Connect with us

AI Research

How AlphaChip transformed computer chip design

Published

on


Science

Published
Authors

Anna Goldie and Azalia Mirhoseini

Our AI method has accelerated and optimized chip design, and its superhuman chip layouts are used in hardware around the world

In 2020, we released a preprint introducing our novel reinforcement learning method for designing chip layouts, which we later published in Nature and open sourced.

Today, we’re publishing a Nature addendum that describes more about our method and its impact on the field of chip design. We’re also releasing a pre-trained checkpoint, sharing the model weights and announcing its name: AlphaChip.

Computer chips have fueled remarkable progress in artificial intelligence (AI), and AlphaChip returns the favor by using AI to accelerate and optimize chip design. The method has been used to design superhuman chip layouts in the last three generations of Google’s custom AI accelerator, the Tensor Processing Unit (TPU).

AlphaChip was one of the first reinforcement learning approaches used to solve a real-world engineering problem. It generates superhuman or comparable chip layouts in hours, rather than taking weeks or months of human effort, and its layouts are used in chips all over the world, from data centers to mobile phones.

AlphaChip’s groundbreaking AI approach revolutionizes a key phase of chip design.

SR Tsai, Senior Vice President of MediaTek

How AlphaChip works

Designing a chip layout is not a simple task. Computer chips consist of many interconnected blocks, with layers of circuit components, all connected by incredibly thin wires. There are also lots of complex and intertwined design constraints that all have to be met at the same time. Because of its sheer complexity, chip designers have struggled to automate the chip floorplanning process for over sixty years.

Similar to AlphaGo and AlphaZero, which learned to master the games of Go, chess and shogi, we built AlphaChip to approach chip floorplanning as a kind of game.

Starting from a blank grid, AlphaChip places one circuit component at a time until it’s done placing all the components. Then it’s rewarded based on the quality of the final layout. A novel “edge-based” graph neural network allows AlphaChip to learn the relationships between interconnected chip components and to generalize across chips, letting AlphaChip improve with each layout it designs.

Left: Animation showing AlphaChip placing the open-source, Ariane RISC-V CPU, with no prior experience. Right: Animation showing AlphaChip placing the same block after having practiced on 20 TPU-related designs.

Using AI to design Google’s AI accelerator chips

AlphaChip has generated superhuman chip layouts used in every generation of Google’s TPU since its publication in 2020. These chips make it possible to massively scale-up AI models based on Google’s Transformer architecture.

TPUs lie at the heart of our powerful generative AI systems, from large language models, like Gemini, to image and video generators, Imagen and Veo. These AI accelerators also lie at the heart of Google’s AI services and are available to external users via Google Cloud.

A row of Cloud TPU v5p AI accelerator supercomputers in a Google data center.

To design TPU layouts, AlphaChip first practices on a diverse range of chip blocks from previous generations, such as on-chip and inter-chip network blocks, memory controllers, and data transport buffers. This process is called pre-training. Then we run AlphaChip on current TPU blocks to generate high-quality layouts. Unlike prior approaches, AlphaChip becomes better and faster as it solves more instances of the chip placement task, similar to how human experts do.

With each new generation of TPU, including our latest Trillium (6th generation), AlphaChip has designed better chip layouts and provided more of the overall floorplan, accelerating the design cycle and yielding higher-performance chips.

Bar graph showing the number of AlphaChip designed chip blocks across three generations of Google’s Tensor Processing Units (TPU), including v5e, v5p and Trillium.

Bar graph showing AlphaChip’s average wirelength reduction across three generations of Google’s Tensor Processing Units (TPUs), compared to placements generated by the TPU physical design team.

AlphaChip’s broader impact

AlphaChip’s impact can be seen through its applications across Alphabet, the research community and the chip design industry. Beyond designing specialized AI accelerators like TPUs, AlphaChip has generated layouts for other chips across Alphabet, such as Google Axion Processors, our first Arm-based general-purpose data center CPUs.

External organizations are also adopting and building on AlphaChip. For example, MediaTek, one of the top chip design companies in the world, extended AlphaChip to accelerate development of their most advanced chips while improving power, performance and chip area.

AlphaChip has triggered an explosion of work on AI for chip design, and has been extended to other critical stages of chip design, such as logic synthesis and macro selection.

AlphaChip has inspired an entirely new line of research on reinforcement learning for chip design, cutting across the design flow from logic synthesis to floorplanning, timing optimization and beyond.

Professor Siddharth Garg, NYU Tandon School of Engineering

Creating the chips of the future

We believe AlphaChip has the potential to optimize every stage of the chip design cycle, from computer architecture to manufacturing — and to transform chip design for custom hardware found in everyday devices such as smartphones, medical equipment, agricultural sensors and more.

Future versions of AlphaChip are now in development and we look forward to working with the community to continue revolutionizing this area and bring about a future in which chips are even faster, cheaper and more power-efficient.

Acknowledgements

We’re so grateful to our amazing coauthors: Mustafa Yazgan, Joe Wenjie Jiang, Ebrahim Songhori, Shen Wang, Young-Joon Lee, Eric Johnson, Omkar Pathak, Azade Nazi, Jiwoo Pak, Andy Tong, Kavya Srinivasa, William Hang, Emre Tuncer, Quoc V. Le, James Laudon, Richard Ho, Roger Carpenter and Jeff Dean.

We especially appreciate Joe Wenjie Jiang, Ebrahim Songhori, Young-Joon Lee, Roger Carpenter, and Sergio Guadarrama’s continued efforts to land this production impact, Quoc V. Le for his research advice and mentorship, and our senior author Jeff Dean for his support and deep technical discussions.

We also want to thank Ed Chi, Zoubin Ghahramani, Koray Kavukcuoglu, Dave Patterson, and Chris Manning for all of their advice and support.



Source link

Continue Reading
Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

AI Research

Google invests £5 billion in AI, research, training and data center in the UK

Published

on


Google plans to invest 5 billion pounds (about 6.78 billion US dollars) in AI infrastructure and other projects in the UK over the next two years, according to the Wall Street Journal.

The company says the funds will also support energy supply, research, engineering, and workforce training. At the same time, Google has opened a new data center north of London to meet the growing demand for services like Cloud, Maps, Workspace, and Search.

Other US tech giants are also ramping up their investments across Europe. Oracle has announced 3 billion dollars for projects in Germany and the Netherlands, Microsoft is putting 4.75 billion dollars into Italy, and Amazon is making multi-billion dollar investments in cloud and logistics centers in Germany and Spain. OpenAI is moving ahead with a major European project as well, called “Stargate Norway.”



Source link

Continue Reading

AI Research

New WalkMe offering embeds training directly into apps – Computerworld

Published

on


Now, said Bickley, “imagine today’s alternative: a system that sees you perform a task correctly the first time, learns from it, and memorializes those sequenced steps, mouse clicks, and keystrokes, such that when the employee gets stuck, the system can pull them through to task completion. With many end users accessing dozens or hundreds of system transactions as part of their job, this functionality is invaluable — a real efficiency driver, and also a means to reduce risk via the built-in guardrails and guidance ensuring accurate data is entered into the system.”

He said, “WalkMe tracks users’ usage of their systems — where they stop, what they do, and where they run into problems. This existing baseline of ‘user context’ is a natural jumping-off point for an AI-assisted evolution of the product. The Visual No Code Editor is where the employee guidance flows are built, and the no-code, visual point and click nature of this tool enables business teams to build the training tools and not be dependent on developers.”

‘I see this as much more than Clippy for SAP’

“[The ability] to build highly granular workflows that can discern across user group segments (for example, role, device type, geography, behavior), coupled with already existing automation for things like auto-completion of form fields as an example, provides a meaningful nudge when an employee gets stuck on a process step,” said Bickley.



Source link

Continue Reading

AI Research

Notre Dame to host summit on AI, faith and human flourishing, introducing new DELTA framework | News | Notre Dame News

Published

on


Artificial intelligence is advancing at a breakneck pace, as governments and industries commit resources to its development at a scale not seen since the Space Race. These technologies have the potential to disrupt every aspect of life, including education, the economy, labor and human relationships.

“As a leading global Catholic research university, Notre Dame is uniquely positioned to help the world confront and understand AI’s benefits and risks to human flourishing,” said John T. McGreevy, the Charles and Jill Fischer Provost. “Technology ethics is a key priority for Notre Dame, and we are fully committed to bringing the wisdom of the global Church to bear on this critical theme.”

In support of this work, the Institute for Ethics and the Common Good and the Notre Dame Ethics Initiative will host the Notre Dame Summit on AI, Faith and Human Flourishing on the University’s campus from Monday, Sept. 22 through Thursday, Sept. 25. This event will draw together a dynamic, ecumenical group of educators, faith leaders, technologists, journalists, policymakers and young people who believe in the enduring relevance of Christian ethical thought in a world of powerful AI.

“As artificial intelligence becomes more powerful, the ‘ethical floor’ of safety, privacy and transparency is simply not enough,” said Meghan Sullivan, the Wilsey Family College Professor of Philosophy and the director of the Institute for Ethics and the Common Good and the Notre Dame Ethics Initiative. “This moment in time demands a response rooted in the Christian tradition — a richer, more holistic perspective that recognizes the nature of the human person as a spiritual, emotional, moral and physical being.”

Sullivan noted that a unified, faith-based response to AI is a priority of newly elected Pope Leo XIV, who has spoken publicly about the new challenges to human dignity, justice and labor posed by these technologies.

The summit will begin at 5:15 p.m. Monday with an opening Mass at the University’s Basilica of the Sacred Heart. His Eminence Cardinal Christophe Pierre, Apostolic Nuncio to the United States, will serve as primary celebrant and homilist with University President Rev. Robert A. Dowd, C.S.C., as concelebrant. All members of the campus community are invited to attend this opening Mass.

Summit speakers include Andy Crouch, Praxis; Alex Hartemink, Duke University; Molly Kinder, Brookings Institution; Andrew Schuman, Veritas Forum; Anne Snyder, Comment Magazine and Elizabeth Dias, The New York Times. Over the course of the summit, attendees will take part in use case workshops, panels and community of practice sessions focused on public engagement, ministry and education. Executives from Google, Microsoft, Apple and many other organizations are among the 200 invited guests who will attend.

At the summit, Notre Dame will launch DELTA, a new framework for guiding conversations about AI. DELTA — an acronym that stands for Dignity, Embodiment, Love, Transcendence and Agency — will serve as a practical resource across sectors that are experiencing disruption from AI, including homes, schools, churches and workplaces, while also providing a platform for credible, principled voices to promote moral clarity and human dignity in the face of advancing technology.

“Our goal is for DELTA to become a common lens through which to engage AI — a language that reflects the depth of the Christian tradition while remaining accessible to people of all faiths,” Sullivan said. “By bringing together this remarkable group of leaders here at Notre Dame, we’re launching a community that will work passionately to create — as the Vatican puts it — ‘a growth in human responsibility, values and conscience that is proportionate to the advances posed by technology.’”

Although the summit sessions are by invitation only, Sullivan’s keynote on DELTA will be livestreamed. Those interested are invited to view the livestream and learn more about DELTA at https://ethics.nd.edu/summit-livestream at 8:30 a.m. EST on Tuesday, Sept. 23.

The Notre Dame Summit on AI, Faith and Human Flourishing is supported with a grant provided by Lilly Endowment Inc.

Lilly Endowment Inc. is a private foundation created in 1937 by J.K. Lilly Sr. and his sons Eli and J.K. Jr. through gifts of stock in their pharmaceutical business, Eli Lilly and Company. While those gifts remain the financial bedrock of the Endowment, it is a separate entity from the company, with a distinct governing board, staff and location. In keeping with the founders’ wishes, the Endowment supports the causes of community development, education and religion and maintains a special commitment to its hometown, Indianapolis, and home state, Indiana. A principal aim of the Endowment’s religion grantmaking is to deepen and enrich the lives of Christians in the United States, primarily by seeking out and supporting efforts that enhance the vitality of congregations and strengthen the pastoral and lay leadership of Christian communities. The Endowment also seeks to improve public understanding of religious traditions in the United States and across the globe.

Contact: Carrie Gates, associate director of media relations, 574-993-9220, c.gates@nd.edu



Source link

Continue Reading

Trending