At its annual developer conference (I/O) on May 10, Google unveiled Duplex, a new technology that allowed its virtual assistant to have a conversation with a human over the phone, completing real-world tasks such as fixing a salon appointment and making a dinner reservation. You could tell the assistant to make an appointment or reservation and it would then make the call, talk to the person respond to questions, negotiate timing and even thank the person at the end. While Duplex is bringing natural language processing, deep learning and text-to-speech technology together, what is remarkable is that the conversation was so natural that the other person didn’t realise that he or she was talking to a bot because it changed annotation according to the flow of the conversation. While computers have been able to understand and generate natural speech, automated phone conversations are still stilted, and Duplex is the fulfilment of a long-standing goal of enabling humans to have a natural conversation with computers.
At the core of Duplex is a recurring neural network which mimics a human brain, internalising past experiences and drawing from them during training. In fact, Google’s artificial intelligence (AI) arm, DeepMind’s latest program was able to beat humans in a maze game after it learnt to find its way like a human. It was faster and was even able to find shortcuts when doors opened up suddenly in the maze. Navigating spaces is second nature to us because of neurons called grid cells in our brains that help us find our way. The program was trained with data on how rodents search for food and it was able to mimic the process of navigating from one place to another. The team, which also trained the network to navigate through unfamiliar mazes through reinforcement learning, found that the neural network developed something similar to grid cells formed in human brains.
While deep-learning algorithms have always outperformed humans in areas such as diagnosis, and future prediction, they take a long time to learn. What takes us few hours to learn takes the best algorithms few hundreds of hours and a lot more data processing. According to researchers, humans are able to learn faster thanks to dopamine, a chemical in our brain that affects our emotions and plays a key role in the learning process. The potential to develop brain-like activity from AI systems, such as the one developed by DeepMind, could give scientists more insights into the complex way human brains work. Figuring out how we control our limbs will help AI researchers come up with technology to help people control prosthetic limbs with their thoughts using brain-to-machine interfaces.
Taking AI where human brains go is the next frontier says Kris Gopalakrishnan, chairman, Axilor Ventures and co-founder, Infosys, who is not only backing academic research in AI in India but also start-ups in this space. “That is the subject of many research papers. We need to discover much less intensive machine learning approaches to augment AI with natural intelligence. Augmented intelligence systems should learn with much less data and have the ability to learn from the ecosystem to accelerate the training process.”
Jim Breyer, Silicon Valley’s legendary VC who now runs his own fund, Breyer Capital, mentions there is a lot of research that suggests by 2050, machine intelligence will surpass human intelligence in many ways. He believes that 60% of the return generated by venture capital will be from AI, calling it the most interesting investment theme of the next decade. The real innovation behind AI is deep learning and every industry is tweaking it to break new boundaries in their space. For instance, Breyer, who is betting big on AI, believes that healthcare will be one of the biggest beneficiaries. “I think the intersection of AI and healthcare is especially compelling because there are numerous areas where AI will be beneficial. I’m especially excited about AI impacting workflow optimisation, imaging (diagnosis and workflow efficiency), clinical trial recruitment and retention, and drug discovery,” he says. In healthcare, while there are several use cases, AI is now increasingly being used in the early detection of cancer by analysis of blood samples. Breyer is a lead investor in Paige.AI, a company revolutionising clinical diagnosis and treatment in oncology through AI. Paige.AI has signed an exclusive licence with the Memorial Sloan Kettering Cancer Center to gain access to its intellectual property in computational pathology as well as exclusive rights to its library of 25 million pathology slides.
Both the US and China want to be at the helm of AI domination and why not. PricewaterhouseCoopers predicts that AI can contribute up to $15.7 trillion to the global economy by 2030. China believes that its domestic AI industry will be worth $150 billion by 2030 and wants to be the world leader in AI by then. It is already home to the most valuable AI company in the world. SenseTime, valued at more than $4.5 billion, makes AI software that recognises objects and faces, and counts China’s biggest smartphone brands as its customers. The company is developing a software code, Viper, which can screen data from thousands of live feeds ranging from traffic cameras to ATMs. This will be used for mass surveillance by the government, which is already using its facial recognition and video analysis software. While the mass surveillance sparks privacy concerns, a lot of its video analysis is being used to improve the safety of driverless cars. “The top AI researchers and entrepreneurs in China are as good as their American peers. Right now we are witnessing a degree of competitive innovation that we haven’t seen since the space race. I don’t know who will win or if there will be a single winner, but I can say with confidence that China is taking AI very seriously and should not be underestimated,” says Breyer.
China has an inherent advantage when it comes to data, the fuel on which an AI engine runs. There is no other country that comes close. Its mobile users are almost 3x that of the US and an internet base of nearly 750 million users who are willing to experiment and adopt new technology in mobile payments or e-commerce. It has a huge pool of engineers who are pioneering AI research at their universities and more importantly, a government that is not afraid to use the huge amounts of data its citizens generate every day, for the advancement of AI. This data is used by AI systems to improve the accuracy of its predictions and become more efficient. “Access to data is a huge advantage because to train your machine, you need a large data set. Paradoxically in India, it is the large tech giants that own these data sets rather than start-ups. So, they have a better chance to work with local partners. For instance, IBM can walk into Manipal Hospitals Group and show them what they have done and get access to oncology reports because Manipal will always trust an IBM over a start-up with their data,” says Parag Dhol, managing director at Inventus Capital Partners. Manipal has deployed IBM intelligence for cancer research at six of its locations. Similarly, Google has tied up with Aravind Eye Hospitals, Sankara Nethralaya and Narayana Nethralaya to train AI with photographs of retinas. India is the diabetes capital of the world with over 70 million suffering from it and over 10% of them lose their eyesight owing to the disease. The algorithm achieved an accuracy of 98.6% in detecting diabetic retinopathy, on par with the best ophthalmologists and retina specialists. The AI, which can throw up the results in seconds, is now being used by Aravind Eye Care to assist manual grading that takes hours, and will eventually replace the manual grading system.
But for some Indian start-ups such as Uniphore having Amazon, Apple and Google in their space has worked well for them. Uniphore provides speech analytics, a virtual assistant and portable voice biometric system for its enterprise clients. “Since Amazon, Apple and Google are making significant investments in speech recognition, it brings a lot of awareness to the space. A lot of enterprises are now looking to add a voice layer to their applications. Enterprises usually follow where consumers go,” says Umesh Sachdev, CEO, Uniphore. He believes all the efforts of the biggies in the area of speech recognition have opened up a $450 billion-$500 billion enterprise market for Uniphore. “We can disrupt the entire call centre industry with our solutions,” he says. Uniphore counts Cisco’s chairman John Chambers, Kris Gopalakrishnan and IDG Ventures among its investors and is one of the earliest start-ups in India to build a deep tech model in speech recognition. “It is all moving from touch to speech,” says Sateesh Andra, managing director, Endiya Partners. “Indian start-ups will do well if they are able to use the technology to solve an existing problem. They can leverage AI not only to build applications for the local market but can also take the product global.” He gives the example of his portfolio company SigTuple, which is addressing a global problem — a shortage of pathologists — with its AI solution. India has one pathologist per 65,000 people and in China, the number is worse with one pathologist for every 130,000. Even in the US there are only 5.8 pathologists for every 100,000. A shortage of pathologists means you have an overworked pathologist or a half-baked technician analysing the slides and there is a good chance they could miss something during diagnosis. SigTuple has developed a low-cost machine that digitises blood and pathology samples by attaching smartphone and mechanical components to a regular microscope. The AI software automatically scans and analyses blood smears as well as other biological samples, tags the visual data and churns out the results in minutes. It also provides visual evidence to easily verify the result. Since it is addressing a larger problem across countries, the company plans to go global and is planning to work towards an USFDA approval.
Distributed gains
Just like AI, blockchain, which was initially created for crypto-currencies, is now being used in healthcare to bring patients, insurers and providers together. With a low-cost and decentralised ledger approach to managing information, blockchain gives all the parties involved access to a single database of encrypted data. It creates an audit trail each time data is changed, helping to ensure the integrity of the information. “Blockchain technology will substantially impact healthcare. I think that value will initially accrue in administrative areas and record keeping before it reaches patient health information. Interesting opportunities exist in provider credentialing and pharmaceutical supply chain management. We could also see blockchain play a role in genomics marketplaces,” says Breyer. Blockchain is also making its way into the world of financial services.
“Block chain can play a very sophisticated role where there is a legal document or when a contract crosses boundaries because there are multiple parties involved. So blockchain can play a very important role in trade finance and lending based on data. It also opens up the market by bringing in the untrusted players in a trusted manner,” says Shekhar Kirani, partner, Accel Partners. Infosys recently set up India Trade Connect, a blockchain-based trade network in India, with seven private banks including Axis Bank, ICICI Bank, RBL Bank, and Kotak Mahindra Bank. The partner banks can take better informed lending decisions since all of them are aware of the exact position of a bill as well as the partner bank’s exposure to a particular company.
A new kind of drive
When Ford unveiled its Model T, the first car to be mass-produced, the world that we lived in changed forever. A century later, self-driving or autonomous cars are all set to do the same. Waymo, Alphabet’s subsidiary, is ahead in the business of driverless cars and is expected to commercially launch its driverless fleet service sometime this year. SoftBank has invested $2.25 billion from its Vision Fund for a 20% stake in GM’s autonomous driving arm, GM Cruise Holdings, which is preparing for a commercial launch in 2019. Connected cars will communicate with one another to avoid accidents and traffic jams, and riders will be able to spend commuting time focusing on work and other tasks. They will also play a huge role in driving up the revenue of ride hailing business, which is expected to grow from $5 billion now to $285 billion by 2030 as their driverless fleet increases. While autonomous cars are still some time away in India, it offers Indian start-ups opportunities to come up with applications that improve the overall safety of autonomous cars or assisted driving. Take the case of Netradyne, which provides vision-based analytics through its device – Driveri – for fleet management, automotive, security and surveillance industries. From tailgating and lane detection to speed and time to collision, the device tracks, records and analyses a whole lot of data. The device also alerts the drivers in real-time. With most of clients in the US, the company is looking at working with OEMs and fleet owners in the domestic market.
Automated future
A lot of advancement in AI is on account of the huge amount of data generated by connected devices, made possible by the internet of things (IoT). As the number of connected devices and industrial machines increase, the global IoT market is expected to exceed $1.29 trillion by 2020, while the Indian IoT market is poised to touch $15 billion during the same period. The Bengaluru-based Stellapps leverages IoT rather innovatively to improve the supply chain of dairy farms including milk production, procurement, cold chain and farmer payments. Their IoT router and IoT controller collects data through the sensors that are embedded in milking systems, animal wearables and milk chilling systems, which is sent to the cloud and the analytics platform then crunches the data and sends back the analysis to its customers’ mobiles. Zenatix, on the other hand, uses advanced machine learning models to provide significant energy savings for large commercial consumers of electricity. It uses sensors for monitoring energy, temperature, humidity and other parameters, helping retail and BFSI businesses save up to 30% on their energy bills.
Whether IoT, blockchain or the all-pervasive AI, each of these technologies have the ability to help industries such as manufacturing, healthcare and financial services to not only break new barriers but also allow start-ups across the world to redefine the way business is done. The good news is that this is just the beginning and the best is yet to come.
