Tag Archives: chip

#437345 Moore’s Law Lives: Intel Says Chips ...

If you weren’t already convinced the digital world is taking over, you probably are now.

To keep the economy on life support as people stay home to stem the viral tide, we’ve been forced to digitize interactions at scale (for better and worse). Work, school, events, shopping, food, politics. The companies at the center of the digital universe are now powerhouses of the modern era—worth trillions and nearly impossible to avoid in daily life.

Six decades ago, this world didn’t exist.

A humble microchip in the early 1960s would have boasted a handful of transistors. Now, your laptop or smartphone runs on a chip with billions of transistors. As first described by Moore’s Law, this is possible because the number of transistors on a chip doubled with extreme predictability every two years for decades.

But now progress is faltering as the size of transistors approaches physical limits, and the money and time it takes to squeeze a few more onto a chip are growing. There’ve been many predictions that Moore’s Law is, finally, ending. But, perhaps also predictably, the company whose founder coined Moore’s Law begs to differ.

In a keynote presentation at this year’s Hot Chips conference, Intel’s chief architect, Raja Koduri, laid out a roadmap to increase transistor density—that is, the number of transistors you can fit on a chip—by a factor of 50.

“We firmly believe there is a lot more transistor density to come,” Koduri said. “The vision will play out over time—maybe a decade or more—but it will play out.”

Why the optimism?

Calling the end of Moore’s Law is a bit of a tradition. As Peter Lee, vice president at Microsoft Research, quipped to The Economist a few years ago, “The number of people predicting the death of Moore’s Law doubles every two years.” To date, prophets of doom have been premature, and though the pace is slowing, the industry continues to dodge death with creative engineering.

Koduri believes the trend will continue this decade and outlined the upcoming chip innovations Intel thinks can drive more gains in computing power.

Keeping It Traditional
First, engineers can further shrink today’s transistors. Fin field effect transistors (or FinFET) first hit the scene in the 2010s and have since pushed chip features past 14 and 10 nanometers (or nodes, as such size checkpoints are called). Korduri said FinFET will again triple chip density before it’s exhausted.

The Next Generation
FinFET will hand the torch off to nanowire transistors (also known as gate-all-around transistors).

Here’s how they’ll work. A transistor is made up of three basic components: the source, where current is introduced, the gate and channel, where current selectively flows, and the drain. The gate is like a light switch. It controls how much current flows through the channel. A transistor is “on” when the gate allows current to flow, and it’s off when no current flows. The smaller transistors get, the harder it is to control that current.

FinFET maintained fine control of current by surrounding the channel with a gate on three sides. Nanowire designs kick that up a notch by surrounding the channel with a gate on four sides (hence, gate-all-around). They’ve been in the works for years and are expected around 2025. Koduri said first-generation nanowire transistors will be followed by stacked nanowire transistors, and together, they’ll quadruple transistor density.

Building Up
Growing transistor density won’t only be about shrinking transistors, but also going 3D.

This is akin to how skyscrapers increase a city’s population density by adding more usable space on the same patch of land. Along those lines, Intel recently launched its Foveros chip design. Instead of laying a chip’s various “neighborhoods” next to each other in a 2D silicon sprawl, they’ve stacked them on top of each other like a layer cake. Chip stacking isn’t entirely new, but it’s advancing and being applied to general purpose CPUs, like the chips in your phone and laptop.

Koduri said 3D chip stacking will quadruple transistor density.

A Self-Fulfilling Prophecy
The technologies Koduri outlines are an evolution of the same general technology in use today. That is, we don’t need quantum computing or nanotube transistors to augment or replace silicon chips yet. Rather, as it’s done many times over the years, the chip industry will get creative with the design of its core product to realize gains for another decade.

Last year, veteran chip engineer Jim Keller, who at the time was Intel’s head of silicon engineering but has since left the company, told MIT Technology Review there are over a 100 variables driving Moore’s Law (including 3D architectures and new transistor designs). From the standpoint of pure performance, it’s also about how efficiently software uses all those transistors. Keller suggested that with some clever software tweaks “we could get chips that are a hundred times faster in 10 years.”

But whether Intel’s vision pans out as planned is far from certain.

Intel’s faced challenges recently, taking five years instead of two to move its chips from 14 nanometers to 10 nanometers. After a delay of six months for its 7-nanometer chips, it’s now a year behind schedule and lagging other makers who already offer 7-nanometer chips. This is a key point. Yes, chipmakers continue making progress, but it’s getting harder, more expensive, and timelines are stretching.

The question isn’t if Intel and competitors can cram more transistors onto a chip—which, Intel rival TSMC agrees is clearly possible—it’s how long will it take and at what cost?

That said, demand for more computing power isn’t going anywhere.

Amazon, Microsoft, Alphabet, Apple, and Facebook now make up a whopping 20 percent of the stock market’s total value. By that metric, tech is the most dominant industry in at least 70 years. And new technologies—from artificial intelligence and virtual reality to a proliferation of Internet of Things devices and self-driving cars—will demand better chips.

There’s ample motivation to push computing to its bitter limits and beyond. As is often said, Moore’s Law is a self-fulfilling prophecy, and likely whatever comes after it will be too.

Image credit: Laura Ockel / Unsplash Continue reading

Posted in Human Robots

#437337 6G Will Be 100 Times Faster Than ...

Though 5G—a next-generation speed upgrade to wireless networks—is scarcely up and running (and still nonexistent in many places) researchers are already working on what comes next. It lacks an official name, but they’re calling it 6G for the sake of simplicity (and hey, it’s tradition). 6G promises to be up to 100 times faster than 5G—fast enough to download 142 hours of Netflix in a second—but researchers are still trying to figure out exactly how to make such ultra-speedy connections happen.

A new chip, described in a paper in Nature Photonics by a team from Osaka University and Nanyang Technological University in Singapore, may give us a glimpse of our 6G future. The team was able to transmit data at a rate of 11 gigabits per second, topping 5G’s theoretical maximum speed of 10 gigabits per second and fast enough to stream 4K high-def video in real time. They believe the technology has room to grow, and with more development, might hit those blistering 6G speeds.

NTU final year PhD student Abhishek Kumar, Assoc Prof Ranjan Singh and postdoc Dr Yihao Yang. Dr Singh is holding the photonic topological insulator chip made from silicon, which can transmit terahertz waves at ultrahigh speeds. Credit: NTU Singapore
But first, some details about 5G and its predecessors so we can differentiate them from 6G.

Electromagnetic waves are characterized by a wavelength and a frequency; the wavelength is the distance a cycle of the wave covers (peak to peak or trough to trough, for example), and the frequency is the number of waves that pass a given point in one second. Cellphones use miniature radios to pick up electromagnetic signals and convert those signals into the sights and sounds on your phone.

4G wireless networks run on millimeter waves on the low- and mid-band spectrum, defined as a frequency of a little less (low-band) and a little more (mid-band) than one gigahertz (or one billion cycles per second). 5G kicked that up several notches by adding even higher frequency millimeter waves of up to 300 gigahertz, or 300 billion cycles per second. Data transmitted at those higher frequencies tends to be information-dense—like video—because they’re much faster.

The 6G chip kicks 5G up several more notches. It can transmit waves at more than three times the frequency of 5G: one terahertz, or a trillion cycles per second. The team says this yields a data rate of 11 gigabits per second. While that’s faster than the fastest 5G will get, it’s only the beginning for 6G. One wireless communications expert even estimates 6G networks could handle rates up to 8,000 gigabits per second; they’ll also have much lower latency and higher bandwidth than 5G.

Terahertz waves fall between infrared waves and microwaves on the electromagnetic spectrum. Generating and transmitting them is difficult and expensive, requiring special lasers, and even then the frequency range is limited. The team used a new material to transmit terahertz waves, called photonic topological insulators (PTIs). PTIs can conduct light waves on their surface and edges rather than having them run through the material, and allow light to be redirected around corners without disturbing its flow.

The chip is made completely of silicon and has rows of triangular holes. The team’s research showed the chip was able to transmit terahertz waves error-free.

Nanyang Technological University associate professor Ranjan Singh, who led the project, said, “Terahertz technology […] can potentially boost intra-chip and inter-chip communication to support artificial intelligence and cloud-based technologies, such as interconnected self-driving cars, which will need to transmit data quickly to other nearby cars and infrastructure to navigate better and also to avoid accidents.”

Besides being used for AI and self-driving cars (and, of course, downloading hundreds of hours of video in seconds), 6G would also make a big difference for data centers, IoT devices, and long-range communications, among other applications.

Given that 5G networks are still in the process of being set up, though, 6G won’t be coming on the scene anytime soon; a recent whitepaper on 6G from Japanese company NTTDoCoMo estimates we’ll see it in 2030, pointing out that wireless connection tech generations have thus far been spaced about 10 years apart; we got 3G in the early 2000s, 4G in 2010, and 5G in 2020.

In the meantime, as 6G continues to develop, we’re still looking forward to the widespread adoption of 5G.

Image Credit: Hans Braxmeier from Pixabay Continue reading

Posted in Human Robots

#437269 DeepMind’s Newest AI Programs Itself ...

When Deep Blue defeated world chess champion Garry Kasparov in 1997, it may have seemed artificial intelligence had finally arrived. A computer had just taken down one of the top chess players of all time. But it wasn’t to be.

Though Deep Blue was meticulously programmed top-to-bottom to play chess, the approach was too labor-intensive, too dependent on clear rules and bounded possibilities to succeed at more complex games, let alone in the real world. The next revolution would take a decade and a half, when vastly more computing power and data revived machine learning, an old idea in artificial intelligence just waiting for the world to catch up.

Today, machine learning dominates, mostly by way of a family of algorithms called deep learning, while symbolic AI, the dominant approach in Deep Blue’s day, has faded into the background.

Key to deep learning’s success is the fact the algorithms basically write themselves. Given some high-level programming and a dataset, they learn from experience. No engineer anticipates every possibility in code. The algorithms just figure it.

Now, Alphabet’s DeepMind is taking this automation further by developing deep learning algorithms that can handle programming tasks which have been, to date, the sole domain of the world’s top computer scientists (and take them years to write).

In a paper recently published on the pre-print server arXiv, a database for research papers that haven’t been peer reviewed yet, the DeepMind team described a new deep reinforcement learning algorithm that was able to discover its own value function—a critical programming rule in deep reinforcement learning—from scratch.

Surprisingly, the algorithm was also effective beyond the simple environments it trained in, going on to play Atari games—a different, more complicated task—at a level that was, at times, competitive with human-designed algorithms and achieving superhuman levels of play in 14 games.

DeepMind says the approach could accelerate the development of reinforcement learning algorithms and even lead to a shift in focus, where instead of spending years writing the algorithms themselves, researchers work to perfect the environments in which they train.

Pavlov’s Digital Dog
First, a little background.

Three main deep learning approaches are supervised, unsupervised, and reinforcement learning.

The first two consume huge amounts of data (like images or articles), look for patterns in the data, and use those patterns to inform actions (like identifying an image of a cat). To us, this is a pretty alien way to learn about the world. Not only would it be mind-numbingly dull to review millions of cat images, it’d take us years or more to do what these programs do in hours or days. And of course, we can learn what a cat looks like from just a few examples. So why bother?

While supervised and unsupervised deep learning emphasize the machine in machine learning, reinforcement learning is a bit more biological. It actually is the way we learn. Confronted with several possible actions, we predict which will be most rewarding based on experience—weighing the pleasure of eating a chocolate chip cookie against avoiding a cavity and trip to the dentist.

In deep reinforcement learning, algorithms go through a similar process as they take action. In the Atari game Breakout, for instance, a player guides a paddle to bounce a ball at a ceiling of bricks, trying to break as many as possible. When playing Breakout, should an algorithm move the paddle left or right? To decide, it runs a projection—this is the value function—of which direction will maximize the total points, or rewards, it can earn.

Move by move, game by game, an algorithm combines experience and value function to learn which actions bring greater rewards and improves its play, until eventually, it becomes an uncanny Breakout player.

Learning to Learn (Very Meta)
So, a key to deep reinforcement learning is developing a good value function. And that’s difficult. According to the DeepMind team, it takes years of manual research to write the rules guiding algorithmic actions—which is why automating the process is so alluring. Their new Learned Policy Gradient (LPG) algorithm makes solid progress in that direction.

LPG trained in a number of toy environments. Most of these were “gridworlds”—literally two-dimensional grids with objects in some squares. The AI moves square to square and earns points or punishments as it encounters objects. The grids vary in size, and the distribution of objects is either set or random. The training environments offer opportunities to learn fundamental lessons for reinforcement learning algorithms.

Only in LPG’s case, it had no value function to guide that learning.

Instead, LPG has what DeepMind calls a “meta-learner.” You might think of this as an algorithm within an algorithm that, by interacting with its environment, discovers both “what to predict,” thereby forming its version of a value function, and “how to learn from it,” applying its newly discovered value function to each decision it makes in the future.

Prior work in the area has had some success, but according to DeepMind, LPG is the first algorithm to discover reinforcement learning rules from scratch and to generalize beyond training. The latter was particularly surprising because Atari games are so different from the simple worlds LPG trained in—that is, it had never seen anything like an Atari game.

Time to Hand Over the Reins? Not Just Yet
LPG is still behind advanced human-designed algorithms, the researchers said. But it outperformed a human-designed benchmark in training and even some Atari games, which suggests it isn’t strictly worse, just that it specializes in some environments.

This is where there’s room for improvement and more research.

The more environments LPG saw, the more it could successfully generalize. Intriguingly, the researchers speculate that with enough well-designed training environments, the approach might yield a general-purpose reinforcement learning algorithm.

At the least, though, they say further automation of algorithm discovery—that is, algorithms learning to learn—will accelerate the field. In the near term, it can help researchers more quickly develop hand-designed algorithms. Further out, as self-discovered algorithms like LPG improve, engineers may shift from manually developing the algorithms themselves to building the environments where they learn.

Deep learning long ago left Deep Blue in the dust at games. Perhaps algorithms learning to learn will be a winning strategy in the real world too.

Image credit: Mike Szczepanski / Unsplash Continue reading

Posted in Human Robots

#437182 MIT’s Tiny New Brain Chip Aims for AI ...

The human brain operates on roughly 20 watts of power (a third of a 60-watt light bulb) in a space the size of, well, a human head. The biggest machine learning algorithms use closer to a nuclear power plant’s worth of electricity and racks of chips to learn.

That’s not to slander machine learning, but nature may have a tip or two to improve the situation. Luckily, there’s a branch of computer chip design heeding that call. By mimicking the brain, super-efficient neuromorphic chips aim to take AI off the cloud and put it in your pocket.

The latest such chip is smaller than a piece of confetti and has tens of thousands of artificial synapses made out of memristors—chip components that can mimic their natural counterparts in the brain.

In a recent paper in Nature Nanotechnology, a team of MIT scientists say their tiny new neuromorphic chip was used to store, retrieve, and manipulate images of Captain America’s Shield and MIT’s Killian Court. Whereas images stored with existing methods tended to lose fidelity over time, the new chip’s images remained crystal clear.

“So far, artificial synapse networks exist as software. We’re trying to build real neural network hardware for portable artificial intelligence systems,” Jeehwan Kim, associate professor of mechanical engineering at MIT said in a press release. “Imagine connecting a neuromorphic device to a camera on your car, and having it recognize lights and objects and make a decision immediately, without having to connect to the internet. We hope to use energy-efficient memristors to do those tasks on-site, in real-time.”

A Brain in Your Pocket
Whereas the computers in our phones and laptops use separate digital components for processing and memory—and therefore need to shuttle information between the two—the MIT chip uses analog components called memristors that process and store information in the same place. This is similar to the way the brain works and makes memristors far more efficient. To date, however, they’ve struggled with reliability and scalability.

To overcome these challenges, the MIT team designed a new kind of silicon-based, alloyed memristor. Ions flowing in memristors made from unalloyed materials tend to scatter as the components get smaller, meaning the signal loses fidelity and the resulting computations are less reliable. The team found an alloy of silver and copper helped stabilize the flow of silver ions between electrodes, allowing them to scale the number of memristors on the chip without sacrificing functionality.

While MIT’s new chip is promising, there’s likely a ways to go before memristor-based neuromorphic chips go mainstream. Between now and then, engineers like Kim have their work cut out for them to further scale and demonstrate their designs. But if successful, they could make for smarter smartphones and other even smaller devices.

“We would like to develop this technology further to have larger-scale arrays to do image recognition tasks,” Kim said. “And some day, you might be able to carry around artificial brains to do these kinds of tasks, without connecting to supercomputers, the internet, or the cloud.”

Special Chips for AI
The MIT work is part of a larger trend in computing and machine learning. As progress in classical chips has flagged in recent years, there’s been an increasing focus on more efficient software and specialized chips to continue pushing the pace.

Neuromorphic chips, for example, aren’t new. IBM and Intel are developing their own designs. So far, their chips have been based on groups of standard computing components, such as transistors (as opposed to memristors), arranged to imitate neurons in the brain. These chips are, however, still in the research phase.

Graphics processing units (GPUs)—chips originally developed for graphics-heavy work like video games—are the best practical example of specialized hardware for AI and were heavily used in this generation of machine learning early on. In the years since, Google, NVIDIA, and others have developed even more specialized chips that cater more specifically to machine learning.

The gains from such specialized chips are already being felt.

In a recent cost analysis of machine learning, research and investment firm ARK Invest said cost declines have far outpaced Moore’s Law. In a particular example, they found the cost to train an image recognition algorithm (ResNet-50) went from around $1,000 in 2017 to roughly $10 in 2019. The fall in cost to actually run such an algorithm was even more dramatic. It took $10,000 to classify a billion images in 2017 and just $0.03 in 2019.

Some of these declines can be traced to better software, but according to ARK, specialized chips have improved performance by nearly 16 times in the last three years.

As neuromorphic chips—and other tailored designs—advance further in the years to come, these trends in cost and performance may continue. Eventually, if all goes to plan, we might all carry a pocket brain that can do the work of today’s best AI.

Image credit: Peng Lin Continue reading

Posted in Human Robots

#436774 AI Is an Energy-Guzzler. We Need to ...

There is a saying that has emerged among the tech set in recent years: AI is the new electricity. The platitude refers to the disruptive power of artificial intelligence for driving advances in everything from transportation to predicting the weather.

Of course, the computers and data centers that support AI’s complex algorithms are very much dependent on electricity. While that may seem pretty obvious, it may be surprising to learn that AI can be extremely power-hungry, especially when it comes to training the models that enable machines to recognize your face in a photo or for Alexa to understand a voice command.

The scale of the problem is difficult to measure, but there have been some attempts to put hard numbers on the environmental cost.

For instance, one paper published on the open-access repository arXiv claimed that the carbon emissions for training a basic natural language processing (NLP) model—algorithms that process and understand language-based data—are equal to the CO2 produced by the average American lifestyle over two years. A more robust model required the equivalent of about 17 years’ worth of emissions.

The authors noted that about a decade ago, NLP models could do the job on a regular commercial laptop. Today, much more sophisticated AI models use specialized hardware like graphics processing units, or GPUs, a chip technology popularized by Nvidia for gaming that also proved capable of supporting computing tasks for AI.

OpenAI, a nonprofit research organization co-founded by tech prophet and profiteer Elon Musk, said that the computing power “used in the largest AI training runs has been increasing exponentially with a 3.4-month doubling time” since 2012. That’s about the time that GPUs started making their way into AI computing systems.

Getting Smarter About AI Chip Design
While GPUs from Nvidia remain the gold standard in AI hardware today, a number of startups have emerged to challenge the company’s industry dominance. Many are building chipsets designed to work more like the human brain, an area that’s been dubbed neuromorphic computing.

One of the leading companies in this arena is Graphcore, a UK startup that has raised more than $450 million and boasts a valuation of $1.95 billion. The company’s version of the GPU is an IPU, which stands for intelligence processing unit.

To build a computer brain more akin to a human one, the big brains at Graphcore are bypassing the precise but time-consuming number-crunching typical of a conventional microprocessor with one that’s content to get by on less precise arithmetic.

The results are essentially the same, but IPUs get the job done much quicker. Graphcore claimed it was able to train the popular BERT NLP model in just 56 hours, while tripling throughput and reducing latency by 20 percent.

An article in Bloomberg compared the approach to the “human brain shifting from calculating the exact GPS coordinates of a restaurant to just remembering its name and neighborhood.”

Graphcore’s hardware architecture also features more built-in memory processing, boosting efficiency because there’s less need to send as much data back and forth between chips. That’s similar to an approach adopted by a team of researchers in Italy that recently published a paper about a new computing circuit.

The novel circuit uses a device called a memristor that can execute a mathematical function known as a regression in just one operation. The approach attempts to mimic the human brain by processing data directly within the memory.

Daniele Ielmini at Politecnico di Milano, co-author of the Science Advances paper, told Singularity Hub that the main advantage of in-memory computing is the lack of any data movement, which is the main bottleneck of conventional digital computers, as well as the parallel processing of data that enables the intimate interactions among various currents and voltages within the memory array.

Ielmini explained that in-memory computing can have a “tremendous impact on energy efficiency of AI, as it can accelerate very advanced tasks by physical computation within the memory circuit.” He added that such “radical ideas” in hardware design will be needed in order to make a quantum leap in energy efficiency and time.

It’s Not Just a Hardware Problem
The emphasis on designing more efficient chip architecture might suggest that AI’s power hunger is essentially a hardware problem. That’s not the case, Ielmini noted.

“We believe that significant progress could be made by similar breakthroughs at the algorithm and dataset levels,” he said.

He’s not the only one.

One of the key research areas at Qualcomm’s AI research lab is energy efficiency. Max Welling, vice president of Qualcomm Technology R&D division, has written about the need for more power-efficient algorithms. He has gone so far as to suggest that AI algorithms will be measured by the amount of intelligence they provide per joule.

One emerging area being studied, Welling wrote, is the use of Bayesian deep learning for deep neural networks.

It’s all pretty heady stuff and easily the subject of a PhD thesis. The main thing to understand in this context is that Bayesian deep learning is another attempt to mimic how the brain processes information by introducing random values into the neural network. A benefit of Bayesian deep learning is that it compresses and quantifies data in order to reduce the complexity of a neural network. In turn, that reduces the number of “steps” required to recognize a dog as a dog—and the energy required to get the right result.

A team at Oak Ridge National Laboratory has previously demonstrated another way to improve AI energy efficiency by converting deep learning neural networks into what’s called a spiking neural network. The researchers spiked their deep spiking neural network (DSNN) by introducing a stochastic process that adds random values like Bayesian deep learning.

The DSNN actually imitates the way neurons interact with synapses, which send signals between brain cells. Individual “spikes” in the network indicate where to perform computations, lowering energy consumption because it disregards unnecessary computations.

The system is being used by cancer researchers to scan millions of clinical reports to unearth insights on causes and treatments of the disease.

Helping battle cancer is only one of many rewards we may reap from artificial intelligence in the future, as long as the benefits of those algorithms outweigh the costs of using them.

“Making AI more energy-efficient is an overarching objective that spans the fields of algorithms, systems, architecture, circuits, and devices,” Ielmini said.

Image Credit: analogicus from Pixabay Continue reading

Posted in Human Robots