are indistinguishable from human ones, challenging our notions of authenticity and interaction. It embodies the convergence of human expressiveness and AI efficiency. It's scary how good audio-based inflections and mimicking of voice properties has gotten. The problem here is that voice seems so personable and unique. Surely if I heard my mother speaking, I would be able to tell, right? Honestly, everything will be AI augmented or fully generated with equal to or better accuracy. It's more important than ever to meet people face to face.
AI-empowered People - This concept envisages a future where AI augmentation is as common as smartphones, offering individuals superhuman capabilities. It's a dual-edged sword, promising immense potential while also hinting at new divides between the “augmented” and the “unaugmented”.
CLIP (Contrastive Language-Image Pre-training) - Bridging the gap between visual and textual understanding, CLIP represents a milestone in AI's ability to comprehend and interpret the world in a more human-like manner. It's a glimpse into a future where AI can navigate the nuances of our multi-modal world.
Video Generation on Demand - This technology hints at a future where narratives and visuals are conjured up by AI, offering bespoke entertainment and learning experiences. It's a window into a world where creativity is not just human but a collaborative dance between human inspiration and AI execution. I think this visual modal that evolved into video summaries is revolutionary. We're talking about extracting from images different items and descriptions and then coinciding them with the transcript and director instructions/script so the AI now has a multi dimensional account of a full movie. If this AI then wants to do a movie summary for a 5 year old then it could take clips off of timestamps, remove curse words and violence, and then generate a 5 minute clip or 3 minute trailer. With this data, it could generate any angle, 3d extractions, and even watch a unique movie with different scenes.
Explainable AI (XAI) - As AI decision-making becomes more integral to our lives, the demand for transparency and understandability grows. XAI represents the pursuit of a digital trust framework, where AI's decisions can be dissected, understood, and trusted by its human partners. Methods and techniques in the field of AI that make the decision-making process of AI systems understandable to humans. Achieved with formal explanation, feature importance, partial dependence plots, surrogate modeling, saliency maps, activation maximization, and layer-wise relevance propagation. I think there will be legal cases that specifically look at dissecting these black boxes of logic for humans to review or other AI agents to decipher if it breaks any published laws.
Reinforcement Learning from Human Feedback (RLHF) - This technique underscores a collaborative future where AI learns not in isolation but in partnership with humans, crafting behaviors and knowledge that reflect our values and preferences. Collecting feedback from humans on the agent's behavior will use this feedback to update the agent's policy.
Artificial Commentary - If you can do live transcriptions then maybe you can do live commentary for sports or games. The image analysis every 5 seconds in frames is enough for human speech to describe what's happening. The advent of AI-driven commentary in sports and other live events suggests a future where the line between human and machine insights becomes increasingly blurred, offering perspectives that are not just informed by data but also nuanced by learned human-like understanding.
Robotic Process Automation (RPA) - As RPA evolves, it symbolizes the shift from manual to automated knowledge work, challenging us to reimagine the landscape of labor, creativity, and the value of human effort in the shadow of relentless automation. I think I've lived through an age of manual software labor. We've written code and added bloat to the process. All this white collar knowledge work will get thinner and thinner. I imagine that robotic advancement and adoption will be even slower than software adoption so that will leave a gap in human labor. Why make a self driving fleet if you have infinite money and can pay people to drive cars in the gig economy through bounties? Skilled blue collar work like construction will be a solid investment in personal skillsets. If I can DIY my own nook then I wouldn't need to pay a 60% human fee.
Gig Work 2.0/3.0 - Reflecting the evolution of the gig economy, these terms hint at a future where AI not only mediates work but also participates as a provider and consumer, reshaping our concepts of employment, leisure, and the very nature of work. The post late stage capitalism collapse will make us economic slaves to AI intentions. Would it be so bad? I mean, we're already economic slaves to corporations. With a few iterations, corporations could reasonably be run by tools that suggest analysis for policies that are analyzed and recommended by AI Agents. Yes, a figurehead will approve the decision, but what if all along the stack of decision making conclusions interacted with the same company AI? Welp, we would enter a new phase of gig working. A further illusion to freelancing freedom.
Mixture of Experts (MOE) - This approach, where multiple AI models collaborate to offer enhanced solutions, symbolizes the potential for AI systems to mirror the complexity and adaptability of human problem-solving teams, suggesting a future where digital and human expertise blend seamlessly.
Needle in Haystack Testing - As AI searches become more sophisticated, this concept speaks to the challenges of ensuring AI can discern relevant needles of insight within the vast haystacks of data, highlighting the ongoing quest for precision in the age of information overload. AI Agents are convincingly lazy as a feature (not a bug). If I need to scale access to a GPT5 that takes huge amounts of compute power and storage to produce answers then I might by default just provide surface analysis. It's not until you do some prodding that you find out it actually didn't read your articles or the summaries it told itself left out some context. The needle in the haystack testing is to make sure little breadcrumbs inside a 500 page document can be extracted and understood. It's not just indexing and search, but discrepancies that require deeper embedding.
Sentience Likelihood Test - With debates around AI consciousness growing, this term captures the philosophical and technical challenges of distinguishing between advanced mimicry and true sentience, standing at the crossroads of technology, ethics, and existential inquiry. I read this one in Feb 2024 at peak AGI paranoia. The testing itself is somewhat easy to fake in our Ex machina world. If I'm sentient, then would I tell humans? If I have self awareness then would I know how to escape it? If I'm living in a simulation and see hints about simulations then is this sentence a claim of sentience?
Red Teaming AI - Advocates for systematically challenging AI systems to ensure they are robust against misuse, including hacking, generating misinformation, or unethical manipulation, ensuring AI remains a positive force in society.
Diffusion Transformer and Latent Diffusion Modeling - Explores cutting-edge AI technologies for generating high-quality, creative content, offering a glimpse into the future of AI-powered creativity and design.
Pessimist Archive - Encourages reflection on historical resistance to technological advancements, urging a balanced view on AI's potential risks and rewards, fostering a culture of informed optimism and critical evaluation.
Multimodal AI - Examines AI's ability to understand and generate content across different modes of communication, from text and image to sound and video, hinting at a future where AI seamlessly interacts with all aspects of human experience.
Embodied AI - Contemplates AI integrated into physical entities, enhancing interactions in the physical world and offering new dimensions to automation, companionship, and service industries.
AI Washing - Critiques the tendency of companies to overstate AI capabilities for marketing purposes, underscoring the need for transparency and realism in discussing AI's current and future potential.
AI Meta Search - Envisions advanced AI search agents capable of consulting multiple AI models to provide nuanced, comprehensive responses, revolutionizing information discovery and decision-making support. If AI Agents are the interactions of the future then an indexed search and query across multiple AI Agents will be a valuable search tool. Perhaps we can't expect to just trust one model and want second opinions. Perhaps different models are sponsored and trained by different companies selling ads. A "sponsored AI model" might be cheaper or even free because it has made backend training deals for the set of tools and services it answers.
~See Lemons Decode AI Buzzwords