More Gemini Features Coming to Google TV
Google TV integrates Gemini AI for image/video generation, enhances Google Photos, and adds a dedicated YouTube Shorts feed to the homepage.
Read moreConversations about artificial intelligence typically revolve around technology-centric topics: machine learning, conversational interfaces, autonomous agents, and other aspects of data science, mathematics, and implementation. However, the history and evolution of AI is not just a story of technology. The story of artificial intelligence is also intertwined with waves of innovation and research breakthroughs that first encounter economic and technical barriers. There seems to be a continuous pattern of discovery, innovation, interest, investment, cautious optimism, boundless enthusiasm, realization of limitations, technical barriers, waning interest, and AI research retreating back into academic backgrounds, much like the ebb and flow of waves on the shore.
This cycle of interest, investment, hype, followed by decline, rinse, and repeat is particularly frustrating for technologists and investors, as it does not follow the typical technology adoption lifecycle. Popularized in Geoffrey Moore's book "Crossing the Chasm," technology adoption usually follows a clear path. The technology undergoes development and is first discovered by innovators and early adopters, if it can cross the "chasm," it is then adopted by the early majority market, and eventually enters competition with the late majority, and finally laggards. If the technology fails to cross the chasm, then it will ultimately end up in the dustbin of history. However, what sets AI apart is that it does not fit into the pattern of the technology adoption lifecycle.
But AI is not a discrete technology. Instead, it is a collection of technologies, concepts, and methods, all aligned with the pursuit of intelligent machines. This pursuit has inspired scholars and researchers to propose theories about how the brain and intelligence work, as well as their concepts for how to mimic these aspects using technology. Artificial intelligence is the progenitor of technologies that span the technology lifecycle. Investors are not investing in "AI" per se, but rather in AI research and technological advancements that can help achieve AI goals; as researchers discover new insights that help them overcome previous challenges or eventually become the technological infrastructure that catches up with previously infeasible concepts, new technologies emerge and update the investment cycle.
Understanding is Needed
Clearly, intelligence is like an onion with many layers. Once one layer is understood, it is found to only explain a limited aspect of intelligence. We discover another layer that is not fully understood, and then we go back to our research institutions to figure out how it works. In Cognilytica's exploration of the intelligence of voice assistants, the benchmark is set to tease out the next layer: understanding. That is, knowing what something is, recognizing images in a trained concept category, converting audio waveforms into words, identifying patterns in a set of data, or even playing advanced games, is different from actually understanding these things. The lack of understanding is why users get frustrated with voice assistants' responses to their questions and why we often fail to truly achieve autonomous machine capabilities. Without understanding, there is no common sense. Without common sense and understanding, machine learning is just a collection of learned patterns that cannot adapt to the constantly changing real world.
One of the visual concepts that helps understand this increasing value is the "DIKW pyramid":
Although the above Wikipedia entry conveniently skips the "understanding" step in its article, we believe that understanding is the next logical threshold for AI capabilities. Dealing with this layer, like all the previous layers of the AI onion, will require new research breakthroughs, significant improvements in computational power, and data volume. What? Don't we have almost infinite data and infinite computational power?
Pursuing Common Sense: Machine Reasoning
In the early days of artificial intelligence development, researchers realized that for machines to navigate successfully in the real world, they must understand how the world operates and how various things are interconnected. In 1984, the world's longest-running AI project began. The Cyc project focused on generating a comprehensive "ontology" and common sense knowledge about how the world works, basic concepts, and "empirical rules." The Cyc ontology uses a knowledge graph to construct relationships between different concepts and uses an inference engine to enable the system to reason about facts.
The main idea behind Cyc and other efforts to build understanding is the recognition that systems cannot truly be intelligent if they do not understand what they are identifying or classifying. This means that we need to delve deeper into intelligence than just machine learning. We need to peel back another layer of the onion and uncover another delicious layer of frozen cake. We need not only machine learning but also machine reasoning.Machine reasoning is the concept that empowers machines to establish connections between facts, observed outcomes, and all the magical things we can train machines to learn. Machine learning has achieved broad capabilities and functionalities and has opened up a possible world that would not be possible without the ability to train machines to recognize and identify patterns in data. However, these systems are unable to truly apply information to higher-end functionality or apply learning from one domain to another without human intervention, which weakens this capability. Even the application of transfer learning has its limitations.
Indeed, we are rapidly facing the reality that overcoming current limitations with machine learning-centric AI will require more than just machine learning. To reach higher levels, we need to break through this wall and shift from machine learning-centric AI to machine reasoning-centric AI. However, this will require research breakthroughs that we have yet to realize.
The Cyc project, being the longest-running AI project, is an exception to this fact. The reason the Cyc project has persisted for so long is that the pursuit of common sense knowledge has proven to be elusive over the past few decades. Encoding common sense into a machine-processable form is a massive challenge. You not only need to encode the entities themselves in a way that the machine knows what you are talking about, but you also need to encode all the relationships between these entities. Machines need to know millions or even billions of "things." Some of these things are tangible, like "rain," while others are intangible, like "thirst." The task of encoding these relationships is partially automated but still requires human verification of the accuracy of connections. Because, after all, if machines could do this, we would have solved the challenge of machine recognition. It's a bit of a chicken and egg problem. Without some way of organizing the relationships between information, the challenge of machine recognition cannot be solved. But without some form of automation, it is impossible to organize all the relationships that machines need to know.Are we still limited by data and computational power?
It turns out that machine learning is very data and computationally intensive. Over the past decade, many iterative enhancements have alleviated the computational burden and helped improve data efficiency. GPUs, TPUs, and emerging FPGAs are helping provide the necessary raw computational power. However, despite these advancements, complex machine learning models with many dimensions and parameters still require significant computation and data. Machine reasoning is likely to be one or more complexities beyond machine learning. The task of completing the complex relationships between things and truly understanding these things may exceed the computational and data resources available today.
The current wave of interest and investment in artificial intelligence shows no signs of slowing down or stopping anytime soon, but inevitably, for some simple reason, it will slow down at some point: we still do not understand intelligence and how it works. Despite the amazing contributions made by researchers and technologists, we are still guessing in the dark about the mysterious nature of cognition, intelligence, and consciousness. At some point, we will face the limitations of assumptions and approaches, and we will strive to peel back another layer of the onion to address the next set of challenges. Machine reasoning is rapidly approaching, and it is the next challenge we must overcome in our quest for artificial intelligence. If we can apply our research and investment talent to address this problem, we can keep the momentum going through AI research and investment. Otherwise, the cycle of AI will repeat, and the current wave will come to an end. Perhaps not now, or even in the next few years, but the rise and fall of AI is inevitable.
Google TV integrates Gemini AI for image/video generation, enhances Google Photos, and adds a dedicated YouTube Shorts feed to the homepage.
Read moreAlphabet Q1 paid subscribers reached 350M, fueled by YouTube & Google One. YouTube ad revenue slightly missed at $9.88B amid Premium growth. Total Q1 revenue hit $109.9B, beating forecasts.
Read moreAnthropic releases Claude connectors for Adobe, Blender, and Ableton, enabling AI workflow automation and app control. It also sponsors the Blender Development Foundation.
Read moreApple introduces an App Store model splitting 12-month plans into monthly installments. On iOS 26.4+, it lowers upfront costs without reducing total price, adding billing tracking.
Read moreGoogle launches Aadhaar digital IDs in Wallet for India, enabling secure QR/NFC verification. A phased partner rollout targets key sectors, while passport-based credentials expand globally.
Read moreYouTube tests "Ask YouTube," an AI tool blending text and video for complex queries. Initially for US Premium users, it supports interactive follow-ups, expanding Google's AI search.
Read moreGoogle grants US DoD classified AI access, joining OpenAI and xAI in defense deals amid Anthropic's dispute. Contractual safeguards against surveillance and lethal weapons remain unenforceable.
Read moreMistral AI launches Workflows, a Temporal-powered orchestration platform for secure, Python-driven enterprise AI automation featuring human-in-the-loop approvals.
Read moreCustom subliminal audio maker

AI Social Media Manager

AI Search Visibility

Workout and Calorie Tracker
Stock research platform

AI interview assistant

Scientific figure generator

AI courses with real code