According to management consulting firm Zinnov, there are around 170 AI startups in India. These enterprises have received a total funding of $ 36 million. Bengaluru leads the list with 64 startups based there. “In the last five years computing capability and capacity has improved significantly, aiding in the growth of AI startups,” says Hardik Tiwari, engagement lead, Zinnov.
According to him, there are three parts to AI: infrastructure, platform and applications. The startups in India focus on the third, that is, building AI applications.
AI, the science of making computers think and act like humans, is behind the buzz about automation, often discussed with trepidation. Given that AI is essentially machine learning that drives driver less cars and other machines that are steadily taking over human tasks, there is a gnawing fear that jobs will be usurped. The backers of AI say no doubt it will take away the mundane, repetitive jobs, but it will end up creating new ones.
The goal of AI startups is not to replace jobs, but “improve the way humans do their jobs,” says Talhagato Rai Dastidar, founder, SigTuple, a Bengaluru-based AI startup. SigTuple was launched in July 2015 and received seed funding from Flipkart founders Sachin Bansal and Binny Bansal. The 17-people startup has an ambitious goal to have all aspects of healthcare diagnostics such as blood tests and X-ray scans interpreted by AI tools.
In India, several Indian startups are at the leading edge of AI applications.Mumbai-based Arya.ai was the first AI startup from India to be selected by Paris&Co, a French innovation agency as one of the 21 global companies that do standout innovations. Founded by two IIT Bombay researchers, Arya’s goal to create systems that will automate task execution.
“A I is a technology mega trend,” says Ashootosh Chand , partner, PricewaterhouseCoopers India, a consultancy. He points to a recent World Bank report that suggested that of all the jobs being done in India, 69% could be done by machines. “That’s AI systems. The market will be huge. Self-driving cars are powered by AI systems. Anything that is repetitive will be done by AI.”
Smart Use of Data
Indian enterprises have turned out to be a breeding ground for AI startups because they are in businesses with a surfeit of data. The common thread running between healthcare e-commerce companies, oil refineries and airlines is that they thrive on a proliferation of data. A large e-commerce company like Amazon India or Flipkart, for example, collects about two terabyte of data daily. A F1 car generates 1.2 gb of data per second. Seema Kumar, country leader, developer ecosystem & startups, IBM India says startups apply AI to convert data into information and scale that up to knowledge to eventually create cognitive systems. “This is relevant across domains and industries.”
By 2018, more than 50% of the applications built will be cognitive, says research firm IDC.
In India, IBM works with AI startups in areas including healthcare, finance and education. For example, IBM is work ing with Signzy, an AI startup focused on verifying digital documents like eKYC (Know You r Customer) . I n Ju ne, Signzy won Reserve Bank of India’s payment system innovation contest. IBM also works with TalView, which uses speech recognition to gauge personality of applicants.
Narendra Bhandari, Microsoft says the core of AI is fuelled by data and the goal is to get to human level of accuracy. “AI gives life to intelligence and that’s what the startups in the space aim to do.“
For instance, ClearTax, an e-filing startup, uses AI for document processing. ClearTax dosen’t have humans doing data entry -it’s done by machines.In 2011, the accuracy was 50-55% and now it’s 95-98% and it processes about one million tax returns.
Other e-commerce companies such as Amazon, Flipkart and Paytm rely on AI tools to check fraud, improve conversion rates, customise selection and get a better hang on what the customer wants.
“We are just beginning to see the powerful impact of machine learning across industries, from fraud protection and customer service to instant translation via speech and text and energy efficiency in data centers,“ says a Google spokesperson.
Machines that See and Talk
Bengaluru-based AIndra Systems works in the domain of computer vision to solve problems like facial recognition and screening and detection of cervical cancer.The 13-people company has been bootstrapped by founders. Abhishek Mishra, co-founder AIndra Systems says, “The more data we get, the better the system gets. We see and understand patterns to improve the algorithm.”
For example, to enable error free facial recognition, the more pictures the system sees the better it gets–like a selflearning system. Facebook uses facial recognition to help people tag users and their friends in photos they upload. To improve outcomes in cervical cancer AIndra has collaborated with Kidwai Hospital in Bengaluru to provide patient records.
For facial recognition, AIndra has four customers, including the Skill Council of India.
Chennai based Mad Street D en , launched in 2013, is developing AI solutions for the retail industry. Funded by Sequoia Capital Mad Street Den “teaches machines how to see”, says co -founder Ashwani Asokan. T he goal of the visual-intelligence based platform is to build an AI tool, for both online and off line use that can understand users preferences. Its team of 50 is spread across India, Europe and the US and comprises data scientists, neuro scientists, computer vision experts, marketers and designers. Mad Street Den’s counts TataCliq.com and Craftsvilla.com -online fashion and handicraft retailers respectively as its customers.
Niki.ai, a startup funded by Ratan Tata, aims to build a machine users can talk to and pay bills–simply shoot `pay my Airtel bill’ and the system will check the bill amount and make the payment via mobile wallet, debit card or credit card. The user just has to feed in the information one time and the rest is as easy as giving a command to the Niki.ai app.
Sachin Jaiswal, co-founder, Niki.ai says, “the 4MB app is a concierge that helps users book hotel rooms, pay bills, book tickets and so on. The key is understanding human language and patterns.” Niki.ai says it has 50 partners, including Burger King, MobiKwik, Ticket Goose and Clear Trip.
Like elsewhere, much of the work that AI startups in India do at present is basic. “AI is at super early stage. It’s like creating a new category of work,” says Asokan.
Jaiswal of Niki.ai dismisses Siri and Cortana–digital assistants from Apple and Microsoft-as play tools. “We have to bring AI to mainstream.”
There are two main challenges that AI faces–lack of data (the proliferation is still not qualitative enough) and lack of talent. The former is needed to improve systems–the more they (AI systems) see the more they learn and better they become. For instance, an automated system that interprets CT Scans must see millions of scans to understand and interpret patterns. “There’s lack of richness of data to make a system cognitive,” says Kumar of IBM.
The second issue, talent, will get tackled as the esoteric area expands. But at present, startups find hard to get machine learning experts.
The twin challenges haven’t dented the enthusiasm of AI geeks. Jaiswal says in a few years, all existing web interfaces will go away and be replaced by AI. “We will be the gateway to the internet.”
Source – The Economic Times (Delhi)