Home Blog Page 298

Inside Intel’s Loihi 2 Neuromorphic Chip: The Upgrades and Promises

October 2, 2021

Intel Loihi 2, Loihi, neuromorphic chip ai , Intel lava — Image Credit: Analytics Drift Team

The second-generation “Loihi” processor from Intel has been made available to advance research into neuromorphic computing approaches that more closely mimic the behavior of biological cognitive processes. Loihi 2 outperforms the previous chip version in terms of density, energy efficiency, and other factors. This is part of an effort to create semiconductors that are more like a biological brain, which might lead to significant improvements in computer performance and efficiency.

Intel Announces Loihi 2, Lava Software Framework For Advancing Neuromorphic Computing - Phoronix

The first generation of artificial intelligence was built on the foundation of defining rules and emulating classical logic to arrive at rational conclusions within a narrowly defined problem domain. It was ideal for monitoring and optimizing operations. The second generation is dominated by the use of deep learning networks to examine the contents and data that were mostly concerned with sensing and perception. The third generation of AI focuses on drawing similarities to human cognitive processes, like interpretation and autonomous adaptation.

This is achieved by simulating neurons firing in the same way as humans’ nervous systems do, a method known as neuromorphic computing.

Neuromorphic computing is not a new concept. It was initially suggested in the 1980s by Carver Mead, who coined the phrase “neuromorphic engineering.” Carver had spent more than four decades building analytic systems that simulated human senses and processing mechanisms including sensation, seeing, hearing, and thinking. Neuromorphic computing is a subset of neuromorphic engineering that focuses on the human-like systems’ “thinking” and “processing” capabilities. Today, neuromorphic computing is gaining traction as the next milestone in artificial intelligence technology.

Intel Rolls Out New Loihi 2 Neuromorphic Chip: Built on Early Intel 4 Process

In 2017, Intel released the first-generation Loihi chip, a 14-nanometer chipset with a 60-millimeter die size. It has more than 2 billion transistors and three orchestration Lakemont cores. It also features 128 core packs and a configurable microcode engine for asynchronous spiking neural network-on-chip training. The benefit of having spiking neural networks enabled Loihi to be entirely asynchronous and event-driven, rather than being active and updating on a synchronized clock signal. When a charge builds up in the neurons, “spikes” are sent along active synapses. These spikes are mostly time-based, with time being recorded as part of the data. The core fires out its own spikes to its linked neurons when spikes accumulate in a neuron for a particular amount of time and reach a certain threshold.

Even though Loihi 2 has 128 neuromorphic cores, each core now has 8 times the number of neurons and synapses. Each of the 128 cores has 192 KB of flexible memory, compared to the prior limit of 24. Each neuron may now be assigned up to 4096 states depending on the model, compared to the previous limit of 24. The Neuron model can now be entirely programmable, similar to an FPGA, which gives it more versatility – allowing for new sorts of neuromorphic applications.

One of the drawbacks of Loihi was that spike signals were not programmable and had no context or range of values. Loihi 2 addresses all of these issues while also providing 2-10x (2X for neuron state updates, up to 10X for spike generation) faster circuits, eight times more neurons, and four times more link bandwidth for increased scalability.

Loihi 2 was created using the Intel 4 pre-production process and benefited from the usage of EUV technology in that node. The Intel 4 process allowed to halve the size of the chip from 60 mm2 to 31 mm2, with the number of transistors rising to 2.3 billion. In comparison to previous process technologies, the use of extreme ultraviolet (EUV) lithography in Intel 4 has simplified the layout design guidelines. This has allowed Loihi 2 to be developed quickly.

Support for three-factor learning rules has been added to the Loihi 2 architecture, as well as improved synaptic (internal interconnections) compression for quicker internal data transmission. Loihi 2 also features parallel off-chip connections (that enable the same types of compression as internal synapses) that may be utilized to extend an on-chip mesh network across many physical chips to create a very powerful neuromorphic computer system. Loihi 2 also features new approaches for continual and associative learning. Furthermore, the chip features 10GbE, GPIO, and SPI interfaces to make it easier to integrate Loihi 2 with traditional systems.

Loihi 2 further improves flexibility by integrating faster, standardized I/O interfaces that support Ethernet connections, vision sensors, and bigger mesh networks. These improvements are intended to improve the chip’s compatibility with robots and sensors, which have long been a part of Loihi’s use cases.

Another significant change is in the portion of the processor that assesses the condition of the neuron before deciding whether or not to transmit a spike. Earlier, users had to make such conclusions using a simple bit of arithmetic in the original processor. Now, they only need to conduct comparisons and regulate the flow of instructions in Loihi 2 thanks to a simpler programmable pipeline.

Intel claims Loihi 2’s enhanced architecture allows it to be compatible in carrying back-propagation processes, which is a key component of many AI models. This may help in accelerating the commercialization of neuromorphic chips. Loihi 2 has also been proven to execute inference calculations, with up to 60 times fewer operations per inference compared to Loihi – without any loss in accuracy. Often inference calculations are used by AI models to interpret given data.

The Neuromorphic Research Cloud is presently offering two Loihi 2-based neuromorphic devices to researchers. These are:

Oheo Gulch is a single-chip add-in card that comes with an Intel Arria 10 FPGA for interfacing with Loihi 2 which will be used for early assessment.
Kapoho Point, an 8-chip system board that mounts eight Loihi 2 chips in a 4×4-inch form factor, will be available shortly. It will have GPIO pins along with “standard synchronous and asynchronous interfaces” that will allow it to be used with things like sensors and actuators for embedded robotics applications

These will be available via a cloud service to members of the Intel Neuromorphic Research Community (INRC) and Lava via GitHub for free.

Intel has also created Lava to address the requirement for software convergence, benchmarking, and cross-platform collaboration in the realm of neuromorphic computing. As an open, modular, and extendable framework, it will enable academics and application developers to build on one other’s efforts and eventually converge on a common set of tools, techniques, and libraries.

Lava operates on a range of conventional and neuromorphic processor architectures, allowing for cross-platform execution and compatibility with a variety of artificial intelligence, neuromorphic, and robotics frameworks. Users can get the Lava Software Framework for free on GitHub.

Edy Liongosari, chief research scientist and managing director for Accenture Labs believes that advances like the new Loihi-2 chip and the Lava API will be crucial to the future of neuromorphic computing. “Next-generation neuromorphic architecture will be crucial for Accenture Labs’ research on brain-inspired computer vision algorithms for intelligent edge computing that could power future extended-reality headsets or intelligent mobile robots,” says Edy.

For now, Loihi 2 has piqued the interest of the Queensland University of Technology. The institute is looking to work on more sophisticated neural modules to aid in the implementation of biologically inspired navigation and map formation algorithms. The first generation Loihi is already being used at Los Alamos National Lab to study tradeoffs between quantum and neuromorphic computing. It is also being used in the backpropagation algorithm, which is used to train neural networks.

Facebook Launches Dynatask to Boost better Usability of Dynabench and customize NLP tasks

Preetipadma K

October 2, 2021

Facebook AI Research Dynabench Dynadesk NLP — Image Caption: TechBeacon

Last September, Facebook had unveiled Dynabench, an AI data collection and benchmarking tool that creates complex test datasets by putting humans and models “in the loop.” They are now unveiling ‘Dynatask,’ a new feature that unlocks Dynabench’s full potential for the AI community.

By allowing human annotators to engage organically with NLP models, Dynatask assists researchers in identifying flaws in the models. Dynatask has created a new artificial intelligence model benchmarking system that is more accurate and impartial than previous techniques. Researchers will be able to take advantage of the Dynatask platform’s powerful features and compare models on the dynamic leaderboard. This includes measuring techniques for fairness, robustness, compute, and memory, in addition to accuracy.

Dynabench analyses how easily humans can deceive AI using dynamic adversarial data collection technique, which Facebook says is a stronger determinant of a model’s quality than current benchmarks. Last year, academics from Facebook and the University College London showed evidence that 60 to 70 percent of responses produced by models evaluated on open-domain benchmarks are embedded somewhere in the training sets.

Meanwhile, in the past few years, the AI community has taken an interest in open-domain question-answering for its practical uses, and more recently, to assess language models’ understanding of factual information. However, a thorough grasp of the types of questions that models can answer remains unresolved. Further, unknown factors regarding the distribution of questions and responses in benchmark corpora make it challenging to interpret the results.

Fortunately, these benchmarks have been quickly saturating in recent years in NLP. Looking through the annals, it took the research community 18 years to attain human-level performance on MNIST and roughly six years to outperform humans on ImageNet. Meanwhile, beating humans on the GLUE language understanding benchmark took almost a year.

Despite these impressive feats, we are still a long way from having robots that can thoroughly comprehend the essence of natural language.

When Dynabench was first released, it included four tasks: natural language inference, question answering, sentiment analysis, and hate speech identification. With its recent breakthroughs, the Facebook AI research team has powered the multilingual translation challenge at Workshop on Machine Translations. These attempts to acquire dynamic data resulted in eight articles being published and over 400K raw samples being collected.

Dynabench provided new complex datasets that combine people and models to test NLP models properly. This approach identifies gaps in current models, allowing the next generation of AI models to be trained in the loop. It also assesses how readily people may deceive AI algorithms in a dynamic rather than static setting. Dynatask takes these features a step further.

The Facebook Dynatask platform’s potential is limitless, thanks to its highly flexible and configurable feature platform. It opens up a whole new universe for task designers. They may simply create annotation interfaces to allow interactions with models hosted on any number or machine learning competition such as Dynabench and can set up their own challenges with no coding skills. This will enable researchers to acquire dynamic adversarial data.

Each task in Dynatask will have one or more owners who will define the task parameters. They will be able to choose the required assessment indicators/metrics from a list of options, including accuracy, robustness, fairness, compute, and memory. According to the Facebook newsroom blog, anyone can upload models to the task’s evaluation cloud, which is set up to compute scores and other metrics on specified data sets. They may be moved into the loop for dynamic data gathering and human-in-the-loop evaluation after uploading the model, calculation, and evaluation. Task owners may also gather data via the dynabench.org web interface or using annotators (such as Mechanical Turk).

Let’s say you wanted to start a Natural Language Inference task but there weren’t any yet.

Step 1: Log into your Dynabench account and go to your profile page to fill out the “Request new task” form.

Step 2: Once approved, you will have a dedicated task page and corresponding admin dashboard that you control, as the task owner.

Step 3: After uploading the model, select the existing datasets you want to use to evaluate models from the dashboard, as well as the metrics you want to use.

Step 4: Finally, propose or request baseline models from the community.

Step 5: Upload fresh contexts to the system and begin collecting data using the task owner interface to gather a new round of dynamic adversarial data, in which annotators are instructed to produce instances that deceive the model.

Step 6: Once you’ve collected enough data and established that training on the data improves the system, you may upload improved models and then put them in the data collecting loop to create even stronger ones.

The same fundamental process was used by the Facebook AI Research team to create various dynamic data sets, such as Adversarial Natural Language Inference. With the tools available to the larger AI community, the team believes that anybody can build data sets with people and models in the loop.

DeepMind’s AI predicts When and Where it’s going to Rain

Dipayan Mitra

October 2, 2021

Google’s DeepMind and the University of Exeter have partnered with Med Office to develop a deep learning and artificial intelligence-powered predictive tool named ‘nowcasting system’ that can accurately predict where and when it’s going to rain.

Broadly used methodologies generate results that forecast rain only above six hours. But nowcasting can predict rain possibilities in the next ninety minutes. Not every meteorological organization can accurately predict the likelihood of heavy rainfall as it is a complicated task.

The nowcasting system will allow scientists to generate weather forecasts quickly to help many enterprises spread across various sectors to plan their operations effectively. Industries including aviation, water management, agriculture, emergency services, and many others would heavily benefit from this new product of DeepMind.

DeepMind mentioned in a blog, “Today’s weather predictions are driven by powerful numerical weather prediction (NWP) systems. By solving physical equations, NWPs provide essential planet-scale predictions several days ahead.”

It also mentioned that NWP systems struggle to generate high-resolution predictions for short lead times under two hours. DeepMind’s nowcasting tackles this challenge by filling the performance gap. Earlier, many deep learning algorithms have been developed for weather prediction, but they only excelled in predicting the intensity and location of rainfall but not its time.

The head of partnership and product innovation at Med Office, Niall Robinson, said, “One forecast gets precipitation in the right location but at wrong intensity, or another gets the right mix of intensities, but in the wrong place and so on. We went to a lot of effort in this research to assess our algorithm against a wide suite of metrics.”

DeepMind’s research scientist, Suman Ravuri, said that the practical end-user inputs provided by Med Office pushed the model development in a different way than what they intended to do on their own.

Hugging Face releases 900 unique Datasets to standardize NLP

Amit Kulkarni

October 1, 2021

Hugging Face has released the largest hub of ready-to-use datasets for ML models to anchor challenges of NLP. The Dataset contains 900 unique datasets, more than 25 metrics and has more than 300 contributors. This library will support many novel cross-data research projects to standardize end-user interface and provide a lightweight frontend for internet-scale corpora.

When considering NLP use cases, Datasets play a crucial role in evaluating and benchmarking results. While supervised datasets help in fine-tuning models, large unsupervised datasets assist in pertaining and language modelling. A practitioner faces several challenges when dealing with different versions and documentation of Datasets, the primary cause being the lack of uniformity.

To put an end to this, Hugging Face designed a Dataset that not only addresses the associated challenges of Dataset management but also provides access to support community culture. As this Dataset is developed by community contribution, it inherently got the bootstrapping takeaway that consisted of a variety of languages, including — continuous data types, multi-dimensional arrays for images, audios, and videos.

The primary intent of Hugging Face while creating Dataset from public hackathon are:

Each Dataset in the library has a standard tabular format, that facilitates proper version and citation. It needs just one line of code to download all the datasets.
As large datasets are computation and memory efficient, Hugging Face can stream Datasets through the same interface facilitating tokenization and featurization.
All the Datasets are tagged and documented with their usage, types, and construction.

In a short while, the company has managed to get a phenomenal rise as a startup with its transformer library that is backed by TensorFlow and PyTorch. It consists of many pre-trained models to perform text classification, information retrieval, and summarisation tasks. This library is extensively used by researchers at Google, Facebook, and Microsoft and downloaded a billion times.

As high-end technology is concentrated in the hands of a few powerful big companies, Hugging Face wants democratisation of AI to extend the benefits of emerging technologies to smaller organisations. Hugging Face Co-founder and CEO Clement Delangue believes that there is a disconnect between the research and engineering team in NLP, hence it aims to be the GitHub for Machine Learning.

New tool fuses ML and DL features to detect sleep apnea

Amit Kulkarni

October 1, 2021

Individuals who suspect having sleep apnea and doctors who diagnose them could soon have a more effective way to detect the condition at home. Penn State College of Information Sciences and Technology researchers have developed a ConCAD method that can detect sleep apnea by incorporating expert knowledge into deep learning techniques. This research was presented at the European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML-PKDD), virtually held between Sept. 13-17.

Researchers believe the new tool outperforms all existing baseline methods because deep learning technology is brewed with expert knowledge. This tool automatically learns patterns from electrocardiogram (ECG) data collected by at-home devices, making it a faster and more ideal solution than any other existing sleep apnea diagnostics.

“The standard approach to detect sleep apnea involves polysomnography (sleep study) — that involves a patient staying in hospital overnight — under supervision of a clinical practitioner. This process is time-consuming, tedious, and intrusive,” said Guanje Huang, lead author of the paper.

Huang explained that a patient’s data is collected through a sleep study that includes — measuring brain waves, blood oxygen levels, heart rate, respiration, and body movements. Later, clinicians devote their time and various resources to analyze data. “It is very crucial to design an accurate model to automatically analyze data, and help doctors detect sleep apnea swiftly,” said Huang

There also exist other tools that automatically detect sleep apnea through at-home devices using computer models. They are either built through traditional machine learning methods that rely on prior knowledge from human experts or through deep learning methods that eliminate the need for such experts. The former requires hand-crafted features, and the latter consists of immense amounts of data resulting in limitations to these standalone approaches.

“The traditional machine learning method usually requires a small amount of data to learn a robust classifier but requires a careful feature extraction and selection process,” Huang explained.“ Whereas the deep learning models usually achieve better performance but require a large dataset.

Huang’s ConCAD (Contrastive Learning-based Cross Attention for Sleep Apnea Detection) model detects sleep apnea precisely by simultaneously leveraging deep learning features and traditional machine learning’s expert knowledge. This model explicitly requires an expert understanding of RR interval (RRI) and R peak envelope (RPE). Existing methods for detecting sleep apnea involve measuring the intervals between and peak of the R wave, which measures cardiac rhythm in a patient’s ventricular walls, in a standard ECG. Whereas ConCAD utilizes a cross-attention mechanism — a deep learning model that assigns weights to each part based on their importance — to fuse the deep learning features with the expert knowledge features, emphasizing the important features automatically.

The working of ConCAD consists of fours steps:

Pass the original raw ECG data through feature extractors to automatically learn patterns from both expert knowledge and deep learning methods that could indicate sleep apnea.
The patterns or features are automatically and synergistically fused and assigned a weight based on each important part.
The model undergoes a contrastive learning process to match similar features closely.
At last, the data is classified based on final features of ECG and corresponding expert knowledge to calculate the patient’s probability of sleep apnea.

Researchers used two publicly available ECG datasets to test the ConCAD model. These datasets comprise more than 26,000 segments of 30-second and two-and-a-half-minute inputs annotated by experts, identifying apnea or regular sleep events. As compared to six existing state-of-the-art sleep apnea detection methods, ConCAD outperforms all the models. For the first dataset, ConCAD accurately identifies sleep apnea events for 88.75% in one-minute segments, and 91.22% in the five-minute segments, and 82.5% and 83.47% respectively in the second dataset.

Fenglong Ma, assistant professor of information sciences and technology and the principal investigator, said, “If patients use personal ECG devices at home, they may monitor their sleep apnea conditions with the ConCAD model.” “This is a new attempt to assimilate expert knowledge into a deep learning model that will assist doctors in simplifying the diagnostic process of sleep apnea,” added Ma.

US to offer grants to Small Businesses and help them participate in setting Global AI Standards

Dipayan Mitra

October 1, 2021

United States representatives Scott Franklin, Jay Obernolte, and others have introduced a bill for providing government grants to small and medium-sized businesses to help them participate in setting global artificial intelligence (AI) standards.

Government officials believe that while small startups are developing revolutionary innovations and breakthroughs, they don’t have adequate funding and resources to complete the developmental process.

This new bill will ensure that the United States maintains its number one spot as the global leader in artificial intelligence innovators. Earlier, the President of China, XI Jinping, said the Chinese Communist Party (CCP) aims to set a global dominance in the artificial intelligence industry by 2030.

The U.S. government has passed this new bill to counter China’s AI strategies. U.S. representatives have derived the bill from a recommendation in the National Security Commission on Artificial Intelligence’s final report that was given to Congress earlier this year.

U.S. representative Scott Franklin said, “Advances in Artificial Intelligence will play an increasingly defining role in U.S. global competitiveness. That’s why we need to create an environment that allows American small businesses on the cutting edge of technological innovation to participate in creating global standards and regulation in the A.I. field.”

He added that the new bill would promote economic growth by ensuring that artificial intelligence standards adhere to American Business Standards. The bill is a bipartisan solution for maintaining U.S. supremacy in the tech industry and countering Chinese efforts to influence the artificial intelligence sector.

“As a former small business owner during the early days of a burgeoning industry, I know first-hand how critical it is to include the voices of small business owners and engineers in the standard setting process,” said U.S. representative McNerney. He also mentioned that being the co-chair of the House A.I. Caucus, his top priority is to ensure that small businesses have a seat at the table.

IIT Roorkee to start School for Data Science and Artificial Intelligence

Dipayan Mitra

October 1, 2021

IIT Roorkee School Data Science Artificial intelligence

Indian Institute of Technology (IIT) Roorkee is set to start a new school for data science and artificial intelligence in collaboration with the Mehta Family Foundation, USA, named Mehta Family School of Data Science and Artificial Intelligence.

The school has been established with the purpose of training talents in the field of data analytics and artificial intelligence to meet the increasing demand for such job roles in the global market. It also aims to support government initiatives by seeding startups and entrepreneurship.

The first batch will begin from September 2022 and will offer undergraduate, postgraduate, and doctoral programs. IIT Roorkee plans to invite numerous experts to design the curriculum and facilities of the Mehta Family School of Data Science and Artificial Intelligence.

The institute will also provide the best available mentorship to students to let them conduct innovative research. Director of IIT Roorkee, Ajit K Chaturvedi, and CEO of Mehta Family Foundation, Rahul Mehta, signed a contract on 27th September 2021 for this collaborative effort. The school will also facilitate student scholarship and faculty exchange programs.

“AI-driven technologies are rapidly transforming our world. Academic collaborations between international faculty and institutes can produce individuals poised to address such ongoing global challenges as climate change, resource sustainability, and security,” said Rahul Mehta.

Ajit K Chaturvedi said, “I am sure they will nurture the growth of the Mehta Family School of Data Science and Artificial Intelligence at IIT Roorkee.” He also mentioned that a large number of faculty members had contributed their efforts to conceiving the Mehta Family School of Data Science and Artificial Intelligence.

Leena AI raises US$30 million in Series B Funding Round

Dipayan Mitra

October 1, 2021

Human resource technology developing company Leena AI raised US$30 million in its series B funding round led by Bessemer Venture Partners. Other investors like B Capital Group and Greycroft also participated in the funding round.

The company plans to use the fresh funds to accelerate its product development process in order to keep up with the global demand of its platform. Leena AI will also aim to increase its market share and add new products to its suite dedicated to sales, finance, and IT teams.

The company’s product comes with an HR service delivery suite, an employee engagement module, and a Covid Response suite that helps businesses to provide a better employee experience.

Partner at Greycroft, Mark Terbeek, said, “The Leena AI team has been able to bring HR into the modern era through automation and machine learning, enabling better employee engagement.” He further mentioned that human resource is a field that is in desperate need of disruption as most of the Hr operations are slow and done manually.

New York-based Software-as-a-Service (SaaS) startup Leena AI was founded by Adit Jain, Anand Prajapati, and Mayank Goyal in 2018. The firm specializes in developing autonomous conversational AI-backed enterprise solutions to enable companies to provide a better experience to their employees.

Last year, in November, Leena AI raised US$8 million in its series A funding round. This fresh funding brings the company’s total amount raised till date to US$40 million. Leena AI’s platform is used by customers spread across sixty countries, and it is available in over 40 languages. Many industry-leading companies, including Coca-Cola, Nestle, Puma, P&G, Abbott, Al Jazeera, HDFC Bank, Reserve Bank of India, and many others, use Leena AI’s service.

“Leena AI Employee Experience Suite deeply understands enterprise HR support tickets to solve this very difficult problem at the world’s top enterprises,” said the co-founder and CEO of Leena AI, Adit Jain.

Haptik launches AI Agent Assist to provide better Customer Experience

Dipayan Mitra

October 1, 2021

One of the world’s largest conversational AI developing firms and a subsidiary of Reliance Jio, Haptik launches its new artificial intelligence-powered Agent Assist for improving agent performance and providing an enhanced customer experience.

The AI tool provides critical information to agents that help them solve customer issues seamlessly. AI Agent Assist enables customer service executives to quickly reply to queries and provide a personalized experience to their customers.

The platform eliminates the need for agents to access multiple channels to retrieve information by displaying all the necessary data in a single screen; This drastically reduces the time consumed in query resolutions that enable agents to increase their productivity.

The Director of Product Management at Haptik, Vikram Kamath, said, “AI Agent Assist brings the best of both worlds by offering recommendations to agents which aren’t user-facing, unless the agent approves, making AI Agent Assist the most low-risk path to leverage AI and make contact centers more efficient and productive.”

He further mentioned that artificial intelligence is becoming a key differentiator that companies are looking forward to investing in. The tool provides real-time text suggestions with sentence-level completion for agents to help them tackle customer queries tactfully and quickly.

AI Agent Assist comes with an easy-to-use interface that reduces the training time of new agents that allows companies to focus more on operations. Haptik also provides numerous out-of-the-box integration options with leading CRM platforms like Salesforce, Freshworks, and Zendesk.

Mumbai-based conversational AI company Haptik was founded by Aakrit Vaish and Swapan Rajdev in the year 2013. To date, the firm has raised total funding of $12.2 million over two rounds. Haptik has a massive customer base that includes industry-leading companies like Oyo, Coca Cola, Tata Group, KFC, Club Mahindra, Zurich Insurance, HP, Ola, and many more.

“By choosing AI Agent Assist, businesses can create ‘Powerful Agents’ equipped with more knowledge and a higher capacity to handle multiple conversations, with the same accuracy, speed, and precision,” said Kamath.

Google is redesigning Search using AI technology

Amit Kulkarni

October 1, 2021

Google has recently announced that it will be applying AI advancements, including Multitask Unified Model (MUM) to improve Google Search. The company demonstrated new features at the Search On event to leverage MUM for web searches by providing relevant content while enhancing web searches to become more natural and intuitive.

One of the new features introduced is “Things to know,” which will focus on making web searches more convenient to understand for people. As this feature understands the behaviour of how people explore topics, it will show search results that align with the topic people are most likely to look at first.

Google explained this use case by searching “acrylic painting,” which may suggest “Things to know” like — how to get started with painting, different styles of acrylic painting, and many more. For the mentioned example, Google is able to identify over 350 different topics that relate to acrylic painting.

This feature will be rolled out in the coming months that will unlock new ways to — refine search — understanding deeper insights — connect visually rich pages of search results. As images turn people’s visual inspiration into action, these pages would be competing with Pinterest.

To extend further, Google is also upgrading video search. Already, the company identifies key moments inside a video using AI, it now plans to launch features that will identify topics in a video. This feature will provide relevant links that allow users to dig deeper for topics of interest even if the topic is not explicitly mentioned in the video. MUM will understand what the video is about and make suggestions when a related search is made.

In an example, Google showed how a Macaroni penguin’s video could be used to point to a range of related videos for those searching about how Macaroni penguins find their family. As the majority of GenZ learn from YouTube, an initial version of this feature will be introduced on YouTube Search, which will have more visual enhancements in coming months, says Google.

1...297298299...358 Page 298 of 358