London-based AI research firm DeepMind makes a profit for the first time in 2020 of £43.8 million after recording a loss of hundreds of millions of dollars over the past few years. After Google acquired the UK artificial intelligence lab, DeepMind, in 2014 for 400 million pounds, this is the first time that DeepMind has recorded a profit. The company specializes in breaking new ground in AI and machine learning research that Google and its parent company can then commercialize later. DeepMind reported a loss of $649 million in 2019, and Alphabet wrote off 1.1 billion debt in the same year.
The profit comes after DeepMind had increased sales from £560 million last year to a record hit of £826 million this year. According to the Companies House annual results filing, DeepMind had tripled its turnover from just £265.5 million in 2019 to £826.2 million in 2020. Elsewhere, staff costs and other related costs rose modestly from £467 million to £473 million, suggesting that DeepMind’s hiring frenzy may have ended.
DeepMind generates most of its revenue from research carried out for companies under the Google umbrella. DeepMind’s services and solutions aren’t directly sold to customers. Instead, DeepMind can only sell them to companies under the Alphabet Umbrella.
DeepMind, which Google parent Alphabet owns, did not announce deals with private companies outside Alphabet and neither provided any reason for the sudden revenue jump. It does, however, sell its technology, software, and services to Alphabet’s companies, including YouTube, Google, and X, which is the moonshot division.
A DeepMind spokesperson told CNBC the company is “powering products and infrastructure that enrich the lives of billions through the many collaborations we have worked on across Alphabet over the years.”
Google, which generates 99% of Alphabet’s revenue, uses various technologies from DeepMind like WaveNet to make its virtual assistant speak more like a real person. Technologies like Google’s Text-to-Speech application, personalizing app recommendations in Google’s Play Store on Android devices, and Adaptive Battery and Brightness in Android devices have also been taken from DeepMind.
According to research company International Data Corporation, the artificial intelligence (AI) industry in India is anticipated to expand at a CAGR of 20.2 percent over the next five years, reaching US$7.8 billion in total revenues by 2025. (IDC). For the next five years, Indian enterprises will accelerate the use of both AI-centric and AI-non-centric applications, according to IDC.
By the end of 2025, the AI software sector will have dominated the market, growing at a CAGR of 18.1 percent from US$2.8 billion in 2020. Applications accounted for the biggest part of revenue in the software (AI) sector, with a 52% increase in revenue by 2020.
The survey discovered that businesses were using a variety of AI tools, including CRM, ERM, and others, to manage operations, grow supply chains in response to real-time or projected demand, and enhance ROI and cost savings.
According to IDC India Associate Research Director (Cloud and AI) Rishu Sharma, “Indian organisations plan to invest in AI to address current business scenarios across functions, such as customer service, human resources (HR), IT automation, security, recommendations, and many more.”
“Increasing business resilience and enhancing customer retention are among the top business objectives for using AI by Indian enterprises,” he adds.
Artificial Intelligence has already made its mark as a transformational technology in the digital age. This is unsurprising, considering that AI has the potential to bring about radical—and maybe unprecedented—changes in people’s lives and work. Although the AI revolution is far from over, the majority of its economic impact has yet to be realized.
The India AI market report gives an overview of the various data types that businesses are processing for AI-ML solutions. The report also shows the current state of AI projects in organizations and addresses the main concerns about AI-ML deployment methods. Indian organizations cited the cloud to be their preferred deployment location for their AI/ML solutions.
The study also discovered that disruptions in current business processes and a lack of follow-up from business units were two of the most common causes for AI initiatives failing.
Swapnil Shende, an AI Senior Market Analyst, stated With data being one of the most crucial components in an AI/ML project, businesses use a variety of databases to handle large data volumes for making real-time business decisions. Swapnil emphasized that organizations must concentrate on obtaining high-quality training data for AI and machine learning models.
NIT Warangal is now accepting applications from interested faculty and participants for a 10-day online Faculty Development Program (FDP) on “AI and Machine Learning for Biomedical Applications.” The Department of Biotechnology and the Department of Computer Science and Engineering is organizing the course.
The Faculty Development Programme will help disseminate the knowledge in AI, machine learning, and neural networks with python for Biomedical Applications. It empowers the participants to understand how they can use AI to innovate and improve medical-related applications.
Machine learning is a fast-growing field of artificial intelligence concerned with studying and designing computer algorithms for learning good representations of data at multiple levels of abstraction. Since data is overwhelming, organizations struggle to extract the powerful insights they need to make smarter business decisions. The FDP program will train the participants with a hands-on approach that’ll provide an in-depth understanding of the domain of AI & ML and expose them to feasibility & future scope.
The faculty members from NIT Warangal will conduct the program. Industry experts and academicians from IITs/NITs/IIITs are invited to deliver lectures as well. Selection for the FDP on AI and machine learning for Biomedical Applications will be on a first-come-first-serve basis. Interested faculty and working professionals can find the form here.
The maximum number of seats for the program is 60, and there are ten seats allotted to participants from the industry. NIT Warangal will intimate the list of selected participants through e-mail, and the DD will be sent back if a candidate is not selected. All the participants will be issued satisfactory certificates on successful completion of the course. The selection committee will follow reservations as per GOI norms.
Most of the marketing hacks emphasize paying attention to last-click attribution. However, today focusing solely on last-click attribution may not help businesses in the age of big data. With the transition to data-driven attribution, now Google has jumped the bandwagon. Google Ads, its buy-side ad network, will no longer employ last-click attribution as the default conversion model, the company announced in a blog post on last Monday.
While Google previously offered data-driven attribution, marketers and advertisers were unable to use it due to two factors: minimal data requirements and conversion types that were not supported. Google has now reduced the data restrictions and introduced support for more sorts of interactions to allow marketers to make the most of attribution and, as a result, enhance performance. In addition, it has made data-driven attributions the default attribution methodology for all-new conversion activities starting in October.
Marketers and advertisers may choose how much credit each ad interaction receives for conversions using an attribution strategy. As a result, they can gain a better knowledge of how advertising functions and aids in conversion journey optimization.
Early approaches for evaluating Internet marketing data were far too basic to account for the client’s emotional journey. It didn’t offer marketers the advice they needed, whether it was the first-click attribution model, which states the initial interaction on the purchasing path is the most important, or the last-click attribution model, which credits the final click.
Hence mapping the customer journey from their first click till the purchase is made (or the last viewed item before leaving the site) is important to identify what ‘clicked’ with the customers and what went wrong.
Data-driven attribution compares the pathways of consumers who converted with those who did not by looking at all touchpoints, such as clicks and video interactions, on your Google Advertising Search (including Shopping), YouTube, and Display ads. After then, the model looks for patterns in the encounters that result in discussions. For many advertisers, Google Ads will be moving current conversion activities to data-driven attribution in the coming months.
Through understanding which channels are contributing along the buyer journey toward a final conversion (as defined by the brand), the new additions imply that when used in conjunction with automatic bidding strategies or modifications to your manual bidding, data-driven attribution helps to drive subsequent conversions at the same CPA as of the last click.
Last-click attribution, according to Google, falls “short of marketers’ objectives.” On the other hand, data-driven attribution will be more accurate overall by analyzing relevant data about the marketing moments leading up to a conversion. While it’s tough to think about the entire route to buy, the new tool will look for similarities in ad interactions.
It claims that when combined with its automated bidding process, it can deliver more conversions at the same cost to advertisers than before because it considers various interactions people have with a brand before the last click and employs Google’s special predictive analytics and ad targeting optimization sauce. In comparison with machine learning tools, such as conversion modeling, data-driven attribution helps marketers get accurate insights into how each marketing touchpoint contributes to a conversion, without breaching user privacy.
By early 2022, all Google Ads accounts will be using the modeled attribution method, as per the company.
With the introduction of the new model, Google Ads will no longer require data for campaigns, allowing marketers to leverage data-driven attribution for every conversion activity.
Furthermore, Google Ads, which now only supports data-driven attribution for Search, Shopping, Display, and YouTube ads, is expanding support for new conversion types, such as in-app and offline conversions. As a result, the improved data-driven attribution methodology will assist marketers in fully comprehending the value of their Google ads. Google intends to continue to advance machine learning in order to improve existing measurement tools and build new ones that will assist marketers in delivering performance while respecting customer preferences.
Advertisers will still be able to toggle off data-driven attribution and choose for one of Google’s five rules-based attribution methods: last-click, first-click, linear (which rewards every impression equally), time-decay (which credits depending on the time between an impression and conversion), and position-based (40% credit each to the first and last impressions, and 20% distributed over the remainder).
Marketers have struggled with attribution for a long time. This problem is exacerbated by the belief that FLoC (Federated Learning of Cohorts) threatens to strip search advertisers of even more data, forcing them to piece together data on their own. Unlike cookies which are an invasive nuisance, FLoC allows advertisers to track internet users without revealing their identities.
The machine learning attribution methodology in Google Ads appears to be Google’s answer to this data shortage. “Data-driven attribution is privacy-focused, as it learns on real conversion routes and utilises machine learning to assess and predict conversion credits across touchpoints, even when cookies aren’t present,” asserts Google.
India’s national policy think tank, The National Institution for Transforming India, NITI Aayog establishes Experience Studio with Amazon Web Services (AWS) and Intel. The new experience studio is housed at the NITI Aayog’s New Delhi headquarters. The studio will open under NITI Aayog Frontier Technologies Cloud Innovation Center (CIC). It’ll be a hub for collaboration and research to facilitate innovation among startups, businesses, industry domain experts, and government stakeholders.
The studio will demonstrate the possibilities and applications of technologies including artificial intelligence (AI), machine learning (ML), the Internet of Things (IoT), blockchain, robotics, augmented reality (AR), and virtual reality (VR) in the public sector. The experience studio at NITI Aayog will serve as a platform for the Indian government, education, healthcare, and nonprofit enterprises for developing solutions to build the digital agriculture ecosystem, enable digital healthcare, and develop the digital infrastructure for India’s smart cities.
Various industry leaders will show their solutions in the experience studio in industries like AR/VR, drone, geospatial, and IoT technologies in multiple verticals, including agriculture, healthcare, and intelligent infrastructure. There will be leaders such as Raphe mPhibr Pvt. Ltd. in the unmanned aerial vehicles industry and MapMyIndia in geospatial solutions. Centre for Advanced Research in Imaging, Neuroscience, and Genomics, which provides AI in healthcare and Dassault Systemes, will also demonstrate their solutions. Vizara Technologies and Agatsa Software Private Ltd will present their new products.
The studio will undertake events like hackathons, grand challenges, and other capacity-building efforts (AIC) in partnership with the Atal Innovation Mission (AIM) and Atal Incubation Centres to encourage startups to actively participate in the studio.
Dr. Rajiv Kumar, Vice Chairman, NITI Aayog, Shri Amitabh Kant, CEO, NITI Aayog, and Rahul Sharma, President, Public Sector, Amazon Internet Services Pvt. Ltd. (AISPL), AWS India and South Asia, officially opened the studio on 30th Sept, 2021. Prakash Mallya, VP and MD of Intel India’s Sales, Marketing, and Communications Group, virtually attended the inauguration.
NITI Aayog, AWS, and Intel collaborate to promote NITI Aayog’s aim to discover and deploy cutting-edge technology that can enable continuous innovation in delivering citizen services. AWS Cloud Innovation Centers model will be beneficial in efficiently solving public sector challenges through cooperation.
It is no brainer that China has begun its crusade to outshine the rest of the world in the global AI tech industry. As the tension against the USA ensues, both nations are striving to become the world leader in AI by rapidly developing breakthrough technologies that will revolutionize the field and ensure their dominance over others. While earlier China prioritized innovation, recently, it released its first set of ethical guidelines governing artificial intelligence. Not only does it emphasize protecting user rights and preventing risks, it also aligns with Beijing’s goals of reining in Big Tech’s influence and becoming the global AI leader by 2030.
The guidelines, titled “New Generation Artificial Intelligence Ethics Specifications,” were prepared by an AI governance group created under China’s Ministry of Science and Technology (MOST) in February 2019. In June of that year, the group presented a set of guiding principles for AI governance, which were significantly shorter and broader than the recently disclosed criteria. The document was issued last Sunday.
In 2017, China announced its AI Development Plan (AIDP), aiming to make itself an AI powerhouse by 2030, surpassing its competitors to become “the world’s top artificial intelligence innovation hub.” The country also wants to turn AI into a trillion-yuan business and become the driving force behind the development of ethical AI rules and standards. Following that, the Chinese government-backed up its AI initiatives with substantial government funding. All these factors were pivotal in boosting China’s global proportion of AI research publications from 4.26 percent (1,086) in 1997 to 27.68 percent (37,343) in 2017, outpacing every country in the world, including the United States.
But simply laying out a bevy of milestones is not enough. SenseTime, Unisound, iFLYTEK, and Face++ indeed are just a few of China’s world-leading businesses in computer vision, speech recognition, and natural language processing. The country also benefits from its vast population, which presents an enormous potential workforce and unique opportunities to train AI systems, like large patient datasets for training software to predict disease. While the United States has open-source platforms like TensorFlow and Caffe to drive innovation, Baidu’s PaddlePaddle is primarily used in China for the rapid development of AI products.
However, the mandarin nation falls behind in hardware and has been widely criticized for using AI as a way to monitor citizens, especially the Uighur Muslim community in Xinjiang.
In recent years, many governments, research groups, and tech behemoths have released ethical standards, principles, and suggestions for the ethical use of AI. It is critical to create AI governance technologies, such as AI interpretation, rigorous AI safety testing and verification, and AI ethical evaluation, to impose them in existing AI systems and products. This is necessary because many AI technologies are still in the early stages of development and are not yet ready for widespread commercial use. Some of the factors that go into building ethical AI may be found in guidelines and legal literature. Security and privacy, safety and dependability, openness, responsibility, and fairness are among them.
While the trade war between the USA and China will continue in the coming years, they need to find common ground when addressing ethical AI concerns.
In 2019, the Beijing AI Principles were released by the Beijing Academy of Artificial Intelligence (BAAI), supported by the Chinese Ministry of Science and Technology and the Beijing city government. They stated that “human privacy, dignity, freedom, autonomy, and rights should be properly respected” as guiding principles for AI research and development.
The Cyberspace Administration of China (CAC), China’s internet watchdog, published proposed regulations in August this year to govern the use of algorithmic recommender systems by online information services. The recommendations are the most thorough effort to govern recommender systems by any country to date, and they might serve as a model for other countries considering similar laws. Unfortunately, this three-year plan is also an attempt to limit the use of algorithms, signaling Beijing’s latest move to strictly control the country’s internet economy.
The latest move by the Chinese government is another attempt to exercise control over the nation’s tech sector without putting user security and privacy at risk. The idea is to give users control over how their interactions with AI are handled. Hence the data security, personal privacy and the right to opt-out of AI-driven decision-making are also included in the document.
The document states that preventing risks necessitates identifying and fixing technical and security vulnerabilities in AI systems while ensuring that relevant organizations are held accountable and that AI product quality management and control are enhanced. The guidelines also prohibit AI products and services from engaging in unlawful actions or placing national, public, or manufacturing security at risk. They should also not be allowed to undermine the public interest, according to the document.
To surmise, protecting user privacy and data should be paramount for every nation, and China is beginning to take infant steps in this direction.
Researchers at the University of Waterloo have developed a unique artificial intelligence-powered technology that automatically analyzes videos of Hockey. The technology can identify the name of the players by their jersey numbers.
It uses various deep learning algorithms to generate results with accuracy as high as ninety percent. Researchers at the University of Waterloo combined two existing deep learning models to develop this automatic player recognition system that helps in effortless and seamless game analysis.
The University of Waterloo had partnered with a data and statistics providing company named Stathletes Inc to develop this technology that allows to analyze player performance and generate other valuable insights regarding hockey matches.
Kanav Vats, a Ph.D. student in systems design engineering who led the project, said, “That is significant because the only major cue you have to identify a particular player in a hockey video is jersey number.” He further added that prior to this technology, it was very challenging to identify players in a game as they look similar to each other because of helmets and jerseys.
Researchers also created a new dataset consisting of over 50,000 images to test their developed artificial intelligence and deep learning algorithm. Vats mentioned that they generated better results when they used different representations to teach the same thing to their AI algorithm. The two types of representations used during the training process were holistic and numerical.
The machine learning technology automates the annotation process and generates results in a few minutes, which, if done manually, would consume several hours. It can be expected that the researchers soon release such tools for other sports like football.
Kanav Vats will present this paper at the 4th International ACM workshop on Multimedia analysis in Sports, which is to be held in late October 2021.
To counter rising cases of social engineering attacks, a four-member team of computer science students from Rajagiri School of Engineering and Technology (RSET) have developed a deep learning-based solution. The primary intent is to detect websites with phishing motives that are spread widely across various social platforms. This solution is taken into consideration by Kerala Police Cyberdome and will be released (in the process) as an additional feature in the BSafe application.
Ever since the pandemic last year, the probability of companies and individuals being hit by phishing attacks has increased manifold. Since then, many innocent WhatsApp users who unknowingly have clicked on innocuous-looking links have become prey to phishing attacks. Phishing attacks are intended to steal information like login credentials or credit card numbers by an act of spoofing humans by whaling, Email phishing, or vishing.
Realizing the importance of this problem, RSET students developed a deep learning-based solution as a part of their B.Tech final year project. As this solution was the need of the hour, it instantly caught the attention of Kerala Police Cyberdome, which is currently in the process of adding this project as a module to their existing application — BSafe — alerting fraud and spam calls.
Sangeetha Jamal, Assistant Professor, RSET, guided four-member teams: Nithin Valiyaveedu, Roshan Reju, Nithin K.M, and Vysakh Murali. “Unlike the existing models that detect phishing attacks only based on website URL, their solution also collects HTML and includes a script to classify a malicious website. They have also developed a browser extension to quickly run phishing checks on a given URL,” said Nithin Vailyaveedu, team member.
This team further reached Cyberdome with two presentations that impressed Cyberdome representatives about the solution to counter phishing. The team has shared their API (application programming interface) code of solution with Cyberdome for adding it as a module in the BSafe app.
Cyberdome sources said, “BSafe — a mobile and web-based application, is capturing a number of scams in a database, helping users of the app to alert/block such numbers automatically. We plan to incorporate the phishing attack detection as an additional feature in the same app (under process).”
Unlike existing machine learning-based models for phishing attacks, Mr. Roshan said their innovation is more advanced, accurate, and has greater computation power. Although all youngsters have got a job, they have not marketed their solution as a separate product and chose to associate with Cyberdome for the greater benefit of the public.
Chandigarh plans to adopt its new artificial intelligence-powered traffic management system by February 2022. The new technology named Adaptive Traffic Control System (ATCS) will be added as a feature to the city’s Integrated Command Control Center (ICCC) that would enable artificial intelligence-enabled traffic management.
The technology will allow traffic signals to analyze road traffic conditions and operate in accordance with those situations for optimizing the flow of traffic. ATCS has already been deployed at over forty heavily congested junctions and will be operational soon.
According to the union territory administration, the artificial intelligence system will prioritize mobilizing heavy traffic areas while allowing maximum green signals. Additional Chief Executive Officer of Chandigarh Smart City Limited, Anil Gard, said, “The system will become fully operational in February 2022. Once the ICCC, which is under construction at Sector 17, is ready, the facility will become fully operational.”
Adaptive Traffic Control System will use various types of sensors and cameras to scrutinize surroundings, including individual traffic lanes, the total number of vehicles, and many more aspects for calculating the duration of traffic lights to maintain a smooth and uninterrupted flow of traffic.
The joint project being run by Chandigarh Smart City Limited and Bharat Electronics Limited has been besting this artificial intelligence-powered technology for the past month and a half at the Tribune Chowk, and company officials said that the results obtained were up to the mark.
“This AI-based system will reduce the travel time within the city. This will reduce the waiting time at intersections and also save fuel wastage due to idling at intersections,” said NP Sharma, Chief General Manager of CSML.
Apart from ATCS, ICCC will also include other technologies like Integrated Traffic Management System (ITMS), Public Address System (PAS), smart CCTV surveillance, Dynamic Messaging System (DMS), and many more. NP Sharma added that all these technologies would be installed phase-vise to ensure seamless integration.
Saudi Data and Artificial Intelligence Authority (SDAIA) has started its new data and AI accelerator program for startups operating in Saudi Arabia. SDAIA is now accepting applications for its new initiative that will provide a three-month-long intensive training program to the selected startups in Riyadh.
The new program will focus on the development of smart cities and will strengthen the entrepreneurship ecosystem of the country. This development is a step forward towards the country’s goals of Saudi Vision 2030 as the program will considerably help Saudi to diversify its economy and boost the adoption rate of digital technologies.
The program will be conducted in three stages, which would include registrations, screening, selection process, and training. A total of ten startups will be selected who will receive specialized training, attend workshops, and also showcase their projects to potential investors to seek financial assistance.
President of SDAIA, Dr. Abdullah Alghamdi, said, “At SDAIA, we work to enhance the accelerators’ ecosystem in the Kingdom by targeting local and international startups in the field of Data and AI, we are determined to place the Kingdom at the forefront of data and AI enabling countries.”
He also encourages startups to participate in this program and develop groundbreaking artificial intelligence solutions that can be implemented in smart cities. With this AI accelerator program, Saudi Arabia wants to get among the top five contributing countries in the world in the field of artificial intelligence.
“The accelerator program is one of SDAIA’s initiatives in cooperation with Plug and Play, to accelerate startups growth in the field of data and AI,” said the deputy director of the National Information Center and the CEO of strategy management at SDAIA, Dr. Mishari Almishari.
He further mentioned that the accelerator program is meant to aid startups to grow by providing them with ample exposure, networking, and training to help them acquire the necessary skills.
