Trend 2: Continued Efficiency Improvements for ML
Improvements in efficiency — arising from advances in computer hardware design as well as ML algorithms and meta-learning research — are driving greater capabilities in ML models. Many aspects of the ML pipeline, from the hardware on which a model is trained and executed to individual components of the ML architecture, can be optimized for efficiency while maintaining or improving on state-of-the-art performance overall. Each of these different threads can improve efficiency by a significant multiplicative factor, and taken together, can reduce computational costs, including CO2 equivalent emissions (CO2e), by orders of magnitude compared to just a few years ago. This greater efficiency has enabled a number of critical advances that will continue to dramatically improve the efficiency of machine learning, enabling larger, higher quality ML models to be developed cost effectively and further democratizing access. I’m very excited about these directions of research!

Continued Improvements in ML Accelerator Performance
Each generation of ML accelerator improves on previous generations, enabling faster performance per chip, and often increasing the scale of the overall systems. Last year, we announced our TPUv4 systems, the fourth generation of Google’s Tensor Processing Unit, which demonstrated a 2.7x improvement over comparable TPUv3 results in the MLPerf benchmarks. Each TPUv4 chip has ~2x the peak performance per chip versus the TPUv3 chip, and the scale of each TPUv4 pod is 4096 chips (4x that of TPUv3 pods), yielding a performance of approximately 1.1 exaflops per pod (versus ~100 petaflops per TPUv3 pod). Having pods with larger numbers of chips that are connected together with high speed networks improves efficiency for larger models.

ML capabilities on mobile devices are also increasing significantly. The Pixel 6 phone features a brand new Google Tensor processor that integrates a powerful ML accelerator to better support important on-device features.

Our use of ML to accelerate the design of computer chips of all kinds (more on this below) is also paying dividends, particularly to produce better ML accelerators.

Continued Improvements in ML Compilation and Optimization of ML Workloads
Even when the hardware is unchanged, improvements in compilers and other optimizations in system software for machine learning accelerators can lead to significant improvements in efficiency. For example, “A Flexible Approach to Autotuning Multi-pass Machine Learning Compilers” shows how to use machine learning to perform auto-tuning of compilation settings to get across-the-board performance improvements of 5-15% (and sometimes as much as 2.4x improvement) for a suite of ML programs on the same underlying hardware. GSPMD describes an automatic parallelization system based on the XLA compiler that is capable of scaling most deep learning network architectures beyond the memory capacity of an accelerator and has been applied to many large models, such as GShard-M4, LaMDA, BigSSL, ViT, MetNet-2, and GLaM, leading to state-of-the-art results across several domains.

Human-Creativity–Driven Discovery of More Efficient Model Architectures
Continued improvements in model architectures give substantial reductions in the amount of computation needed to achieve a given level of accuracy for many problems. For example, the Transformer architecture, which we developed in 2017, was able to improve the state of the art on several NLP and translation benchmarks while simultaneously using 10x to 100x less computation to achieve these results than a variety of other prevalent methods, such as LSTMs and other recurrent architectures. Similarly, the Vision Transformer was able to show improved state-of-the-art results on a number of different image classification tasks despite using 4x to 10x less computation than convolutional neural networks.

Machine-Driven Discovery of More Efficient Model Architectures
Neural architecture search (NAS) can automatically discover new ML architectures that are more efficient for a given problem domain. A primary advantage of NAS is that it can greatly reduce the effort needed for algorithm development, because NAS requires only a one-time effort per search space and problem domain combination. In addition, while the initial effort to perform NAS can be computationally expensive, the resulting models can greatly reduce computation in downstream research and production settings, resulting in greatly reduced resource requirements overall. For example, the one-time search to discover the Evolved Transformer generated only 3.2 tons of CO2e (much less than the 284t CO2e reported elsewhere; see Appendix C and D in this joint Google/UC Berkeley preprint), but yielded a model for use by anyone in the NLP community that is 15-20% more efficient than the plain Transformer model. A more recent use of NAS discovered an even more efficient architecture called Primer (that has also been open-sourced), which reduces training costs by 4x compared to a plain Transformer model. In this way, the discovery costs of NAS searches are often recouped from the use of the more-efficient model architectures that are discovered, even if they are applied to only a handful of downstream uses (and many NAS results are reused thousands of times).

NAS has also been used to discover more efficient models in the vision domain. The EfficientNetV2 model architecture is the result of a neural architecture search that jointly optimizes for model accuracy, model size, and training speed. On the ImageNet benchmark, EfficientNetV2 improves training speed by 5–11x while substantially reducing model size over previous state-of-the-art models. The CoAtNet model architecture was created with an architecture search that uses ideas from the Vision Transformer and convolutional networks to create a hybrid model architecture that trains 4x faster than the Vision Transformer and achieves a new ImageNet state of the art.

The broad use of search to help improve ML model architectures and algorithms, including the use of reinforcement learning and evolutionary techniques, has inspired other researchers to apply this approach to different domains. To aid others in creating their own model searches, we have open-sourced Model Search, a platform that enables others to explore model search for their domains of interest. In addition to model architectures, automated search can also be used to find new, more efficient reinforcement learning algorithms, building on the earlier AutoML-Zero work that demonstrated this approach for automating supervised learning algorithm discovery.

Use of Sparsity
Sparsity, where a model has a very large capacity, but only some parts of the model are activated for a given task, example or token, is another important algorithmic advance that can greatly improve efficiency. In 2017, we introduced the sparsely-gated mixture-of-experts layer, which demonstrated better results on a variety of translation benchmarks while using 10x less computation than previous state-of-the-art dense LSTM models. More recently, Switch Transformers, which pair a mixture-of-experts–style architecture with the Transformer model architecture, demonstrated a 7x speedup in training time and efficiency over the dense T5-Base Transformer model. The GLaM model showed that transformers and mixture-of-expert–style layers can be combined to produce a model that exceeds the accuracy of the GPT-3 model on average across 29 benchmarks using 3x less energy for training and 2x less computation for inference. The notion of sparsity can also be applied to reduce the cost of the attention mechanism in the core Transformer architecture.

The use of sparsity in models is clearly an approach with very high potential payoff in terms of computational efficiency, and we are only scratching the surface in terms of research ideas to be tried in this direction.

Each of these approaches for improved efficiency can be combined together so that equivalent-accuracy language models trained today in efficient data centers are ~100 times more energy efficient and produce ~650 times less CO2e emissions, compared to a baseline Transformer model trained using P100 GPUs in an average U.S. datacenter using an average U.S. energy mix. And this doesn’t even account for Google’s carbon-neutral, 100% renewable energy offsets. We’ll have a more detailed blog post analyzing the carbon emissions trends of NLP models soon.

In this post, I’ll be showing some amazing ways the Vision API can extract meaning from your images  – keep reading, or jump directly into a tutorial using Python, Node.js, Go, or Java! This tutorial can be completed at no cost within the Google Cloud Free Tier.

They say a picture is worth a thousand words. But how do you make those words available and useful? Around the world, we are generating more images than ever before, and it’s no surprise that businesses are turning to image recognition technology to help meet the immense opportunities created with this growing set of data.

Cloud Vision API is a powerful tool that enables you to perform a variety of tasks including label detection, text recognition, and object tracking on your image data. Whether it’s identifying products in a retail store, analyzing social media posts for brand mentions, or scanning through millions of images to find a specific object, the Cloud Vision API can help businesses automate their image analysis workflows and gain valuable insights from their visual data.  To protect privacy, and help you build responsibly, the Cloud Vision API offers features to limit personal identification, such as person blur, which hides identifiable features. 

Let’s explore a few of the key features of the Cloud Vision API.

Detect famous landmarks

Landmark detection allows you to analyze images to identify specific landmarks such as buildings, natural features, and other recognizable locations. Cloud Vision API recognizes landmarks and provides information about them, including their name, location, and other relevant details. Perhaps you are trying to identify the landmarks in images shared by customers as part of social campaigns, or want to build a mobile app that provides information to tourists on famous landmarks.

In the below left-hand side image, Cloud Vision API has detected the Eiffel Tower, shown in the visualized response. Not shown in this visualization here, but also detected, were Pont de Bir-Hakeim (the bridge) and Champs de Mars (the park in front of the Eiffel Tower).

Detect objects and label images

Object detection and labels are two related features that enable you to identify and classify objects within an image. Object detection detects and locates objects within an image, and provides information such as the position, size, and orientation of each object.  Labels, on the other hand, provide a general classification of the content within an image.

Object detection has practical applications in many industries such as self-driving vehicles (where it’s critical), retail, manufacturing and more, while labels can be used to help classify and organize large collections of images, or to categorize and filter content.

You can see the similarities and differences in the responses provided by the object detection and labeling features in this image taken in Setagaya.

Detect text

Cloud Vision API detects and extracts text from any image, even if it’s handwritten or in different languages. Once it detects text, the API can provide information about the position, orientation, and size of each text element, as well as individual words, and their bounding boxes.

In this image of a traffic sign, Cloud Vision API has detected the text and provided it in the response.

Detect explicit content

Cloud Vision API can automatically identify and flag explicit or inappropriate content within an image using five categories: adult, spoof, medical, violence, and racy. The API provides a score that indicates the likelihood for each category in the image, which you can use to set thresholds in your application and decide how to handle those that exceed them. This feature is particularly useful for filtering or moderating user-generated content. 

Luckily for the images I shared here, each category has been deemed “very unlikely” to be present. Phew!

Next Steps

These are just a few features of the Cloud Vision API and how it can help your business with automating image analysis workflows and gaining valuable insights from your visual data. 

Head to the interactive walkthrough tutorials in Python, Node.js, Go, and Java to see step-by-step how to access the API and learn more about all the features that you can integrate into your own applications! Again, this tutorial can be completed at no cost within the Google Cloud Free Tier.

It’s no secret that many organizations and job seekers find the hiring process exhausting. It can be time consuming, costly, and somewhat risky for both parties. Those are just some of the experiences we wanted to change when we started Gyfted, a pre-vetted talent marketplace for people who complete tech training or degree programs and are looking for the right career move. At the same time, we’re helping businesses save time and improve recruiting outcomes with our automated candidate screening and sourcing tools.

Our vision is clear: To take a candidate through one structured hiring process, and then put them in front of thousands of companies. It’s similar to the common app system in higher education. Sounds simple, but it is a herculean technical and UX task. To succeed we had to combine advanced psychometric testing, machine learning, the latest in behavioral design, and develop the highest quality structured, relational dataset to represent candidate and manager profiles and preferences on our network. Fortunately, we were cofounded by world-leading experts in these areas including Dr. Michal Kosinski, one of the world’s top computational psychologists, and Adam Szefer, a gifted young technologist. We’ve been joined by a group of equally talented employees, most of whom work remotely in Poland, US, Switzerland, UK, Israel, and Ukraine.

The influence of dating platforms and matching

When seeking inspiration, we were influenced by the success of dating platforms, especially Bumble with its focus on commitment. These platforms have done a great job using design to match people together.

We like to think we’re doing the same for recruiters and candidates in terms of not only role and culture fit matching, but also through a fundamental feature of Gyfted, which is that our job-seekers are anonymous. This helps recruiters meet one of their goals today, which is to minimize bias in the hiring process and enable objective, diversity-oriented recruiting.

Another of our unique selling points is that we conduct candidate screening that is gamified and automated, with a structured interview, where the interview remains with the candidate’s profile. We roughly estimate that just for the 15 million open jobs on LinkedIn, if companies fill in 50% of those via external recruiting, and conduct a screening interview with 10 candidates per job filled at $50/hour paid to an employee to do the screening, that’s $3.75 billion and 75 million hours in direct costs. This is on top of applying for jobs, selecting CVs, and coordinating the process, which takes an even bigger financial and time toll on both applicants and recruiters. Instead, it would be better to take one interview for 1000 companies. The impact of what we want to achieve with our vision is enormous.

We also offer career discovery and career search tools for job candidates. This includes free, personalized feedback for every job-seeker. Right now, we’re aiming the service at students, bootcamp graduates and juniors, helping them to land jobs in tech and the creative industry at large. Next, we’ll expand into mid and senior roles. In the long run we want to reshape how recruiting happens through a common app that saves everyone in the market significant time and resources, helping people find jobs not only faster, but jobs that truly fit them.

Developing advanced AI applications with Google Cloud

We obtained our original funding from angel and institutional investors, and we were selected into the Fall 2021 batch of StartX, the non-profit start-up accelerator and founder community associated with Stanford University. But like most startups our budgets are tight, and we need to find ways to operate as efficiently as possible, especially when building out our technology stack and developer environment.

That’s where Google Cloud comes in. It’s a lot more affordable and flexible than competing solutions, and our developers love it. We use Google App Engine for the hosting and development of our applications giving us enormous flexibility. Vertex AI enables us to build, deploy, and scale machine learning models faster, within a unified artificial intelligence platform. On top of that we use Google Vertex AI Workbench as the development environment for the data science workflow, which allows us to have everything that we need to host and develop innovative AI-based applications.

BigQuery, Google Cloud’s serverless data warehouse, is another stand-out solution for us. We use it to crunch big data from all our systems and the UX is very intuitive and easy to use, allowing us to use it across the business and get insights from a wide range of employees, not just technical experts.

Above all, Google Cloud helps us solve the main platform challenges facing Gyfted including scalability and identity management, so we are perfectly positioned for growth. Right now, we handle about 2 million candidate interactions, a volume we expect to grow exponentially. As that number grows, we rely on Google Cloud to help us scale securely and with reliability.

Eliminating bias from the hiring process

Our technology partners have also been integral to helping us get to an advanced stage of our beta program. MongoDB on Google Cloud takes the data burden off our teams and reduces time to value of our applications. We can stay nimble and can scale database capacity at the push of a button.

Our collaboration with the Google team has been fantastic. Our Startup Success Manager is an expert when it comes to Google Cloud solutions, and he also understands our business from his own experience as an entrepreneur and an investor. It’s great to have an internal point of contact who can help us navigate all of Google’s resources.

I’d also stress the extent to which Google Cloud values align with ours. For example, a key benefit for our customers is the ability to strip unconscious bias out of the hiring process. Google Cloud tools support this commitment to diversity, especially when we are building out our AI models.

On a team level, we also appreciate the support that Google Cloud has shown through its Google Support Fund for Start-ups in Ukraine. This has helped many Ukrainian businesses to continue to operate at a very challenging time, including startups with remote, distributed teams in Poland where most of us stem from.

If I had to sum up Google Cloud and our collaboration with Google for Startups in a phrase, I’d say that it adds enormous value to our business while removing much of the risk when scaling up a start-up. We’ve seen the addition of many new tools and features in the past two years and our Google mentors are always looking at the best way these can be integrated with Gyfted’s own roadmap. That means that we can continue to transform recruiting and hiring processes with the support of one of the world’s most advanced tech companies as a strategic growth partner.

If you want to learn more about how Google Cloud can help your startup, visit our page here to get more information about our program, and sign up for our communications to get a look at our community activities, digital events, special offers, and more.

Startups need to move quickly and focus their limited resources on areas where they can differentiate. If infrastructure isn’t your differentiator, don’t put a lot of energy into your infrastructure when someone else can do it for you. What’s more, time-to-market matters to startups more now than ever. Successful early stage companies know this, and leverage existing tools, libraries, frameworks and innovations whenever possible.  

Next week Google Cloud will be featured at TechCrunch Disrupt, the iconic annual event where “founders and investors shaping the future of disruptive technology” come together to share stories, network, and learn from each other. Google Cloud will host a session and Google Cloud engineers will be available in a virtual booth to answer startup questions. 

The session, “Demo Derby – How startups are disrupting the status quo with innovative data analytics, AI and modern app development” will be a fun, fast-paced set of presentations showing short demos of startups and startup projects built with Google Cloud. 

Demo Derby Presenters

• Andrea Le Vot, Chief Data Protection Officer with startup BlueZoo will deliver a demo showing how they are revolutionizing the property insurance industry with AI and Cloud – and doing so while actually protecting people’s privacy.
• Dale Markowitz and Zack Akil, Applied AI Engineers, will show how easy it is to use AI for everything from video search to editing to automated translation.
• Vidya Nagarajan, Group Product Manager from Google Cloud will deliver a fast paced demo that shows how startups can drive developer productivity with serverless innovations.

“Disrupt is iconic for its engagement with founders, investors, and the broad community of early stage companies,” said Andrea Le Vot, Chief Data Protection Officer at BlueZoo. “We are thrilled to share our insights with others in our community, and to show how we have partnered with Google Cloud to innovate faster than most of the larger, well funded insurance companies in our market.”

The series of snippet sized demos will be followed by a roundtable discussion with the demo developers focused on lessons learned, best practices, how to reduce time-to-market, and how to focus on where you can most effectively differentiate.

The session will air on Day One of the event: Tuesday September 21, 2021 at 1pm PT and will be available on demand after the event for all Disrupt attendees.

Siemens aims to shape the future—and has the scale, culture, and know-how to realize its ambitions. The Germany-headquartered industrial manufacturing business operates in more than 120 countries and focuses on electrification, automation, and digitalization. Siemens’ extensive portfolio includes industrial production digitalization and automation technologies, diagnostic and therapeutic imaging for healthcare, and building automation and management technologies.

“All our solutions are united by the fact they impact the way people live their lives,” says Stephanie Morton, Siemens’ Global Talent Acquisition Manager: Strategy, Technology & Talent Relationship Management.

“‘Engineered for life’ is one of our brand claims.”

Hiring “future makers”

Hiring thinkers, dreamers, and doers hungry to transform businesses, industries, and lives is key to Siemens’ success. The business calls its people “future makers” and employs more than 370,000 of them worldwide.

Recruiting for a workforce this size is no easy task. Siemens’ global jobs and careers website has about 5,000 open positions at any one time, and the organization receives 2 million applications to fill 35,000 positions per year.

However, by 2017, the website was experiencing problems that hampered the talent acquisition team’s efforts to fill roles promptly with the right candidates. Its standard keyword matching technology could not optimize the results provided to job seekers, frustrating potential candidates and increasing the workload for recruiters.

“With Cloud Talent Solution, we saw a 30 percent uplift in candidate conversions from search to application.”

—Stephanie Morton, Global Talent Acquisition Manager: Strategy, Technology & Talent Relationship Management, Siemens

“We found the language we used about our jobs internally was often not the same language job seekers used,” explains Morton. “For example, we may have an engineering position open for our MindSphere cloud-based IoT operating system. We would post this on our jobs and careers website as ‘MindSphere engineer.’ Unfortunately, a job seeker using more general terms such as ‘IoT engineer’ may miss this advertisement.”

In addition, without intelligence and context, the keyword matching technology could not determine job seeker search intentions from misspelled terms. This became more of a problem as job seekers increasingly used mobile devices to conduct searches.

With keyword searches also generating pages and pages of results, job seekers often succumbed to the temptation to apply for every position returned. This left the talent acquisition team drowning in messages and applications.

“We experienced issues around seniority as well—for example, people looking for senior legal roles were receiving irrelevant results for internships or junior roles,” says Morton.

Change needed

These problems could not persist for a team charged with continuously improving candidate experiences. The team opened discussions with Jibe, a recruitment platform provider that operated Siemens’ external career websites. Jibe had collaborated with Google to integrate Cloud Talent Solution—a solution that uses machine learning to better understand job content and job seeker intent—into its platform.” Jibe offered the integrated solution to Siemens as one of its most forward-thinking customers,” says Morton.

Morton’s team reviewed Cloud Talent Solution and was immediately excited by its potential to quickly make a profound difference in its recruitment activities—without requiring recruiters and other team members to invest considerable time and effort. Cloud Talent Solution could enable the website to understand the broad intent of a job seeker—including synonymous positions—and deliver considerably more relevant results.

“We can authoritatively say Cloud Talent Solution is improving the experience for job seekers; they are conducting more searches, those searches are more effective, and more of them are converting to job applications.”

—Stephanie Morton, Global Talent Acquisition Manager: Strategy, Technology & Talent Relationship Management, Siemens

After Cloud Talent Solution expanded to encompass more than 100 languages—enabling Siemens to offer a consistent experience to candidates from a range of countries and backgrounds—Morton’s team gave Jibe and Google the green light to proceed.

“We had two key issues to address during the implementation,” says Morton. “The first was to make sure Google Cloud was aware of any Siemens-specific recruitment terms. For example, we have German-language posts that include terms commonly used internally regarding internships and junior positions. We worked closely with our partners to make sure these terms were represented in the algorithm to complement the vast constellation of already-mapped job titles in Cloud Talent Solution.”

Siemens also needed to quantify Cloud Talent Solution benefits to sell the service internally within the business and win support to proceed beyond a pilot. “Because Cloud Talent Solution was effectively an invisible layer behind our website, we had to test its impact,” says Morton.

A 30 percent uplift

Working with Jibe, Siemens conducted A/B testing. Five percent of candidates in the test group used the keyword-based search experience and 95 percent of candidates the search experience powered by Cloud Talent Solution. The testing accommodated changes in candidate queries over time and context.

“With Cloud Talent Solution, we saw a 30 percent uplift in conversions from search to application,” says Morton.

“We can authoritatively say Cloud Talent Solution is improving the experience for job seekers; they are conducting more searches, those searches are more effective, and more of them are converting to job applications,” she adds. “Anecdotally, recruiters are seeing fewer applications from job seekers who have applied to long lists of positions returned from search queries.”

“Partnering with market-leading providers like Google Cloud can help us stay one step ahead of the competition over the journey.”

—Stephanie Morton, Global Talent Acquisition Manager: Strategy, Technology & Talent Relationship Management, Siemens

In the longer term, Siemens plans to monitor the impact of Cloud Talent Solution on metrics such as candidate retention rates and manager satisfaction.

With Cloud Talent Solution well established within the business, Siemens is now working with Google Cloud to add new features—including voice assistants—to enhance the candidate experience. “Helping candidates laser in on the right positions and make successful connections is something we’re doing everything we can to make happen,” says Morton.

More broadly, Cloud Talent Solution and Jibe are helping Siemens’ talent acquisition and the rest of its human resources function build its reputation and execute its strategy. “We are becoming a function driven by delivering amazing technology-based solutions to our employees and our candidates,” says Morton. “People within Siemens are very impressed to see human resources leading the way in working with Google Cloud to deliver a solution like this at a global scale.”

“The more light-touch exercises like this we can do, the fewer big, expensive, time-consuming initiatives we need to take,” she says. “Furthermore, partnering with market-leading providers like Google Cloud can help us stay one step ahead of the competition over the journey.”