AI-powered filter app Prisma wants to sell its tech to other companies

Prisma, the Russian company best known for its AI-powered photo filters, is shifting to B2B. The company won’t retire its popular app, but says in the future, it will focus on selling machine vision tools to other tech firms.

“We see big opportunities in deep learning and communication,” Prisma CEO and co-founder Alexey Moiseenkov told The Verge. “We feel that a lot of companies need expertise in this area. Even Google is buying companies for computer vision. We can help companies put machine vision in their app because we understand how to implement the technology.” The firm has launched a new website — prismalabs.ai — in order to promote these services.

Prisma will offer a number of off-the-shelf vision tools, including segmentation (separating the foreground of a photo from the background), face mapping, and both scene and object recognition. The company’s expertise is getting these sorts of systems — powered by neural networks — to run locally on-device. This can be a tricky task, but avoiding using the cloud to power these services can result in apps that are faster, more secure, and less of a drain on phone and tablet battery life.

Although Prisma’s painting-inspired filters were all the rage last year (the app itself was released in June 2015), they were soon copied by the likes of Facebook, which might account for the Russian company’s change in direction.

Moiseenkov denies this is the case, and says it wasn’t his intention to compete with bigger social networks. “We never thought we were a competitor of Facebook — we’re a small startup, with a small budget,” he said. But, he says, the popularity of these deep learning filters shows there are plenty of consumer applications for the latest machine vision tech.

Moiseenkov says his company will continue to support the Prisma app, and that it will act as a showcase for the firm’s latest experiments. He says the app still has between 5 million and 10 million monthly active users, most of which are based in the US. The company also started experimenting with selling sponsored filters on its main app last year, and says it will continue to do so. It also launched an app for turning selfies into chat stickers.

There have been rumors that Prisma would get bought out by a bigger company. Moiseenkov visited Facebook’s headquarters last year, and the US tech giant has made similar acquisitions in the past — buying Belarus facial filter startup MSQRD in March 2016. When asked if the company would consider a similar deal, co-founder Aram Airapetyan replied over email: “We want to go on doing what we do and what we can do best. The whole team is super motivated and passionately committed to what we do

Source:-theverge

Education in 2020s: your AI-powered teacher will take your attendance and predict your grades

The steady march of technology is making some jobs obsolete. While reskilling is one solution, can AI be used to change the way students learn and make reskilling easier for future generations?

Andrew NG, Co-founder of Coursera and former chief scientist at Baidu recently likened Artificial Intelligence (AI) to ‘electricity of the 21st century’. He said at a Wall Street Journal event,

Whatever industry you work in, AI will probably transform it, just as 100 years ago the rise of electricity transformed industry after industry. So, I hope that whatever industry you’re in, you’ll figure out how to leverage AI, because I think it will create new winners and losers in almost every category.

What about the new generation of students just entering academia? There is a great need to prepare them for this new, ‘AI-first world’, as Google CEO Sundar Pichai calls it. Academic courses though haven’t been able to keep pace with some of the real-world needs. What is the best way to transform education? Through AI, of course!

Anshul Bhagi and Nikhila Ravi, computer scientists and education entrepreneurs who met as graduate students at Harvard University, recently co-founded OpenEd.ai, a non-profit to develop and promote open-source AI for education. Here is their story.

Taking AI from the dorms of Harvard University to IIT Delhi and beyond

Anshul, a Computer Science graduate from Massachusetts Institute of Technology and an MBA from Harvard, founded and runs Camp K12, an education startup bringing coding to K-12 education in India. Nikhila on the other hand is an alumna of University of Cambridge and Harvard University (ML/AI), and is now an engineer at Facebook.

Talking about how OpenEd.ai came about, Anshul says,

Throughout these projects, they found open-source code, pre-trained models, datasets and tutorials created by the likes of Denny Britz (WildML.com) and Andrej Karpathy to be of tremendous value. They came to the conclusion that given the steep learning curve ML often requires, and the fast pace at which the field moves within academic/research silos, the ML community will continue to benefit immensely from open-source initiatives that bridge learning and informational gaps.

With a long-term goal of democratising AI in education, Nikhila and Anshul decided to set up OpenEd.ai, a non-profit dedicated to developing and promoting open-source AI for education.

The team then added Shreyas Deshmukh, a Microsoft Product Manager and computer scientist/AI developer, to the organising committee for OpenEd.ai’s inaugural event, and together they orchestrated the AI for Education HackWeek that is currently in progress (July 28-Aug 11). The first offline leg of the event took place at IIT Delhi from from July 28 to August 4.

During the global HackWeek, OpeEd.ai offered free computing resources and research resources for all participants courtesy Amazon, IBM Watson, and Digital Ocean; special online workshops courtesy Google, IBM Watson, and Amazon. OpenEd.ai is also offering $19,000 in prizes, including a $6,000 grand prize for the best solution, which will be evaluated by Harvard CS faculty and members of the global ML community.

Amitabh Kant, CEO of NITI Aayog, in his keynote address, talked about the challenges in the Indian education system and how technologies like AI could help the country leapfrog. He also announced two additional challenges for the participants, which are also open to the general public.


Related read: Can you solve these ‘AI in education’ challenges? You may get to present the solution to government officials


Here is an overview of some of the other workshops at the event.

Can AI lead to better learning outcomes?

Vishal Dixit and Sarvesh Kanodia from Omidyar Network, a philanthropic investment firm established by eBay founder Pierre Omidyar and his wife Pam, talked about unsolved problems and startup opportunities in education AI. Omidyar Network has committed more than $1 billion to for-profit and non-profit organisations. Some of their bets in the education space include ventures like like Code.org, Khan Academy, AltSchool, Teach for India, and Aspiring Minds.

Vishal remarked that every child should have individualised access to education. The challenge is to go from self-paced e-learning solutions to more adaptive and conversational tutoring methodologies. He said,

AI is the process and not the product…. The end game has to be better learning outcomes. I don’t believe that teachers will be redundant because of AI, it will instead act as force multipliers for teachers.

Roopa Kudva, Partner and Managing Director for India at Omidyar Network, in a statement said,

Our support to OpenEd.ai HackWeek is aligned with our commitment to supporting innovations in technology. This is an attractive platform for developers to learn from experts working in the cutting edge of AI and leverage their skills towards solving some of the key problems in education.

Man and machine, power extreme

Danish Contractor from IBM Research gave the audience an overview of how IBM is leveraging AI in education to predict future performance and aid childhood learning. He said that by gathering data from schools from Class I to VIII and studying the patterns, they are now able to predict with great accuracy what a grade 4 student would be likely to achieve by grade 8, assuming the same learning trajectory.

Danish also spoke about how learning content analytics and cognitive career counselors could help decrease the burden on human teachers. With Watson’s cognitive solutions, IBM aims to transform the learning experience through personalisation, and help educators gain insights into learning styles, preferences, and aptitude of every student.

Talking about the complementary and symbiotic role of AI in education, Sriram Raghavan, Director, IBM India Research & CTO, IBM India/South Asia, in a statement said,

We see AI as Intelligence Augmentation (IA) — cognitive systems that work together with humans to enhance (rather than replace) human capabilities. Our goal is to develop technology that works seamlessly with teachers and students and uses AI-driven insights to assist human stakeholders in performing their tasks better — whether it is pedagogy, learning, or administration.

TensorFlow and the basics of machine learning

Manoranjan Padhy and Rohit Gupta from Google gave the audience an overview of artificial intelligence and machine learning. Starting with the basic terminologies used in machine learning they went on to explain the different categories in machine learning, which include supervised, unsupervised, semi-supervised, and reinforcement learning.

With most of the explaining done, they then introduced the audience to TensorFlow, an open-source software library developed by Google for Machine Intelligence and took them through a few demo challenges.

Six challenges to hack away at…

After the workshops concluded, participants had the option of working on any of the challenges put forth, including two by the Government of India, three by IBM, and one wildcard challenge. It read as:

Apply your knowledge of CS, Data Science, Machine Learning, and NLP to build open-source AI for education, solve any problem in education you’re passionate about.

A participant’s output can be any of the following:
* Usable products (e.g. chatbots, web apps, mobile apps).
* Data sets (eg scrape public sources to collect data useful for future AI work in education).
* Tutorials and how-to articles for AI-education work you’ve previously done.

Submissions are open until 11.59 pm on Friday, August 11. Learn more about all six challenges and submit your hack here at OpenEd.ai.

Source:-Yourstory