2020 will be the year of augmented reality
Greg Nichols looks back and takes a look at the biggest trends in AR and spatial computing. Read more: https://zd.net/2EfpRyz
We will continue where we left off with part one of the top technology trends for the 2020s, this is what the data landscape will form for the coming years.
2. AI: It’s all about data and hardware
The last part of the years 2010 was all about AI and the 2020s will be no different. We will see that AI extends its reach and influences every imaginable field. However, we have already seen the AI hype rise, but we must also be prepared for a kickback. And it is very important to know what “AI” actually means.
In essence, what we call AI today is an umbrella term for different techniques for matching patterns. Machine learning and its various sub-domains, such as deep learning, essentially come down to pattern recognition. We have seen several breakthroughs in the 2010s, but the seeds for most techniques and algorithms were planted decades ago and remain essentially the same.
Yet we have seen the performance of AI systems in many domains range from worse than human to overtaking and surpassing people. How is that possible? The answer is two-fold: data and calculation.
The digitization of almost all aspects of human activity has led to an explosion of the amount of data being generated. Algorithms now have much more data to work with, and that alone means they can perform much better. At the same time, however, progress has been made in areas such as image recognition: adjustments in neural networks brought about by vibrant communities have improved the accuracy of the algorithms. ImageNet is a good example of this.
Much of what is sold today as “AI” is hose oil. It doesn’t work and can’t work. Arvind Narayanan, a Princeton Professor, recently made waves by calling this out.
AI continues to break into new domains at breakneck speed. In 2019 alone we have made great progress in the field of natural language processing, games and common sense, to name just a few. New performance in terms of result quality and execution speed have been achieved almost monthly. The amount of resources allocated is huge and the research is progressing faster than ever. So should we all prepare for the brave new AI world? Well, maybe not that fast.
The problem with the AI frenzy is the gap between the haves and the have nots is widening. And not only because of the resources and expertise of the major players. It is a self-reinforcing loop of sorts: being data-driven, designing and producing data-driven products means that these products can not only have a head start, but that they also bring in more data when they work.
Because there is an evolutionary link that connects data and AI, more data is used to develop a better AI, leading to better products, more data, and so on. Facebook is an archetypal and widely recognized example of this, but it is not the only one. When people like the economist ask for a new approach to antitrust rules for the data economy, this should be cause for concern.
However, data is only part of the AI comparison. The other part is hardware. Without the huge advances in hardware that the 2010s have seen, AI would not be possible. In the past, access to the computing power needed to handle the huge amounts of data needed for machine learning was a reserved privilege.
Although the type of hardware that Big Tech has access to remains unintelligible to most people, the democratization of species seems to have been transcribed. The combination of cloud, with on-demand access to processing power and specialized hardware for AI workloads, has made AI chips accessible to more organizations than ever, assuming they can afford it.
The “Wafer-Scale Engine” from Cerebras covers almost the entire surface of a 12-inch silicon wafer and is therefore 57 times larger than the largest graphic processing unit of NVIDIA.
NVIDIA was the major innovator and winner in 2010 AI hardware. The company that most people got to know as a creator of GPUs, specialized hardware that is commonly used by gamers for fast graphics, has reinvented itself as an AI superpower. The architecture of GPUs appears to be very suitable for performing AI workloads.
Intel became complacent in its dominance of traditional CPU hardware and other GPU makers could not perform, so NVIDIA became the leader in AI hardware. However, that is not certain, and the hardware space is already experiencing rapid innovation.
While NVIDIA dominates AI hardware and has also built a software ecosystem around it, waves of disruption affect the AI chip market. Just a few days before the end of the 2010, Intel fell back by acquiring Habana Labs. Habana Labs is one of the many startups on the AI chip market that wants to come up with new designs, built from the ground up to accommodate AI workloads.
Although Habana Labs is unknown to many people, its chips are already being used in production by cloud suppliers and autonomous vehicle manufacturers. GraphCore, the first AI chip unicorn at the end of 2018, recently announced that its chips are now being used in Microsoft Azure Cloud. The AI chip race has not started yet.
1. The future is graph, knowledge graph
Until the beginning of the 2010s, the world mainly relied on relational databases and spreadsheets. To a large extent still. But if the 2010 years brought the first traces of disagreement in the monoculture of data structures in tabular form, the 2020s will bring the last nail in the coffin. The NoSQL database of databases has largely succeeded in getting developers, administrators, CIOs, CTOs, and business people out of their comfort zone and imparted the mentality of “best tool for the job”.
Polyglot persistence, like the language for using data models and data management, interchangeable depending on the task being performed, becomes the new standard. After relational, key / value, document, column and time series databases, the newest link in this evolutionary proliferation of data structures is a graph. Graph databases and knowledge graphs have expanded enormously in recent years and have been included in hype cycles.
While it is understandable why many people tend to view graphics as a new technology, the truth is that this technology is at least 20 years old. And it was largely initiated by none other than Tim Berners Lee, who is also credited as the inventor of the web, in 2001 with the publication of his Semantic Web manifest in Scientific American. Lee also coined the term Giant Global Graph to describe the next phase in the evolution of the web.
Since we have been working with this technology since the early 2000s, it is exciting to see it getting steam with technical progress, financing and use cases that are piling up into a snowball effect. It is also funny to see the washing of the charts begin. In essence, progress is made in the graph along the path of progress in machine learning.
It is not so much that there was a major breakthrough in the technology that made it possible, but more about the right circumstances that made it boom. Many of the concepts, formats, standards and technology with which graphic databases and knowledge graphs can flourish have been developed for more than 20 years. What has caused the perfect chart storm is a combination of factors.
Google, NASA and leading organizations from every domain use knowledge charts to manage and utilize large amounts of data
Just like AI, the data explosion contributed to bringing the graph forward. Now that Big is no longer a qualification for Data, because we master the art of storing much of it, the question is really how we can get value from data. Using connections in data is a prominent way to extract value from data, and graphing is the best way to use connections.
This is why graphic databases excel in usage situations where connections must be found in data, such as fraud prevention or master data management. This is why graph analyzes, with algorithms such as centrality or PageRank based on the processing of nodes and edges, can provide valuable insights into connected data sets. Because terminology is still moving for many newcomers in this area, a brief history lesson and grounding in semantics may be needed.
Graph analyzes such as PageRank can be applied to data stored in each back end. Graph databases are designed back-end for graphical data structures and offer specialized query languages, APIs and often storage structures. Knowledge graphs, on the other hand, are a specific subclass of graphs, also called semantic graphs, which are supplied with metadata, schemes and global identification options.
Google has played an important role in the rise of graphs and knowledge graphs. Since the web itself is a great use scenario for graphing, PageRank was born. Because crawling and categorizing content on the web is a very difficult problem to solve without semantics and metadata, Google embraced them and coined the term Knowledge Chart in 2012. This and the widespread acceptance of schema.org that accompanied it was the start of the rapid rise of graph technology and knowledge graphs.
Knowledge charts can address important challenges such as data management, but ultimately they can serve as the digital substrate to reconcile the philosophy of knowledge acquisition and organization with the practice of data management in the digital age. The world’s NASAs and Morgan Stanleys manage ontologies and use knowledge graphs.
Charts and knowledge charts are also cross-cut in AI. Much of the AI hardware and software for the 2020s uses graphic data structures. A combination of bottom-up, pattern matching techniques with top-down, knowledge-based approaches is the most promising way for AI to continue making progress.
As Nathan Benaich, author of the State of AI Report, said: “Domain knowledge can effectively help a deeply learning system start its knowledge by coding primitives instead of forcing the model to learn it all over again.” Knowledge graphs are the best technology we have for coding domain knowledge, and the world’s most extensive knowledge base – the web – already functions as such.
Knowledge chart is a technology with which other technologies can accelerate their growth and with which people can also inventory their own knowledge. That is why the future is Knowledge Graph.
To infinity and beyond
Looking back, it becomes clear how far we have come in the relatively short period of the past decade. Counterintuitive, although this may seem, we are not sure if this is a good thing. Somewhere along the way, technological advances allowed people to follow, understand, and digest technology in the dust. At the start of this new decade, we seem to delve into the never-ending race for more: more data, more processing power, more technology.
The belief that more equals better seems to be firmly rooted in most of us. And the signs of what’s coming seem to tell the story of not just more, but immeasurably much more. Quantum computing is leaping forward and promises to unlock computing power above our wildest imagination. DNA storage seems to do the same for storage. More data and data than we would know what to do with. To infinity and beyond. But what for and for whom?
Does this technology make us happier and brings us closer or alienates us? Where do all the huge productivity gains go? Who is in control, who can call the shots and why?
For example, AI is already used to make critical decisions. Who can build those systems, based on what data and according to whose ethics and criteria? Should society as a whole have some control over it? How can society even dream of mastering a technology that it hardly understands, and in what way? Are we sure that more technology is the solution to technological problems? What is a moral compass for the 21st century?
For the time being these are questions that few people want to answer. But if the 2020s develop according to the path they are set to, more and more of us will have to face these questions head-on.