Monday, November 10, 2025
ad
Home Blog Page 349

Top AI News Of The Week [October 11, 2020]

By
Analytics Drift
-
AI News October 11, 2020

AI news of the week (October 11, 2020), although overshadowed by announcements and opinions from the conferences like NVIDIA GTC and RAISE, was packed with other interesting developments by tech-giants like Instagram and Waymo. On the other side, the antitrust committee in the US finds that Facebook, Google, Amazon, and Apple have a monopoly in the tech landscape, which hindered innovation.

Here are the top AI news from the week — October 11, 2020.

NVIDIA GTC 2020

The five-day NVIDIA GTC concluded on Friday had numerous announcements by Jensen Huang, co-founder and CEO of NVIDIA. The GPU provider doubled down on enhancing cloud computing with its BlueField 2 DPU, which is powered by Arm cores. Since the cloud has become foundational for innovation, NVIDIA is striving in the field with its hardware products. Product releases, however, were not limited to improving the cloud’s performance. The company is building the world’s first Drug Discovery Lab to expedite the workflows of scientists for drug or vaccine discovery. NVIDIA GTC had several other announcements that you can read here.

Waymo Launches Its Self-Driving Cars To The Public

Waymo, on October 8, launched its self-driving cars for the public. Users can leverage the app, which is available on Google Play and the App Store to get their rides for friends and family. But, the service is currently limited to the Phoenix area and is only available for Waymo One members. The company, however, is committed to open up the service for others in the coming weeks. And later this year, Waymo will introduce vehicle operators in the vehicle to further expand rapidly by serving in a wider geographical area.

Instagram Will Block Offensive Comments

Instagram rolled a feature that will block offensive comments using machine learning. It is mostly similar to what Twitter does with hiding comments for potentially offensive comments. But, it varies in a way that Instagram will consider the type of comment a user has previously reported as offensive, thereby making it more personalized for the users. This feature is in beta for now.

In addition, Instagram is also using AI to recognize if a user is about to comment that can be against the guidelines of the platform. Following this, Instagram will nudge the user to avoid posting a comment.

NeurIPS 2020 Accepted Papers And The Response

Papers with Code on Twitter: "NeurIPS 2019 Implementations - get all the papers with code in one place here: https://t.co/WmZUoc9xg2… "
NeurIPS 2020

NeurIPS has announced the list of accepted papers for its virtual conference, which will be held from December 6 to 12. A total of 1,903 papers were approved, outnumbering the acceptance count of 1,428 in 2019. Led by Google (169), Stanford (104), and MIT (98), 1,178 papers that got the approval were from the US.

There was a lot of criticism by researchers whose papers were not accepted, but Hima Lakkaraju, Assistant Professor of Harvard, in a series of tweets noted that there is a biased response from the students on Twitter. She mentioned that around 20% of the papers are accepted, and most (80%) of it rejected.

Antitrust

In a 450 page report, the antitrust panel pinpointed the monopoly of Facebook, Amazon, Google, and Apple. These companies have worked towards eliminating competitions by either acquisition or cloning competitors’ products. 

“The Subcommittee’s series of hearings produced significant evidence that these firms wield their dominance in ways that erode entrepreneurship, degrade Americans’ privacy online, and undermine the vibrancy of the free and diverse press. The result is less innovation, fewer choices for consumers, and a weakened democracy,” noted the report.

The report also undermined the efforts of these companies in delivering benefits to the society by pointing that it has come at a cost. “These firms typically run the marketplace while also competing in it—a position that enables them to write one set of rules for others, while they play by another, or to engage in a form of their own private quasi regulation that is unaccountable to anyone but themselves,” the report said.

Advertisement

NVIDIA NeMo: Develop Conversational AI Models In 3 Lines Of Code

By
Analytics Drift
-
NVIDIA NeMo

NVIDIA at GTC released NVIDIA NeMo (Neural Modules), an open-source toolkit to develop Conversational AI models on the fly. As per the company, you can make Conversational AI processing models in just 3 lines of code. The PyTorch-based NVIDIA NeMo consists of building blocks and pre-trained models, which can be leveraged to fine-tune or build upon with minimal effort.

Developers and researchers often go through complex processes: data pre-processing, modifying several networks, and verifying compatibility across input layers while developing NLP-based modes. To address such challenges, NeMo comes with domain-specific–automated speech recognition (ASR), NLP, and text-to-speech (TTS)–modules and building blocks.

NVIDIA NeMo caters to a wide range of Conversational AI workflows by bringing the entire dependencies in one place. Besides, NeMo models can be exported to NVIDIA Jarvis–an application framework for multimodal Conversational AI services that delivers real-time performance on GPUs–for high-performance interface with a single command. “You can expert models in ONNX, PyTorch, and TorchScrip,” noted in the NVIDIA blog.

Users will have access to state-of-the-art models like QuartzNet, Jasper, BERT, TRacotron2, and WaveGlow in just three lines of code from NVIDIA NGC. Most of the models available in NGC are trained for over 100,000 hours on NVIDIA DGX™ across a wide range of datasets. One can effortlessly modify these domain-specific modules according to their requirements and deploy with minimal clutters.

At the time when Conversational AI is proliferating in video calls for real-time transcripts and chatbots for customer services, NeMo can become an enabler for organizations in delivering superior customer experience. Companies can use NeMo and streamline their development and deployment of Conversational AI models in their products and services.

Also Read: A Glimpse At OpenAI’s GPT-3 Pricing

Building NLP models is very complicated as it requires enormous computing power to train the models. With NVIDIA NeMo, you can minimize the cost while ensuring you are deploying superior Conversational AI models.

You can read more about NeMo and check some of the tutorials here.

Credit: NVIDIA

Stay tuned for more announcement from NVIDIA GTC.

Advertisement

IBM Is Offering Free Year-Long Quantum Computing Course

By
Analytics Drift
-
IBM Quantum Computing Course

In association with The Coding School, IBM is offering a free “Qubit by Qubit’s Introduction to Quantum Computing” course. The program is open for high school students or above; you can apply even if you are a working professional. A total of 5,000 learners will be able to enrol in this course for free. However, students from high school would be given priority while evaluating the applications.

The course will include weekly lectures and labs led by TAs, which will help learners to acquire right from the introduction of quantum computing to quantum mechanics and quantum algorithms. Students will also have the opportunity to run a program on a real quantum computer.

Also Read: Udacity Is Offering Yet Another Scholarship For AI, Data, & Cloud Track

“We don’t want education access to be the roadblock in the way of any passionate student, no matter their location or circumstances, to make a potentially groundbreaking impact,” authors noted in the IBM blog post.

Since IBM wants to ensure students are future-ready, aspirants do not need any pre-requisite, such as prior knowledge of quantum computing or STEM (science, technology, engineering, and mathematics). While applying for the program, you will only be asked for necessary details like name, occupation, the country you live in, email id, university/school and grade, and other demographics.

Also Read: Amazon Makes Its Machine Learning Course Free For All

The applications are being reviewed and will be accepted on a rolling basis in two rounds: October 9 and October 15. The sooner you apply, the better. However, if you enroll late, you may have a chance to be accepted based on the course’s available capacity.

IBM Quantum and Qiskit team has become the enabler in the field by providing free programs for students. The team had earlier hosted the Qiskit Global Summer School for two-week to impart knowledge of quantum computing, which was attended by 4084 students from 101 countries.

As IBM is one of the leaders in quantum computing, the free Qubit by Qubit’s Introduction to Quantum Computing course is a must for students and professionals.

Apply for the course here.

Advertisement

Udacity Is Offering Yet Another Scholarship For AI, Data, & Cloud Track

By
Analytics Drift
-
Udacity Scholarship

Udacity collaborates with Bertelsmann to offer yet another scholarship to more than 15,000 learners in the coming months. The scholarship is segmented into two parts, where learners will have to qualify the first phase to obtain the second part of the scholarship. Of the 15,000 learners who will be accepted for the first phase, only 4,800 will be offered advanced courses for AI, Data, and Cloud.

Learners will not have the option to choose among the three tracks–AI, Data, and Cloud–to apply for the scholarship. Udacity will be evaluating the right track based on the answers you provide while applying. Introduction to Azure Applications (cloud track), Introduction to Problem Solving and Advanced Analytics (data track), and Introduction to AI in Business (AI track) are the courses available for the first phase.

Also Read: Amazon Makes Its Machine Learning Course Free For All

However, not everyone completing the course will be offered an advanced course. Udacity usually considers various metrics to evaluate students, which they announce in the slack workspace as the first phase starts. Nevertheless, 1600 learners from each track will be promoted to the next step of the scholarship.

Unlike previous terms, this scholarship by Udacity doesn’t require you to go through a lengthy application phase. Udacity has removed the fields where you had to type in your response. 

The last day to apply for the scholarship is on November 16, 2020, and the first phase will start on December 3, 2020. Eventually, the advanced course will begin on March 15, 2020.

Also Read: A New Free Deep Learning Course By fast.ai

Anyone with age above 18 is eligible for the scholarship, but you would require some prerequisite knowledge to get started. In addition, if you can dedicate 5 hours of your time in a week, you will be able to qualify for the first phase. However, the advanced course will be a bit demanding and require you to spend up to 10 hours per week.

Read about the Udacity scholarship here and apply accordingly.

Advertisement

NVIDIA GTC 2020: Key Announcements

By
Analytics Drift
-
NVIDIA GTC 2020

The first day of NVIDIA GTC 2020 was packed with major announcements by NVIDIA in the artificial intelligence landscape. In the AI conference’s keynote, Jensen Huang, co-founder and CEO of NVIDIA, demonstrated NVIDIA’s effort in revolutionizing machine learning, data centers, autonomous vehicles, robotics, drug discovery, supercomputers, and more.

Here are the top announcements from NVIDIA GTC 2020:

1. NVIDIA BlueField 2 DPU

Today, when the world relies highly on cloud computing, cloud providers need a processor to handle the increased workloads effortlessly. Huang said that 20 to 30 percent of CPU cores are utilized to manage the data center’s infrastructure. To mitigate such challenges, NVIDIA released BlueField 2 DPU, a programmable processor powered with Arm cores, to effectively manage the cloud’s workloads.

Huang also mentioned that the successor of BlueField 2. The third and fourth generation of DPU would further enhance network management in the cloud. While the former is almost finished, the latter will be released in 2023 and will support CUDA parallel programming platform.

2. Jetson Nano 2GB

Announced at NVIDIA GTC 2020, Jetson™, an embedded system for next-generation autonomous machines, has a new member with Jetson Nano 2GB, which will be available for $59. Unlike other Jetson systems, Nano 2GB comes at a lower price; it is almost half of the Jetson Nano (announced last year). Focused at students and hobbyists, Jetson Nano 2GB is bundled with free online training and certification. Powered by 128-core NVIDIA Maxwell™, it is ideal for beginners to make AI-solutions.

You can pre-order Jetson Nano 2GB here.

3. Maxine A Video Call Platform

Due to the pandemic, there is a rapid increase–30 million web meetings every day–in video calls. Understanding the needs of users, NVIDIA at GTC 2020 announced Maxine, a video AI platform. It is not just any other video meeting application. Maxine is equipped with AI that reduces the bandwidth consumption of video calls by a factor of 10. Besides, it can align your face as if you are looking at the camera although you might be watching the other person on the screen.

4. NVIDIA Clara Discovery And World’s First Drug Discovery Lab

NVIDIA Clara Discovery is a suite of tools for scientists to discover drugs efficiently by leveraging artificial intelligence. Huang noted that drug discovery takes decades and costs over $2.5 billion in research. However, COVID-19 has brought the urge among scientists to discover drugs quickly. With NVIDIA Clara Discovery, the company wants to accelerate the workflows for scientists, thereby making the world future proof.

In addition, NVIDIA also collaborated with GSK to build the world’s first AI drug discovery lab. GSK will also get access to NVIDIA’s Cambridge-1–the most powerful AI supercomputer of the UK (announced at the NVIDIA GTC 2020)–to push the boundaries while discovering drugs.

5. NVIDIA EGX AI Platform

NVIDIA EGX AI platform will work in tandem with Ampere architecture GPU and BlueField-2 DPU on a single PCIe card. Using the EGX AI platform, organizations can quickly deploy and manage edge devices. Many firms like KION Group and Northwestern Memorial Hospital are already using EGX AI for supply chain solutions and IoT sensors. “This is the iPhone moment for the world’s industries — NVIDIA EGX will make it easy to create, deploy and operate industrial AI services,” said Huang at NVIDIA GCT 2020.

6. NVIDIA GPUs In Microsoft Azure For Microsoft Office

NVIDIA and Microsoft joined hands to bring its GPUs in Azure for smart experience for Microsoft Office users, including grammar correction and text prediction. Companies like Square, Twitter, eBay, GE Healthcare, and more are already using NVIDIA GPUs for inference.

Every cloud provider offers NVIDIA GPUs to support machine learning workflows. NVIDIA’s inference compute has exceeded all other cloud CPUs. It is envisioned that NVIDIA GPUs will represent 90% of the total cloud inference compute in the near future. The new collaboration with Microsoft Azure will further help NVIDIA gain extensive market share in the inference compute.

7. DOCA

DOCA, a programmable data-center-infrastructure-on-a-chip architecture, was yet another solution focused on data centers. “DOCA SDKs let developers write infrastructure apps for software-defined networking, software-defined storage, cybersecurity, telemetry and in-network computing applications yet to be invented,” Huang said.

Stay tuned for more stories from NVIDIA GTC 2020 announcements.

Advertisement

Tone Transfer By Google Can Convert Your Voice Into Music Using Machine Learning

By
Analytics Drift
-
Tone Transfer

Have you ever wondered how your voice would sound if it came out of a musical instrument like a flute or violin? Google’s team–Magenta–explored the possibility of using machine learning to convert your voice into music with Tone Transfer.

Leveraging Differential Digital Signal Processing (DDSP), the team developed an open-source technology to convert your voice into music. “DDSP allows developers to combine interpretable structure of classical DSP elements (such as filters, oscillators, reverberation, etc.) with the expressivity of deep learning,” noted Magenta.

With DDSP, the researchers created complex realistic signals by controlling various parameters, which was done by training neural networks to adapt to a dataset through standard backpropagation. This novel approach was also presented in ICLR 2020, where the researcher demonstrated the conversion of the sound of Voice → Violin and Violin → Flute. 

DDSP learns by extracting different characters of a musical instrument and maps it with different sounds, thereby creating appealing music. You can use a wide range of audio right from someone singing to dog’s barking and people talking to convert them into a sound from violin, flute or more.

Google last week released a portal where you can upload or record your audio and then transform it into a wide range of sounds of instruments.

Since researchers used western music to train the model, it might give unexpected results. However, DDSP transforms sound by modeling frequencies in the audio, ignoring the standards of western music. This reduces the possibility of distorted or undesired sound.

You can use the tool–Tone Transfer–to listen to how your voice sounds when matched with the sound of instruments. You can also download the audio to share across your network. 

Besides, you can use Magenta, a distributed open-source library for Python and JavaScript, to experiment and make your own machine learning models while playing with music to generate resounding audio.

Watch the video.

Advertisement

Amazon Launches Amazon One For Contactless Payment

By
Analytics Drift
-
Amazon One

Amazon introduces Amazon One–a device for a new approach to enable contactless payment for faster and secure transactions. Users can now scan their palms by imitating prompted gestures to validate the identity. Currently, the scanner will be available at selected Amazon Go stores (two stores in Seattle, South Lake Union at 300 Boren Ave, and more), but the company has plans to provide the device to other organizations as well. Amazon Go can be deployed in retail stores, stadiums, office buildings, among others.

Although it looks like the technology was developed to mitigate challenges caused due to the pandemic, Amazon was working on this novel technology before even the pandemic happened. On December 26, 2019, the largest e-commerce company was granted the patent for Hand ID technology, which was focused toward eliminating the use of cards and applications for transactions. Amazon One uses computer vision technology in real time to authenticate and make the transactions within seconds.

To use Amazon Go, users will have to undergo an initial setup, where they can provide necessary details such as phone number, credit card, and palm scan, to amalgamate for future authentication.

Undoubtedly, Amazon Go has the potential to revolutionize the way we verify identity in our day-to-day lives, but security concerns can be a significant roadblock. Unlike passwords that can be changed in case of hacks to fortify future security attacks, users’ biometrics data at the helm of hackers can haunt forever. 

In order to overcome similar security challenges, Amazon will store the data in its cloud instead of leveraging on-device storage. In addition, customers will have control over their data and can request to delete the information from the cloud, thereby enhancing customer experience.

You can also check your usage history and manage the account if you connect the Amazon One with Amazon account, but connecting your Amazon account is not mandatory to leverage Amazon Go.

Advertisement

GitHub Launches Code Scanning To Find Vulnerability In Your Code

By
Analytics Drift
-
GitHub Code Scanning

GitHub Code Scanning is now generally available for users to evaluate their codes for security flaws. The idea behind this initiative is to eliminate vulnerabilities before the code is in production. By enabling GitHub Code Scanning, every git push is scanned for potential security loopholes. The results are displayed directly in pull requests, making the users aware of the imperfection in code.

The open-source–GitHub Code Scanning solution–is powered by CodeQL, which consists of 2,000+ default queries to scan the code. GitHub has doubled down on security enhancements since it acquired Semmle on September 19, 2019. Semmle allowed developers to write queries to search for vulnerabilities.

GitHub Code Scanning

Now, with Code Scanning, developers can fortify security threats effectively. However, it does not guarantee a 100% secure code since security flaws can vary based on the workflows. Nevertheless, one can eliminate common vulnerabilities in the code or use customized queries to determine security issues.

Since the beta release in May, GitHub Code Scanning developers have scanned 1.4 million times on 12,000 repositories. More than 20,000 security issues such as SQL injection, cross site scripting (XSS), and remote code execution (RCE) vulnerabilities were identified. Of the total flaws, 72% of them were fixed before the pull requests were merged. Such efficiency is way ahead than the industry standards, where it takes more than 30 days to fix at most 30 percent of all the flaws.

“We chose Advanced Security for its out-of-the-box functionality and the custom functionality that we can build off of. Instead of it taking a full day to find and fix one security issue, we were able to find and fix three issues in the same amount of time,” said Charlotte Townsley, Director of Security Engineering of Auth0.

GitHub Code Scanning is open-source and can be used for free with public repositories, but is only available to GitHub Enterprise to scan private repositories. To enable it in public repositories, you can visit here.

Check the announcement here.

Advertisement

D-Wave Launches 5,000+ Qubit Quantum Computing Platform

By
Analytics Drift
-
D-Wave Quantum Computing

D-Wave System Inc, a quantum computing service provider, announces the general availability of its superior quantum computing platform, LeapTM. The advanced system can power the development of deep learning solutions and assist in solving complex material science problems.

LeapTM, with 5000+ qubit and 15-way qubit connectivity, is powered with the AdvantageTM quantum system to enable developers to solve real-world problems. The current iteration of D-Wave’s platform is a leap of 3000+ qubit to empower developers to effectively use hybrid solver service (HSS) for quantum computation with classical resources for application developments.

With HSS, uses can run applications with one million variables, up from 10,000 variables in the previous service. Such capabilities of LeapTM uniquely places it in the computing landscape to support a wide range of applications.

Along with the 5000+ qubit quantum computing platform, D-Wave released a slew of solutions: D-Wave LaunchTM and discrete quadratic model (DQM) solver. While LaunchTM helps businesses get started with hybrid quantum applications, DQM enables developers to solve a new range of problem classes. Along with accepting 0 and 1 as variables, DQM solver works with integers (1 to 500), or red, yellow, and blue. However, DQM will be available in general availability on October 8, 2020.

Companies like Menten AI have been benefiting from D-Wave hybrid quantum solutions. Menten AI has determined protein structure for de novo protein design, outperforming other classical solvers. In addition, Accenture and Volkswagen have been early adopters of the hybrid quantum service of D-Wave for developing applications for banking and paint shop scheduling, respectively.

According to a report, 39% of the surveyed enterprises are experimenting with quantum computing, and 81% of fortune 1000 decision-makers have a quantum computing use-case in mind for the next three years. As quantum computing is gaining momentum, D-Wave is striving at the right time to capture huge market share with its advanced computing services and helping organizations solve complex problems.

Read more here.

Advertisement

Top AI Conferences That You Should Look Ahead In 2020

By
Analytics Drift
-
AI Conferences

This year has delivered a completely different AI conference experience as they have moved virtually due to the pandemic. Many machine learning conferences were rescheduled to the end of the year in order to still host physical events. Unfortunately, things are not under control yet, resulting in the adoption of virtual conferences. Although networking in remote AI conferences is strenuous, it eliminates the geographic barrier of physical events. One can gain interesting insights from top researchers while being at home in 2020.

Here are the top 4 AI conferences that will be organized before the year ends.

RAISE (Responsible AI For Social Empowerment)

Organized by the Government of India, RAISE is a global five-day artificial intelligence summit. Many tech leaders like Arvind Krishna, CEO of IBM, Mukesh Ambani, chairman, MD of Reliance Industries Ltd, N Chandrasekaran, chairman of Tata Sons, and more, will be speaking at the event. Researchers and developers from across the world will join RAISE and talk about a wide range of topics related to artificial intelligence.

Yann LeCun, Chief of AI, Facebook, Dr Milind Tambe, Director “AI for Social Good” at Google Research India, Ms Nivruti Rai, Country Head of Intel, among others, will share their views on topics such as Explainable AI, Responsible AI, NLP, and more.

Starting on October 5, 2020, RAISE is a free AI conference that everyone should sign up for.

NeurIPS (Neural Information Processing Systems) 

NeurIPS is one of the oldest machine learning conferences catering to ML enthusiasts’ needs since 1987. Due to COVID-19, NeurIPS 2020 will be a virtual event, which will be held from December 6, 2020, to December 12, 2020. After ‘industry expo’ on the opening day, the rest of the days will be filled with tutorials, talks, and workshops. You can register for the summit here, which has varying prices for regular attendees and students.

You can access the video archive of NeurIPS here.

GTC (GPU Technology Conference)

GTC is organized by NVIDIA, which will start on October 5, 2020, and end on October 9, 2020. With over 600+ live and on-demand sessions, you will have an opportunity to learn from some of the best minds in the industry. Although the name gives a sense of a hardware event, it is a deep learning conference, covering topics like autonomous machines, conversational AI, AR/VR, and computing.

GTC also includes a full-day workshop from Deep Learning Institute (DLI) to teach learners working with neural networks, deep learning tools, frameworks, and SDKs. However, you will have to pay an additional $99 for attending this remote workshop.

Register for the conference here.

EmTech (MIT Technology Review)

EmTech will be held virtually from October 20, 2020, to October 22 by MIT Technology Review. The AI conference will feature Geoffrey Hinton, Professor Emeritus, University of Toronto, VP and Engineering Fellow, Google, Marc Benioff, Chair, CEO, and Cofounder of Salesforce, Parag Agrawal, CTO of Twitter, and more. The three-day event has been segmented into ‘leading with innovation,’ ‘forces of change,’ and ‘the path forward,’ where researchers and tech leaders will shed light on the possible direction for embracing AI.

Register for the event here.

Advertisement
1...348349350...354Page 349 of 354