A team of researchers at USC is helping Artificial Intelligence (AI) imagine the unseen, a technique that could also lead to fairer AI, new medicines and increased autonomous vehicle safety.

Imagine an orange cat. Now, imagine the same cat, but with coal-black fur. Now, imagine the cat strutting along the Great Wall of China. Doing this, a quick series of neuron activations in your brain will come up with variations of the picture presented, based on your previous knowledge of the world.

In other words, as humans, it’s easy to envision an object with different attributes. But, despite advances in deep neural networks that match or surpass human performance in certain tasks, computers still struggle with the very human skill of “imagination.”

Now, a USC research team comprising computer science Professor and Ph.D. students, has developed an AI that uses human-like capabilities to imagine a never-before-seen object with different attributes. The paper, titled “Zero-Shot Synthesis with Group-Supervised Learning,” was published in the 2021 International Conference on Learning Representations on May 7.

“We were inspired by human visual generalization capabilities to try to simulate human imagination in machines,” said Ge, the study’s lead author.

“Humans can separate their learned knowledge by attributes—for instance, shape, pose, position, color—and then recombine them to imagine a new object. Our paper attempts to simulate this process using neural networks.”

AI’s generalization problem

For instance, say you want to create an Artificial Intelligence (AI) system that generates images of cars. Ideally, you would provide the algorithm with a few images of a car, and it would be able to generate many types of cars—from Porsches to Pontiacs to pick-up trucks—in any color, from multiple angles.

This is one of the long-sought goals of Artificial Intelligence (AI): creating models that can extrapolate. This means that, given a few examples, the model should be able to extract the underlying rules and apply them to a vast range of novel examples it hasn’t seen before. But machines are most commonly trained on sample features, pixels for instance, without taking into account the object’s attributes.

The science of imagination

In this new study, the researchers attempt to overcome this limitation using a concept called disentanglement. Disentanglement can be used to generate deepfakes, for instance, by disentangling human face movements and identity. By doing this, said researcher, “people can synthesize new images and videos that substitute the original person’s identity with another person, but keep the original movement.”

Similarly, the new approach takes a group of sample images—rather than one sample at a time as traditional algorithms have done—and mines the similarity between them to achieve something called “controllable disentangled representation learning.”

Then, it recombines this knowledge to achieve “controllable novel image synthesis,” or what you might call imagination. “For instance, take the Transformer movie as an example” said researcher, “It can take the shape of Megatron car, the color and pose of a yellow Bumblebee car, and the background of New York’s Times Square. The result will be a Bumblebee-colored Megatron car driving in Times Square, even if this sample was not witnessed during the training session.”

This is similar to how we as humans extrapolate: when a human sees a color from one object, we can easily apply it to any other object by substituting the original color with the new one. Using their technique, the group generated a new dataset containing 1.56 million images that could help future research in the field.

Understanding the world

While disentanglement is not a new idea, the researchers say their framework can be compatible with nearly any type of data or knowledge. This widens the opportunity for applications. For instance, disentangling race and gender-related knowledge to make fairer AI by removing sensitive attributes from the equation altogether.

In the field of medicine, it could help doctors and biologists discover more useful drugs by disentangling the medicine function from other properties, and then recombining them to synthesize new medicine. Imbuing machines with imagination could also help create safer AI by, for instance, allowing autonomous vehicles to imagine and avoid dangerous scenarios previously unseen during training.

“Deep learning has already demonstrated unsurpassed performance and promise in many domains, but all too often this has happened through shallow mimicry, and without a deeper understanding of the separate attributes that make each object unique,” said Itti. “This new disentanglement approach, for the first time, truly unleashes a new sense of imagination in A.I. systems, bringing them closer to humans’ understanding of the world.”

The post Enabling the ‘Imagination’ of Artificial Intelligence appeared first on Artificial Intelligence.

Machine learning is enabling a smooth shift in this COVID-19 struck world.

Machine learning is one of the most used technologies in this generation. It has varied capabilities that can transform businesses across industries for the better. From being considered as a niche technology, machine learning is now seeing an increased adoption within companies in all sectors.

From a global perspective, brands are leveraging machine learning to accelerate innovation and better customer experience. For example, Nike uses machine learning for personalized product recommendations. In the F&B industry, Dominos maintains its 10 minutes or less pizza delivery time using machine learning technologies. Another widely used example is how automobile giant BMW uses machine learning to analyze data from vehicle subsystems and predicts the performance of vehicle components and recommends when they should be serviced.

In 2020, machine learning became a priority for tech companies in order to achieve revenue growth while reducing costs. In 2021, those companies are now exploring many matured applications of this technology. Disruptive tech organizations have been leading this technology across many areas like process automation, customer experience, and security.

Following the continuing growth trend, these five industries are likely to adopt machine learning to change their business processes in 2021.

Healthcare Industry

The coronavirus global pandemic has highlighted the importance of investing on and optimizing the healthcare systems. Machine learning is being considered as the most promising technology that enables healthcare providers to generate large volumes of data for insightful clinical decisions. Machine learning also enables huge processes in drug discovery, cutting down the long discovery and development time and reducing overall costs. It can also improve healthcare delivery systems to better the overall quality of healthcare under low costs. In the future, machine learning is predicted to be a critical part of clinical trials. Including pharmaceuticals and the biotech industry, machine learning will be having a huge impact in all aspects.

Banking and Finance Sector

The banking sector is already seeing many advanced use cases of machine learning, especially when it comes to fraud detection and automating processes. Machine learning applications will be proactively explored in areas in trading, investment modeling, risk prevention, and customer sentiment analysis. As countries are making digital transactions their primary mode of payment, machine learning is combining predictive analytics to play a pivotal role in helping financial companies to improve transaction efficiencies within the entire transaction lifecycle. Banks and financial institutions will also use machine learning technology to customize their banking products and offerings to stay up to date in the competitive environment.

Media And Entertainment Industry

Media giants like Amazon and Netflix have already popularized the data-based content consumption channels in recent times. When the world got initially struck with the global pandemic, the demand for new consumption models grew and left companies to leverage their artificial intelligence and machine learning capabilities to create value for the customers. In this process, machine learning is going to be crucial for the media and entertainment industry , whether it’s developing better recommendation engines, delivering hyper-targeted services, or presenting the most relevant content in real-time. Predictive modeling will also be key in communicating with the customers on time, anticipating their future demands, and making good investments.

Retail And Commerce Industry

The retail industry saw a big shift owing to the coronavirus pandemic. The pandemic has disrupted many traditional practices of this industry and machine learning has become a key enabler of change. From the perspective of brick and mortar stores or e-commerce companies, machine learning is helping this sector reinvent their supply chain, inventory management, predicting user behaviour, and analyzing trends. Dynamic pricing is emerging as a key machine learning application to help retailers thrive in the competitive market.

Manufacturing Industry

IoT devices have already flooded this industry and it is only going to increase. Machine learning will be critical to bridge the gaps created by huge amounts of data. It will serve as a building block for the industry along with automation, data connectivity, real-time error detection, supply chain visibility, warehousing efficiency, cost reduction, and asset tracking. Keeping traditional processes aside, machine learning will facilitate innovation and efficiency in the coming days.

The post IN 2021, MACHINE LEARNING IS SET TO TRANSFORM THESE 5 INDUSTRIES appeared first on Artificial Intelligence.

Led by former engineers from Tesla and Faraday Future, Hemant Sikaria and Mayank Sikaria, Sibros is an automotive technology company offering a next-generation vehicle-to-cloud automotive software platform. 

CEO and co-founder Hemant Sikaria had the light bulb moment to start Sibros when he had to make multiple trips to his car dealer due to a software recall. Upon further research, he realised the majority of OEMs did not have the necessary capabilities for advanced full vehicle updates. In fact, global OEMs have been facing significant software and data complexity challenges as vehicles were increasingly shifting towards electrification and self-driving modes. Synergising Hemant’s experience at Tesla in building systems that could solve these problems with Over-the-Air (OTA) programming, and Mayanks’s significant contributions to the software, data and battery management systems at Faraday, the duo teamed up to form Sibros.

Analytics India Magazine got in touch with the founders to understand how Sibros is leveraging machine learning and artificial intelligence to help automakers build new connectivity use cases and business models.

Leading The Way With ML

The advancement in 5G technology along with software-defined vehicles has allowed Over-the-Air software updates to extend beyond just telematics and head unit applications. As a matter of fact, the connected car revolution has made OTA software an industry standard for automakers. For the uninitiated, OTA can reduce consequential software recalls to create new revenue opportunities, and enhance the customer experience by shortening the time required to update, fix, and maintain vehicles throughout their entire lifecycle.

Sibros’ flagship product, OTA Deep Updater executes full vehicle Over-the-Air (OTA) software updates while logging vital vehicle data in real-time from every sensor and microchip. The solution supports any vehicle platform, architecture and hardware, starting from two to four wheelers, commercial trucks/busses to electric, and internal combustion to hybrid vehicles. 

Explaining the process, Hemant stated: Sibros collects data at the edge that is processed with machine learning and artificial intelligence to help automakers build new connectivity use cases, and business models, like variable insurance, driver personalisation, fixing software recalls overnight, crash detection, etc.

With its modular system that provides full vehicle OTA updates to every sensor or engine control unit (ECU) while logging real-time vehicle data, Sibros’ deep connectivity platform really stands out in the market. “As modern vehicles can produce petabytes of data monthly, Sibros selectively capture the most meaningful data based on customisable events and triggers. This data is then further compressed and uploaded to the cloud backend,” explained Hemant. “This method enables efficient cloud computing throughput and enables OEMs to leverage live vehicle data for fleet/service management, developing new connected apps, building third party marketplaces and much more.” 

The company uses machine learning, data pipelines and IoT to abstract, process and interpret vehicle data, such as fault codes, into human-readable formats on visual dashboards. “Our efforts on the AI and analytics front are all geared towards improving capabilities around safety measurement, failure detection and preventative maintenance,” said Mayank. “We offer insights into driving behaviour of each user on the platform by combining vehicle data such as speed or harsh acceleration and braking with external data sources like traffic, data and time.”

Sibros OTA Deep Updater can also automatically detect vehicle movement risks by flagging abnormalities in a vehicle’s manoeuvring patterns. Component usage like blending weather, altitude and region information, battery health, and performance also play a crucial role in providing insights to automakers for determining the optimal configuration of a battery in different markets.

Sibros’ core technology stack leverages Golang as the primary backend language to handle high currency workloads, with an outstanding balance between performance and developer velocity. However, for the frontend, the company leveraged React.js and React Native to deliver the product to multiple platforms without having to rewrite the stack from scratch. “We also heavily leverage cloud-native solutions including SQL/No-SQL and Kubernetes for container orchestration to avoid reinventing the wheel, and build our infrastructure as code using Terraform,” said Mayank.

COVID Impact

“While initially, the COVID pandemic has ground the automotive industry to a halt, slowing the progress with plants shutting down, the things have now started to pick up slowly, and we see a tremendous uptick in business traction,” said Hemant. 

As remote working and contactless interactions are becoming the new norm, Sibros was able to deliver an embedded software and cloud/analytics solution remotely. The solution is also enabling field technicians not to be physically present while performing R&D and testing the beta vehicles. “As our solution is designed to enable fully Over-the-Air remote services, this ultimately benefits end customers to limit their exposure for visiting a dealer for servicing or utilising new connected services, such as in-vehicle payments for goods/food/tolling/parking,” added Hemant.

Despite COVID, Sibros has been able to build a diverse and inclusive work culture. The employees can work directly with clients in their initial months to see their ideas become a reality in their product roadmap and offerings. “Our current and future job requisitions span the gamut — from embedded firmware engineers and full-stack engineers to UI/UX, sales and marketing. We look for people who are coachable and rising stars, instead of necessarily looking for specific skill sets,” said Hemant.

Wrapping up

Sibros currently has presence in multiple locations and is engaged with the largest Indian OEMs of two, three, and four wheeler vehicles. The founders believe that India is a high growth area for Sibros in terms of talent acquisition and building a customer base. The country holds great promise for the future of major automotive trends, particularly in areas of rapid charging infrastructure and EV production/adoption, they said.

The company has grown to over 40 employees and has raised $15 million in a Series A funding round led by Nexus Venture Partners. “Our customer base includes the vehicle manufacturers directly spanning some of the world’s largest funded EV startups to some of the biggest makers of two and four-wheeler vehicles,” added Hemant.

Speaking about the company’s five-year plans, Hemant said Sibros intends to enter other verticals where the technology can be applied with minimal changes. “We also plan to stand up a centre of excellence in Pune as well as expanding our presence in Germany, Japan and China. Our vision is that in five years, the connected vehicle ecosystem will become a mainstream reality, powered by Sibros. Much like smartphone ecosystems gave birth to disruptive services and apps such as Uber or Airbnb, so will the vehicle ecosystem. It’s moving rapidly, and we expect to gain much progress over the next five years to accomplish this vision,” he added.

The post Enabling Connected Vehicle Ecosystem With Machine Learning: The Startup Story Of Sibros appeared first on Artificial Intelligence.