Tag Archives: AI

This Googler’s team is making shopping more inclusive

There’s a lot to love about online shopping: It’s fast, it’s easy and there are a ton of options to choose from. But there’s one obvious challenge — you can’t try anything on. This is something Google product manager Debbie Biswas noticed, as a tech industry veteran and startup founder herself. “Historically, the fashion industry only celebrates people of a certain size and skin color,” she says. “This was something I wanted to change.”

Debbie grew up in India and moved to the U.S. after she graduated college. “I started a company in the women's apparel space, where I learned to solve user pain points around shopping for clothes, sizing and styling.” While working on her startup, Debbie realized how hard shopping was for women, including herself — the models in the images didn’t show her how something would look on her. 

“When I got an opportunity to work at Google Shopping, I realized I could solve so many of these problems at scale using the best AI/ML tech in the industry,” she says. “As a woman of color, and someone who doesn't conform to the ‘traditional beautiful size,’ I feel very motivated to solve apparel shopping problems for people like me.”

A look at Style AI in action.

This was what Debbie and her team wanted to accomplish with Style AI. Style AI is a Shopping feature that helps people see how a product looks on various types of body styles and offers styling advice. Style AI works by using a machine learning algorithm to look at a specific product and visually understand it. “So if someone searches ‘gingham long sleeve shirt,’ Style AI will look at images of long-sleeved gingham shirts, apply our vision recognition technology and understand things like the pattern and the sleeve length and show users fashions that might interest them.” In order to make sure Style AI was inclusive of all different types of shapes, sizes and skin tones Debbie consulted with Google’s Product Fairness, or ProFair, team. ProFair helps teams at Google apply the AI Principles by investigating fairness issues. Together, they find ways to build inclusive services, strengthen equity in data labels and promote fairness and combat bias in AI. 

ProFair held sessions where everyone involved in the project could look for “fairness issues,” which helped Debbie’s team adjust how they designed Style AI. And there was much to consider. “First, we need to be careful of what data we train a model on. If you tell a machine that a certain size and skin color is what it needs to look for, it will,” Debbie explains. “So as responsible product owners, we need to make sure we train it the right way. Even after this, a machine can make many mistakes unknowingly — for example, not realizing that a certain style can be very offensive in one culture and be totally cool in another.” 

For instance, before launching in countries like India and Brazil, ProFair held local focus groups in collaboration with Google’s Product Inclusion team. Debbie says this helped her team find diverse images and clothing for these specific demographics. Debbie’s — and the entire team’s — ultimate goal is that shoppers will feel like they’re seeing themselves when they look for clothing. “Looking at stock product images does not help you decide on your purchase,” she says. “We just always think about what people told us while we were building Style AI: ‘I want to see the product on someone like me!’”

These researchers are bringing AI to farmers

“Farmers feed the entire world — so how might we support them to be resilient and build sustainable systems that also support global food security?” It’s a question that Diana Akrong found herself asking last year. Diana is a UX researcher based in Accra, Ghana, and the founding member of Google’s Accra UX team.

Across the world, her manager Dr. Courtney Heldreth, was equally interested in answering this question. Courtney is a social psychologist and a staff UX researcher based in Seattle, and both women work as part of Google’s People + Artificial Intelligence Research (PAIR) group. “Looking back on history, we can see how the industrial revolution played a significant role in creating global inequality,” she says. “It set most of Western Europe onto a path of economic dominance that was then followed by both military and political dominance.” Courtney and Diana teamed up on an exploratory effort focused on how AI can help better the lives of small, local farming communities in the Global South. They and their team want to understand what farmers need, their practices, value systems, what their social lives are like — and make sure that Google products reflect these dynamics.

One result of their work is a recently published research paper. The paper — written alongside their colleagues Dr. Jess Holbrook at Google and Dr. Norman Makoto Su of Indiana University and published in the ACM Interactions trade journal — dives into why we need farmer-centered AI research, and what it could mean not just for farmers, but for everyone they feed. I recently took some time to learn more about their work.

How would you explain your job to someone who isn't in tech?

Courtney: I would say I’m a researcher trying to understand underserved and historically marginalized users’ lives and needs so we can create products that work better for them. 

Diana: I’m a researcher who looks at how people interact with technology. My superpower is my curiosity and it’s my mission to understand and advocate for user needs, explore business opportunities and share knowledge.

What’s something on your mind right now? 

Diana: Because of COVID-19, there’s the threat of a major food crisis in India and elsewhere. We’re wondering how we can work with small farms as well as local consumers, policymakers, agricultural workers, agribusiness owners and NGOs to solve this problem.

Agriculture is very close to my heart, personally. Prior to joining Google, I spent a lot of time learning from smallholder farmers across my country and helping design concepts to address their needs. 

“Farmers feed the entire world — so how might we support them to be resilient and build sustainable systems that also support global food security?” Diana Akrong
UX researcher, Google

Courtney: I’ve been thinking about how AI can be seen as this magical, heroic thing, but there are also many risks to using it in places where there aren’t laws to protect people. When I think about Google’s AI Principles — be socially beneficial, be accountable to people, avoid reinforcing bias, prioritize safety — those things define what projects I want to work on. It’s also why my colleague Tabitha Yong and I developed a set of best practices for designing more equitable AI products.

Can you tell me more about your paper, “What Does AI Mean for Smallholder Farmers? A Proposal for Farmer-Centered AI Research,” recently published in ACM Interactions

Courtney: The impact and failures of AI are often very western and U.S.-centric. We’re trying to think about how to make this more fair and inclusive for communities with different needs around the globe. For example, in our farmer-centered AI research, we know that most existing AI solutions are designed for large farms in the developed world. However, many farmers in the Global South live and work in rural areas, which trail behind urban areas in terms of connectivity and digital adoption. By focusing on the daily realities of these farmers, we can better understand different perspectives, especially those of people who don’t live in the U.S. and Europe, so that Google’s products work for everyone, everywhere.

Why did you want to work at Google?

Diana: I see Google as home to teams with diverse experiences and skills who work collaboratively to tackle complex, important issues that change real people’s lives. I’ve thrived here because I get to work on projects I care about and play a critical role in growing the UX community here in Ghana.

Courtney: I chose Google because we work on the world's hardest problems. Googlers are  fearless and the reach of Google’s products and services is unprecedented. As someone who comes from an underrepresented group, I never thought I would work here. To be here at this moment is so important to me, my community and my family. When I look at issues I care about the most — marginalized and underrepresented communities — the work we do plays a critical role in preventing algorithmic bias, bridging the digital divide and lessening these inequalities. 

How have you seen your research help real people? 

Courtney: In 2018, we worked with Titi Akinsanmi, Google’s Policy and Government Relations Lead for West and Francophone Africa, and PAIR Co-lead and Principal Research Scientist Fernanda Viegas on the report for AI in Nigeria. Since then, the Ministry of Technology and Science reached out to Google to help form a strategy around AI. We’ve seen government bodies in sub-Saharan Africa use this paper as a roadmap to develop their own responsible AI policies.

How should aspiring AI thinkers and future technologists prepare for a career in this field?

Diana: My main advice? Start with people and their needs. A digital solution or AI may not be necessary to solve every problem. The PAIR Guidebook is a great reference for best practices and examples for designing with AI.

A crossword puzzle with a big purpose

Before the pandemic, Alicia Chang was working on a new project. “I was experimenting with non-traditional ways to help teach Googlers the AI Principles,” she says. Alicia is a technical writer on the Engineering Education team focused on designing learning experiences to help Googlers learn about our AI Principles and how to apply them in their own work.

The challenge for Alicia would be how many people she needed to educate. “There are so many people spread over different locations, time zones, countries!” But when the world started working from home, she was inspired by the various workarounds people were using to connect virtually. 

A photo of Alicia Chang sitting on a bench outside. She is looking into the camera and smiling.

Alicia Chang

“I started testing out activities like haiku-writing contests and online trivia,” Alicia says. “Then one day a friend mentioned an online escape room activity someone had arranged for a COVID-safe birthday gathering. Something really clicked with me when she mentioned that, and I started to think about designing an immersive learning experience.” Alicia decided to research how some of the most creative, dedicated people deliver information: She looked at what teachers were doing. 

Alicia soon stumbled upon a YouTube video about using Google Sheets to create a crossword puzzle, so she decided to make her own — and Googlers loved it. Since the crossword was such a success, Alicia decided to make more interactive games. She used Google Forms to create a fun “Which AI Principle are you?” quiz, and Google Docs to make a word search. Then there’s the Emoji Challenge, where players have to figure out which AI Principles a set of emoji describe. All of this became part of what is now known as the Responsible Innovation Challenge, a set of various puzzle activities built with Google products — including Forms, Sheets, Docs and Sites — that focus on teaching Google’s AI Principles.

The purpose of the Responsible Innovation Challenge is to introduce Google’s AI Principles to new technical hires in onboarding courses, and to help Googlers put the AI Principles into practice in everyday product development situations. The first few puzzles are fairly simple and help players remember and recall the Principles, which serve as a practical framework for responsible innovation. As Googlers start leveling up, the puzzles get a bit more complex.. There’s even a bonus level where Googlers are asked to think about various technical resources and tools they can use to develop AI responsibly by applying them in their existing workflow when creating a machine learning model.

Alicia added a points system and a leaderboard with digital badges — and even included prizes. “I noticed that people were motivated by some friendly competition. Googlers really got involved and referred their coworkers to play, too,” she says. “We had over 1,000 enroll in the first 30 days alone!” To date, more than 2,800 Googlers have participated from across 41 countries, and people continue to sign up. 

It’s been encouraging for Alicia to see how much Googlers are enjoying the puzzles, especially when screen time burnout is all too real. Most importantly, though, she’s thrilled that more people are learning about Google’s AI Principles. “Each of the billions of people who use Google products has a unique story and life experience,” Alicia says. “And that’s what we want to think about so we can make the best products for individual people.” 

Machine Learning GDEs: Q2 ‘21 highlights and achievements

Posted by HyeJung Lee, MJ You, ML Ecosystem Community Managers

Google Developers Experts (GDE) is a community of passionate developers who love to share their knowledge with others. Many of them specialize in Machine Learning (ML).

Here are some highlights showcasing the ML GDEs achievements from last quarter, which contributed to the global ML ecosystem. If you are interested in becoming an ML GDE, please scroll down to see how you can apply!

ML Developers meetup @Google I/O

ML Developer meetup at Google I/O

At I/O this year, we held two ML Developers Meetups (America/APAC and EMEA/APAC). Merve Noyan/Yusuf Sarıgöz (Turkey), Sayak Paul/Bhavesh Bhatt (India), Leigh Johnson/Margaret Maynard-Reid (USA), David Cardozo (Columbia), Vinicius Caridá/Arnaldo Gualberto (Brazil) shared their experiences in developing ML products with TensorFlow, Cloud AI or JAX and also introduced projects they are currently working on.

I/O Extended 2021

Chart showing what's included in Vertex AI

After I/O, many ML GDEs posted recap summaries of the I/O on their blogs. Chansung Park (Korea) outlined the ML keynote summary, while US-based Victor Dibia wrapped up the Top 10 Machine Learning and Design Insights from Google IO 2021.

Vertex AI was the topic of conversation at the event. Minori Matsuda from Japan wrote a Japanese article titled “Introduction of powerful Vertex AI AutoML Forecasting.” Similarly, Piero Esposito (Brazil) posted an article titled “Serverless Machine Learning Pipelines with Vertex AI: An Introduction,” including a tutorial on fully customized code. India-based Sayak Paul co-authored a blog post discussing key pieces in Vertex AI right after the Vertex AI announcement showing how to run a TensorFlow training job using Vertex AI.

Communities such as Google Developers Groups (GDG) and TensorFlow User Groups (TFUG) held extended events where speakers further discussed different ML topics from I/O, including China-based Song Lin’s presentation on TensorFlow highlights and Applications experiences from I/O which had 24,000 online attendees. Chansung Park (Korea) also gave a presentation on what Vertex AI is and what you can do with Vertex AI.

Cloud AI

Cloud AI

Leigh Johnson (USA) wrote an article titled Soft-launching an AI/ML Product as a Solo Founder, covering GCP AutoML Vision, GCP IoT Core, TensorFlow Model Garden, and TensorFlow.js. The article details the journey of a solo founder developing an ML product for detecting printing failure for 3D printers (more on this story is coming up soon, so stay tuned!)

Demo and code examples from Victor Dibia (USA)’s New York Taxi project, Minori Matsuda (Japan)’s article on AutoML and AI Platform notebook, Srivatsan Srinivasan (USA)’s video tutorials, Sayak Paul (India)’s Distributed Training in TensorFlow with AI Platform & Docker and Chansung Park (Korea)’s curated personal newsletter were all published together on Cloud blog.

Aqsa Kausar (Pakistan) gave a talk about Explainable AI in Google Cloud at the International Women’s Day Philippines event. She explained why it is important and where and how it is applied in ML workflows.

Learn agenda

Finally, ML Lab by Robert John from Nigeria, introduces the ML landscape on GCP covering from BigQueryML through AutoML to TensorFlow and AI Platform.


Image of TensorFlow 2 and Learning TensorFlow JS books

Eliyar Eziz (China) published a book “TensorFlow 2 with real-life use cases”. Gant Laborde from the US authored book “Learning TensorFlow.js” which is published by O'Reilly and wrote an article “No Data No Problem - TensorFlow.js Transfer Learning” about seeking out new datasets to boldly train where no models have trained before. He also published “A Riddikulus Dataset” which talks about creating the Harry Potter dataset.

Iterated dilated convolutional neural networks for word segmentation

Hong Kong-based Guan Wang published a research paper, “Iterated Dilated Convolutional Neural Networks for Word Segmentation,” covering state-of-the-art performance improvement, which is implemented on TensorFlow by Keras.

Elyes Manai from Tunisia wrote an article “Become a Tensorflow Certified Developer ” - a guide to TensorFlow Certificate and tips.

BERT model

Greece-based George Soloupis wrote a tutorial “Fine-tune a BERT model with the use of Colab TPU” on how to finetune a BERT model that was trained specifically on greek language to perform the downstream task of text classification, using Colab’s TPU (v2–8).


India-based Aakash Nain has published the TF-JAX tutorial series (Part1, Part2, Part3, Part 4), aiming to teach everyone the building blocks of TensorFlow and JAX frameworks.

TensorFlow with Jax thumbnail

Online Meetup TensorFlow and JAX by Tzer-jen Wei from Taiwan covered JAX intro and use cases. It also touched upon different ways of writing TensorFlow models and training loops.

Neural Networks, with a practical example written in JAX

YouTube video Neural Networks, with a practical example written in JAX, probably the first JAX techtalk in Portuguese by João Guilherme Madeia Araújo (Brazil).


Keras logo

A lot of Keras examples were contributed by Sayak Paul from India and listed below are some of these examples.


Kaggle character distribution chart

Notebook “Simple Bayesian Ridge with Sentence Embeddings” by Ertuğrul Demir (Turkey) about a natural language processing task using BERT finetuning followed by simple linear regression on top of sentence embeddings generated by transformers.

TensorFlow logo screenshot from Learning machine learning and tensorflow with Kaggle competition video

Youhan Lee from Korea gave a talk about “Learning machine learning and TensorFlow with Kaggle competition”. He explained how to use the Kaggle platform for learning ML.


Advances in machine learning and deep learning research are changing our technology, and many ML GDEs are interested and contributing.

Learning Neurl Compositional Neural Programs for Continuous Control

Karim Beguir (UK) co-authored a paper with the DeepMind team covering a novel compositional approach using Deep Reinforcement Learning to solve robotics manipulation tasks. The paper was accepted in the NeurIPS workshop.

Finally, Sayak Paul from India, together with Pin-Yu Chen, published a research paper, “Vision Transformers are Robust Learners,” covering the robustness of the Vision Transformer (ViT) against common corruptions and perturbations, distribution shifts, and natural adversarial examples.

If you want to know more about the Google Experts community and their global open-source ML contributions, please check the GDE Program website, visit the GDE Directory and connect with GDEs on Twitter and LinkedIn. You can also meet them virtually on the ML GDE’s YouTube Channel!

Partnering with the NSF on a research institute for AI to improve elderly care

From the early days of the internet to the development of the Human Genome Project, U.S. government-funded R&D has yielded remarkable progress for society, and today it is an important engine for AI research. That’s why, last year, we were proud to announce our partnership with the U.S. National Science Foundation (NSF) to provide $5M to support the establishment of national research institutes working in the area of Human-AI Interaction and Collaboration (HAIC). This partnership—which is part of a more than $300M NSF investment in AI Research Institutes—will create vibrant research centers across the U.S. to advance how people and AI collaborate through speech, text, gestures, and more. It also builds on our partnership with the NSF on next generation networks, and our AI research collaborations with U.S. federal agencies on weather modeling, robust AI systems, whale population monitoring, and more. 

Today, we are delighted to share that NSF has selected the AI Institute for Collaborative Assistance and Responsive Interaction for Networked Groups (AI-CARING) led by Georgia Tech, along with Carnegie Mellon University, Oregon State University, and University of Massachusetts Lowell to receive the $20M AI Institute for HAIC grant. AI-CARING will improve collaboration and communication in elderly caregiving environments by developing AI systems that adjust to the evolving personal needs and behaviors of those requiring care. With our growing research presence in Atlanta, we’re excited to build on our rich history of collaboration with Georgia Tech and its partners in this effort—most recently supporting some of these universities' work to help vulnerable populations find important information on COVID-19 and monitoring and forecasting disease spread.

With a growing population of older adults in need of caregiving, AI systems can be useful in a variety of contexts, like conversational assistants, health sensing, and improving coordination across the care network. For example, AI can help existing voice assistants better understand people with speech impairments, and can be integrated in home bathrooms to make them more accessible. The AI-CARING Institute will develop assistive AI agents across these types of contexts to help those requiring caregiving to sustain their independence and  improve their quality of life. Additionally, this research will be the product of interdisciplinary teams—with expertise across AI, geriatrics, behavioral sciences, and design—working to ensure that AI is deployed responsibly in this context, with human-centered principles in mind.

Congratulations to the recipient universities of the AI Institute awards and the faculty, listed below. We look forward to learning from the team’s research, sharing our resources and expertise, and building a collaboration to help older adults lead more independent lives and improve the quality of their care.

Recipient university institutions:

  • Georgia Institute of Technology
  • Carnegie Mellon University
  • Oregon State University
  • University of Massachusetts Lowell


  • Sonia Chernova (Georgia Tech) - PI
  • Elizabeth Mynatt (Georgia Tech) - Co-PI
  • Reid Simmons (Carnegie Mellon University) - Co-PI
  • Kagan Tumer (Oregon State University) - Co-PI
  • Holly Yanco (University of Massachusetts Lowell) - Co-PI

Using AI to map Africa’s buildings

Between 2020 and 2050, Africa’s population is expected to double, adding 950 million more people to its urban areas alone. However, according to 2018 figures, a scarcity of affordable housing in many African cities has forced over half of the city dwellers in Sub-Saharan Africa to live in informal settlements. And in rural areas, many also occupy makeshift structures due to widespread poverty.

These shelters have remained largely undetectable using traditional monitoring tools. Machine learning, computer vision and remote sensing have come some way in recognizing buildings and roads, but when it comes to denser neighborhoods, it becomes much harder to distinguish small and makeshift buildings. 

Why is this an issue? Because when preparing a humanitarian response, forecasting transportation needs, or planning basic services, being able to accurately map the built environment - which allows us to ascertain population density - is absolutely key.

Enter Google’s Open Buildings

Google’s Open Buildings is a new open access dataset containing the locations and geometry of buildings across most of Africa. From Lagos’ Makoko settlement to Dadaab’s refugee camps, millions of previously invisible buildings have popped up in our dataset. This improved building data helps refine the understanding of where people and communities live, providing actionable information for state and non-state actors looking to provide services from sanitation to education and vaccination.

Open Buildings uses AI to provide a digital footprint of buildings. This includes producing polygons with the outlines of at least 500 million buildings across the African continent, the majority of which are less than 20 square meters. The full dataset encompasses 50 countries.

The data provides the exact location and polygon outline of each building, its size, a confidence score for it being detected as a valid building and a Plus Code. There is, however, no information about the type of building, its street address, or any identifying data. We have also excluded sensitive areas such as conflict zones to protect vulnerable populations.

Satellite mapping using AI 

The Open Buildings dataset was generated by using a model trained to detect buildings using satellite imagery from the African continent. The information for the buildings detected is then saved in CSV files which are available to download. The technical details of the Open Buildings dataset, including usage and tutorials, are available on the dataset website and the Google AI blog.

Animation showing landscape in Africa being mapped

How will this improve planning?

There are many important ways in which this data can be used, including — but not limited to — the following:

Population mapping: Building footprints are a key ingredient for estimating population density. This information is vital to planning for services for communities. 

Humanitarian response: To plan the response to a flood, drought, or other natural disaster.

Environmental science: Knowledge of settlement density is useful for understanding the human impact on the natural environment. 

Addressing systems: In many areas, buildings do not have formal addresses. This can make it difficult for people to access social benefits and economic opportunities. Building footprint data can help with the rollout of digital addressing systems such asPlus Codes.

Vaccination planning: Knowing the density of population and settlements helps to anticipate demand for vaccines and the best locations for facilities. This data is also useful for precision epidemiology, as well as prevention efforts such as mosquito net distribution.

Statistical indicators: Buildings data can be used to help calculate statistical indicators for national planning, such as the numbers of houses in the catchment areas of schools and health centers, mean travel distances to the nearest hospital or demand forecast for transportation systems.

Google’s AI Center in Accra

This project was led by our team at the AI Research Center in Accra, Ghana. The center was launched in 2019 to bring together top machine learning researchers and engineers dedicated to AI research and its applications. The research team has already been improving Google Maps with AI, adding 120 million buildings and 228,000 km of roads across Africa to Maps in the last year. This work is part of our broader AI for Social Good efforts.

Our quantum processor at the Deutsches Museum

In 2019, our Quantum AI team achieved a beyond-classical computation by outperforming the world’s fastest classical computer. Today, a quantum processor from the Sycamore generation that accomplished this important computing milestone will be donated to the Deutsches Museum of Masterpieces of Science and Technology in Munich, Germany. 

The Deutsches Museum has one of the largest collections of science and technology artifacts in the world. This means that the Sycamore will share the same exhibition space as some of the world’s most important technological achievements: like the roundest object in the world – a silicon sphere that gives the kilogram a new definition; the Z3, one of the earliest computers; the Wright Flyer, considered the first serial motor plane; and automotive history from the first diesel engine to the Waymo Firefly. The museum has a long history of preserving artifacts that mark the start of new eras in science and technology, which is why we’re honored to have the Sycamore processor among these exhibits. The beyond-classical experiment ushered in a new era for exploring near-term quantum algorithms that could have tangible benefits to society, for example—design more efficient batteries, create fertilizer using less energy, and figure out what molecules might make effective medicines.
A picture of the Sycamore processor being handed over by the Google team and Deutsches Museum team involved in the project, in front of Zuse's Z3 computer

Handover of the Sycamore processor in front of Zuse‘s Z3 computer. Luise Allendorf-Hoefer,  Curator electronics, Deutsches Museum, Wolfgang M. Heckl, Director General, Deutsches Museum, Markus Hoffmann, Google Quantum AI Partnerships and Hartmut Neven, Director, Google Quantum AI.

This also marks an important milestone in the collaboration between Google Research and Germany’s burgeoning quantum community.  Since Google has a research presence in Munich and Berlin, it has given us the opportunity to partner with several German organisations to explore the future of quantum computing. For example, the Sycamore processor has already been used by some of our industrial research partners, like Volkswagen and Mercedes-Benz, and will be the foundation for experiments designed with Boehringer Ingelheim, Covestro and BASF. 

If you can’t travel to Munich to visit the Deutsches Museum in person, don’t forget that you can take a virtual trip through Google Arts & Culture.

Advances in TF-Ranking

In December 2018, we introduced TF-Ranking, an open-source TensorFlow-based library for developing scalable neural learning-to-rank (LTR) models, which are useful in settings where users expect to receive an ordered list of items in response to their query. LTR models — unlike standard classification models that classify one item at a time — receive an entire list of items as an input, and learn an ordering that maximizes the utility of the entire list. While search and recommendation systems are the most common applications of LTR models, since its release, we have seen TF-Ranking being applied in diverse domains beyond search, including e-commerce, SAT solvers, and smart city planning.

The goal of learning-to-rank (LTR) is to learn a function f() that takes as an input a list of items (documents, products, movies, etc.) and outputs the list of items in the optimal order (descending order of relevance). Here, green shade indicates item relevance level, and the red item marked with 'x' is non-relevant.

In May 2021, we published a major release of TF-Ranking that enables full support for natively building LTR models using Keras, a high-level API of TensorFlow 2. Our native Keras ranking model has a brand-new workflow design, including a flexible ModelBuilder, a DatasetBuilder to set up training data, and a Pipeline to train the model with the provided dataset. These components make building a customized LTR model easier than ever, and facilitate rapid exploration of new model structures for production and research. If RaggedTensors are your tool of choice, TF-Ranking is now working with them as well. In addition, our most recent release, which incorporates the Orbit training library, contains a long list of advances — the culmination of two and half years of neural LTR research. Below we share a few of the key improvements available in the latest TF-Ranking version.

Workflow to build and train a native Keras ranking model. Blue modules are provided by TF-Ranking, and green modules are customizable.

Learning-to-Rank with TFR-BERT
Recently, pretrained language models like BERT have achieved state-of-the-art performance on various language understanding tasks. To capture the expressiveness of these models, TF-Ranking implements a novel TFR-BERT architecture that couples BERT with the power of LTR to optimize the ordering of list inputs. As an example, consider a query and a list of n documents that one might like to rank in response to this query. Instead of learning an independent BERT representation for each <query, document> pair, LTR models apply a ranking loss to jointly learn a BERT representation that maximizes the utility of the entire ranked list with respect to the ground-truth labels.

The figure below illustrates this process. First, we flatten a list of n documents to rank in response to a query into a list <query, document> tuples. These tuples are fed into a pre-trained language model (e.g., BERT). The pooled BERT outputs for the entire document list are then jointly fine-tuned with one of the specialized ranking losses available in TF-Ranking. Our experience shows that this TFR-BERT architecture delivers significant improvements in pretrained language model performance, leading to state-of-the-art performance for several popular ranking tasks, especially when multiple pretrained language models are ensembled. Our users can now get started with TFR-BERT using this simple example.

An illustration of the TFR-BERT architecture, in which a joint LTR model over a list of n documents is constructed using BERT representations of individual <query, document> pairs.

Interpretable Learning-to-Rank
Transparency and interpretability are important factors in deploying LTR models in ranking systems that can be involved in determining the outcomes of processes such as loan eligibility assessment, advertisement targeting, or guiding medical treatment decisions. In such cases, the contribution of each individual feature to the final ranking should be examinable and understandable to ensure transparency, accountability and fairness of the outcomes.

One possible way to achieve this is using generalized additive models (GAMs) — intrinsically interpretable machine learning models that are linearly composed of smooth functions of individual features. However, while GAMs have been extensively studied on regression and classification tasks, it is less clear how to apply them in a ranking setting. For instance, while GAMs can be straightforwardly applied to model each individual item in the list, modeling both item interactions and the context in which these items are ranked is a more challenging research problem. To this end, we have developed a neural ranking GAM — an extension of generalized additive models to ranking problems.

Unlike standard GAMs, a neural ranking GAM can take into account both the features of the ranked items and the context features (e.g., query or user profile) to derive an interpretable, compact model. This ensures that not only the contribution of each item-level feature is interpretable, but also the contribution of the context features. For example, in the figure below, using a neural ranking GAM makes visible how distance, price, and relevance, in the context of a given user device, contribute to the final ranking of the hotel. Neural ranking GAMs are now available as a part of TF-Ranking,

An example of applying neural ranking GAM for local search. For each input feature (e.g., price, distance), a sub-model produces a sub-score that can be examined, providing transparency. Context features (e.g., user device type) can be utilized to derive importance weights of submodels.

Neural Ranking or Gradient Boosting?
While neural models have achieved state of the art performance in multiple domains, specialized gradient boosted decision trees (GBDTs) like LambdaMART remained the baseline to beat in a variety of open LTR datasets. The success of GBDTs in open datasets is due to several reasons. First, due to their relatively small size, neural models are prone to overfitting on these datasets. Second, since GBDTs partition their input feature space using decision trees, they are naturally more resilient to variations in numerical scales in ranking data, which often contain features with Zipfian or otherwise skewed distributions. However, GBDTs do have their limitations in more realistic ranking scenarios, which often combine both textual and numerical features. For instance, GBDTs cannot be directly applied to large discrete feature spaces, such as raw document text. They are also, in general, less scalable than neural ranking models.

Therefore, since the TF-Ranking release, our team has significantly deepened the understanding of how best to leverage neural models in ranking with numerical features. This culminated in a Data Augmented Self-Attentive Latent Cross (DASALC) model, described in an ICLR 2021 paper, which is the first to establish parity, and in some cases statistically significant improvements, of neural ranking models over strong LambdaMART baselines on open LTR datasets. This achievement is made possible through a combination of techniques, which include data augmentation, neural feature transformation, self-attention for modeling document interactions, listwise ranking loss, and model ensembling similar to boosting in GBDTs. The architecture of the DASALC model was entirely implemented using the TF-Ranking library.

All in all, we believe that the new Keras-based TF-Ranking version will make it easier to conduct neural LTR research and deploy production-grade ranking systems. We encourage everyone to try out the latest version and follow this introductory example for a hands-on experience. While we are very excited about this new release, our research and development journey is far from over, so we will continue to advance our understanding of learning-to-rank problems and share these advances with our users.

This project was only possible thanks to the current and past members of the TF-Ranking team: Honglei Zhuang, ‎Le Yan, Rama Pasumarthi, Rolf Jagerman, Zhen Qin, Shuguang Han, Sebastian Bruch, Nathan Cordeiro, Marc Najork and Patrick McGregor. We also extend special thanks to our collaborators from the Tensorflow team: Zhenyu Tan, Goldie Gadde, Rick Chao, Yuefeng Zhou‎, Hongkun Yu, and Jing Li.

Source: Google AI Blog

Launching the AI Academy for small newsrooms

As people searched for the latest information on COVID-19 last year, including school reopenings and travel restrictions, the BBC recognized they needed to find new ways of bringing their journalism to their audiences. They released a new online tool, the BBC Corona Bot, which uses artificial intelligence to draw on BBC News’ explanatory journalism. It responds with an answer to a reader’s specific question where possible, or points to health authorities’ websites when appropriate. AI technology allowed BBC News to reach new audiences and drive more traffic to their stories and explainers. 

This is one example of how AI can help newsrooms. AI can help build new audiences and automate tasks, freeing up time for journalists to work on the more creative aspects of news production and leaving tedious and repetitive tasks to machines. However, newsrooms around the world have told researchers they worry that access to AI technology is unequal. They fear big publishers likely will benefit most from artificial intelligence, while smaller news organizations could get left behind. 

To help bridge this gap, the Google News Initiative is partnering with Polis, the London School of Economics and Political Science’s journalism think tank, to launch a training academy for 20 media professionals to learn how AI can be used to support their journalism. 

The AI Academy for Small Newsrooms is a six-week long, free online program taught by industry-leading journalists and researchers who work at the intersection of journalism and AI. It will start in September 2021 and will welcome journalists and developers from small news organizations in the Europe, Middle East, and Africa (EMEA) region.

By the end of the course, participants will have a practical understanding of the challenges and opportunities of AI technologies. They will learn examples of how to use AI to automate repetitive tasks, such as interview transcription or image search, as well as how to optimize newsroom processes by getting insights on what content is most engaging.

For example, other newsrooms using AI technology in the region include Schibsted, a Nordic news outlet that developed an innovative model to reduce gender bias in news coverage, while in Spain, El Pais uses an AI-based tool to moderate toxic comments.

Most importantly, participants will create action plans to guide the development of AI projects within their news organizations. JournalismAI will share these plans openly to help other publishers around the world.

This pilot program — which we plan to launch in other regions in 2022 — is part of a broader training effort over the last three years by JournalismAI, a partnership between the GNI and Polis to forester AI literacy in newsrooms globally. More than 110,000 participants have already taken the online training modules available on the Google News Initiative Training Center.

This year, JournalismAI will also create an AI Journalism Starter Pack to make the information about AI in journalism more accessible to small and local publishers. It will include examples of AI tools that can solve small and local publishers' basic needs such as tagging or transcribing.

Find more detailed information on the AI Academy for Small Newsrooms and how to apply on the JournalismAI website. The deadline for applications is 11:59 PM GMT on August 1, 2021.

Kick like a pro with Footy Skills Lab

When I was growing up in Brisbane, Aussie Rules football wasn’t offered as a school sport – and there weren’t any professional female role models to look up to and learn from. Despite these limitations, we got resourceful. We organized football games in our lunch breaks with friends, using soccer or rugby goal posts and adding sticks or cones to serve as point posts. We practised accuracy using rubbish bins as targets.

A decade later, women have truly made their mark in the AFL. There are, however,  many barriers still facing aspiring footy players — including access, cost, mobility and, more recently, lockdown restrictions. We still have to be resourceful to keep active and hone our skills. 

Three years ago, the AFL and Google first teamed up to help footy fans better connect with the games and players they love. Since then, we’ve been thinking about ways we could improve access to Aussie Rules coaching and community participation – regardless of ability, gender, location, culture or socio-economic background. 

A graphic showing a phone with the Footy Skills Lab app open, in front of a Sherrin football.

Today, we’re thrilled to launch Footy Skills Lab to help budding footy players in Australia and all around the world sharpen their AFL skills – straight from their smartphone. 

Footy Skills Lab is a free platform, powered by GoogleAI, which helps you improve your skills through activities in ball-handling, decision-making and kicking across three levels of difficulty.

A screenshot showing the AFL activities available on the app, including ball handling, decision making and kicking.

To give Footy Skills Lab a whirl, all you need is a smartphone with an internet connection, a football, something to prop your phone up (like a water bottle) and space to move. 

You’ll get tips on kicking from me, and tips on ball-handling and decision-making from athletes across the AFLW and AFL Wheelchair competitions – including Carlton’s Madison Prespakis and Richmond’s Akec Makur Chuot. Through audio prompts and closed captioning, these tips and activities are also accessible for people with visual and hearing needs. And when you finish the activity, you’ll get a scorecard that you can share with your friends, family, teammates and coaches. 

Screenshots showing still images of AFL athletes Madison Prespakis and Akec Makur Chuot providing football training tips.

Whether you’re in Manchester, Miami or in lockdown in Melbourne, Footy Skills lab is such an easy, convenient way to get motivated and access coaching from pros.  So go on, join in the fun and give us your best kick!