Asia is leading the pack in AI business deployment compared to less than a third for US companies. The adoption rate in the rest of the world remains low, as firms do not understand the deployment of AI¹ in their operations.
The surveillance behavior of Chinese firms continues and contravenes privacy. MIT’s decision to end its collaboration with iFlytek¹⁰ from China makes sense and will set the trend for other companies. Artificial intelligence does not have to hurt people but rather be ethical, responsible, and accountable.
AI investments continue to increase as businesses understand the value of automation. Digitization, according to a Pew Research Poll, will have positive outcomes for companies that embrace digital transformation². Voice-enabled programs from artificial intelligence are supporting customer service as most companies handle customer queries in real-time.
The COVID-19 pandemic is forcing corporate organizations to adopt remote working models in their operations. Companies are using Zoom to engage with teams as the COVID-19 pandemic continues. Online social interactions have shot up as people communicate because of social distancing measures.
Personalized Prediction to Combat COVID-19
The world has experienced COVID19 outbreaks over the past few months. An initial phase with few infections followed by a take-off of the famous epidemic curve accompanied by a country-wide lockdown. This leads to the governments having to address what President Trump has called “the biggest decision” of his life.
Battling COVID-19 based on the technology of personalized prediction is an alternative approach for policymakers to consider. Data-driven firms³ make personalized recommendations using ML and artificial intelligence technology.
Ming Zeng, Former chief strategy officer, Alibaba¹¹, described how Ant Financial, his company’s small business lending operation, can assess loan applicants in real-time by analyzing their transaction and communications data on Alibaba’s e-commerce platforms in a recent Havard Business Review article.
The same approach could work for COVID-19 pandemics. Multiple sources of data would be used, and machine learning models would be trained to measure an individual’s clinical risk of suffering from COVID: what is the probability they will need intensive care, for which there are limited resources? How likely is it that they will die? There is a great deal that modern data science and AI could mitigate the fallout from this pandemic.
Smaller, Greener Neural Networks
Artificial intelligence has some major sustainability⁴ issues, along with a focus on certain ethical concerns.
Researchers at the University of Massachusetts at Amherst released a startling report estimating that the amount of power required for training and searching a certain neural network architecture involves the emissions of roughly 626,000 pounds of carbon dioxide, which is equivalent to nearly five times the lifetime emissions of the average US car.
Deep neural networks need to be deployed on diverse hardware platforms makes the issue even more severe.
A new automated AI system developed by MIT researchers results cut down the pounds of carbon emissions by improving computational efficiency. The system, once-for-all network, trains one large neural network comprising many pre-trained subnetworks of different sizes that can be tailored to diverse hardware platforms without retraining.
It reduces the energy usually required to train each specialized neural network for new platforms. The process requires roughly 1/1,300 the carbon emissions compared to today’s state-of-the-art neural architecture search approaches.
Song Han, an assistant professor in the Department of Electrical Engineering and Computer Science, says, “The aim is smaller, greener neural networks,”
Learning-based approach for Chip Design
GoogleAI’s lead Jeff Dean, Scientists at Google Research, and the Google chip implementation and infrastructure team to describe a learning-based approach to chip design⁵that can learn from past experience and improve over time, becoming better at generating architectures for unseen components.
It completes designs in under six hours on average, which is significantly faster than the weeks it takes human experts in the loop as they claimed.
It builds upon a technique proposed by Google engineers in a paper published in March. It also advances the state of the art in that the placement of on-chip transistors can be largely automated. The Google technique could enable cash-strapped startups to develop their own chips for AI and other specialized purposes if it is publicly available.
Dean told VentureBeat in an interview late last year, “Basically, right now in the design process, you have design tools that can help do some layout, but you have human placement, and routing experts work with those design tools to kind of iterate many, many times over.”
A framework devised by the researchers directs an agent trained through reinforcement learning to optimize chip placements. It achieved superior PPA on in-production Google tensor processing units as per their claim.
Surveillance behaviors by Chinese Tech Firms
MIT cuts ties with iFlytek, a Chinese artificial intelligence company accused of supplying technology for surveilling⁶ Muslims in the northwestern province of Xinjiang.
It has terminated the relationship in February after reviewing an upcoming project under tightened guidelines governing funding from companies in China, Russia, and Saudi Arabia.
Maria Zuber, Vice President of Research at MIT says,
“We take very seriously concerns about national security and economic security threats from China and other countries, and human rights issues,”
Companies in the United States and universities have built ties with Chinese tech firms in recent years. The relationships between the two countries have soured and come under increasing scrutiny.
The US government banned six Chinese AI companies, including iFlytek in October 2019, from doing business with American firms for reportedly supplying technology used to oppress minority Uighurs in Xinjiang. Human Rights Watch claimed iFlytek supplied police departments in Xinjiang with technology for identifying people using their voiceprints in 2017.
Jiang Tao, a senior VP at iFlytek, says, “We are particularly sorry about this,” US officials are increasingly wary of Chinese companies developing advanced technologies.
AI-based Chest X-ray System
Agrowing number of hospitals faced with staff shortages and overwhelming patient loads turning to AI tools to help them manage the pandemic. Patients often had to wait six hours or more for a specialist to look at their x-rays in the Royal Bolton Hospital, run by the UK’s National Health Service (NHS).
If an AI-based tool could get an initial reading that would dramatically shrink that wait time for an emergency room doctor. A specialist could follow up on the AI system’s reading with a more thorough diagnosis later.
Rizwan Malik, the lead radiologist, Royal Bolton Hospital, identified a promising AI-based chest x-ray system⁷called qXR from the Mumbai-based company Qure. He proposed the x-ray system, and the proposal was finally approved after four months of reviews from multiple hospital and NHS committees and forums.
COVID19 hit the UK before the trial could kick-off. The system began as a pet interest suddenly looked like a blessing. Chest x-rays had become one of the fastest and most affordable ways for doctors to triage patients with shortages & delays in PCR tests.
Qure retooled qXR to detect COVID-induced pneumonia, and Malik proposed a new clinical trial, pushing for the technology to perform initial readings rather than just double-check human ones.
Asia Leading in AI Business Deployment
AI technology is still at the fringes of business processes despite more than half of organizations in the Asia Pacific region that has deployed AI in its operations.
MIT Technology Review Insights thought leadership program “The #globalAI agenda,” examines how organizations are using AI today ⁸ and planning for the future.
A global survey conducted in January and February 2020 featuring 1,004 AI experts explores AI adoption, leading use cases, benefits, and challenges, and seeks to understand how organizations might share data with each other to develop new business models, products, and services in the years ahead.
How do executives see AI playing out in their business within Asia-Pacific? What are the main use cases and challenges do they face in AI deployment?
56% of Asian respondents to the survey had deployed AI in their operations by 2017, compared with less than a third of North American respondents and 35% on average across other regions. Nearly 96% of Asian respondents reported AI deployments, above the 85% average of respondents from the other regions by 2019. Nearly half of survey respondents expect AI to be used in between 21% and 30% of business processes in three years’ time.
Asia sees greater success than other regions within these areas.
AI Identifying vulnerable People
Maccabi Healthcare Services using artificial intelligence to help identify which of the 2.4 million people are most at risk of severe COVID19 complications.
The system developed with AI company @Medial EarlySign¹² has already flagged 2% of its members, amounting to around 40,000 people, individuals are identified and put on a fast track for testing.
The AI was adapted from an existing system trained to identify people most at risk from the flu, using millions of records from Maccabi going back 27 years.
The system draws on a range of medical data, including a person’s age, BMI, health conditions such as heart disease or diabetes, and previous history of hospital admissions to make its predictions. The AI can spot at-risk individuals⁹who might have been missed otherwise.
Whether people that Maccabi flags should be cared for at home, put up in a quarantined hotel, or admitted to hospital can be determined by the use of AI. Major US healthcare providers that are interested in using AI to fast-track their own high-risk patients are discussing with the organization.
Darren Schulte, an MD, and CEO of AI firm Apixio¹³, which develops software to analyze unstructured medical data, says, “Using AI to identify vulnerable people could save lives.”
AI and Automation Investments
Over the past decade, artificial intelligence and automation have been two hot areas of investment. The need for automation, technology, and tools continues to grow as the worldwide workforce increasingly shifts to a remote workforce. Venture capitalists are investing in firms significantly interested in automation and intelligent systems.
There is still no commonly accepted definition of AI despite the fact that it has been around for decades. Some technology firms are focused on how artificial intelligence can better help them manage funds, and other companies are more interested in how AI can supplement their human workforce. Artificial intelligence can help with various tasks that investors need to look at when making investments.
Applications of robotics, particularly to manufacturing and specifically the concept of collaborative robots, are appealing. Collaborative robots can be used to work along with employees.
AI onboard makes the arm easier to use it and a suite of tools to enable anyone to operate the arm without technological training. An iPad or similar device can be used to train the arm through movement on how to carry out tasks. This arm is able to work side by side with humans and falls under the category of collaborative robots.
Drug Treatment Research for COVID 19
The @Ragon Institute of MGH, MIT, and Harvard; The Researchers and the Broad Institute of MIT and Harvard, along with colleagues from around the world, have identified specific types of cells that appear to be targets of the coronavirus that is causing the Covid19 pandemic.
The researchers were able to search for cells that express the two proteins that help the sarscov2 virus enters human cells using existing data on the RNA found in different types of cells.
The researchers found subsets of cells in the lung, the nasal passages, and the intestine that express RNA for both of these proteins much more than other cells.
The researchers hope that their findings will guide scientists who are working on developing new drug treatments or testing existing drugs that could be repurposed for treating Covid-19.
Alex K. Shalek, a core member of MIT’s Institute for Medical Engineering and Science (IMES), and extramural member of the Koch Institute for Integrative Cancer Research, an associate member of the Ragon Institute, and an institute member at the Broad Institute says, “Our goal is to get information out to the community and to share data as soon as is humanly possible so that we can help accelerate ongoing efforts in the scientific and medical communities.”
Voice-enabled Programs and Artificial Intelligence
We grapple with remote workforces and social distancing as the world is undergoing significant change. Education is certainly going through a massive transformation as educators and students transition to learning remotely.
One campus was ahead of the revolution on using AI to augment the learning experience before all this happened.
The future of AI technology and the implementation of voice-enabled programs on the Arizona State University (ASU) campus discussed on an episode of the AI Today Podcast with John Rome, Deputy CIO and Voice Evangelist at Arizona State University.
He explained that he spends his days working on analytics and ensuring that students and staff have access to the technology they need to function. He also advocates for the adoption of voice-enabled programs and artificial intelligence to be implemented on the college campus. Rome is an expert on the developments in this relatively new market of technology as “VoiceEvangelist.”
Rome decided to leverage emerging voice assistant technology to transform the student experience on the ASU campus as part of his role. A group of first-year students was selected to have the opportunity to live in a voice-enabled resident hall that received Amazon Echo Dot devices.
Corporates adopting Remote Working
As online communication and collaboration tools gain traction, the COVID19 crisis has thrust the issue of remote working to the top of corporate agendas globally.
Companies already operated remote “distributed” workforces — WordPress parent Automattic has long embraced remote working, as have GitLab, GitHub, and Basecamp. Payments giant Stripe¹⁴ launched a new remote engineering hub to help it access a bigger global pool of tech talent.
There is a terminal, which helps startups, build-remote engineering teams, by doing all the heavy lifting for them. US companies can use to tap technical talent abroad as terminal operates strategically placed engineering hubs.
Many of terminal’s clients already operate remotely but use its service to sidestep the logistical headaches that come with hiring in other countries, including navigating local legal and tax structures, running recruitment processes, and operating essential services such as payroll and HR, so the physical hubs are only part of the story.
Terminal counts several tech-focused campuses throughout Canada and Mexico, has two corporate hubs in San Francisco and New York. Terminal announced its first acquisition, snapping up Austin, Texas-based AI-powered recruitment startup @Roikoi¹⁵ this week.
The Monte Carlo Tree Search
Scientists at Alphabet’s DeepMind propose a new framework in a paper published on the preprint server Arxiv that learns an approximate best response to players within games of many kinds.
Games are a convenient proving ground to develop algorithms that can be translated into the real world to work on challenging problems asserted by DeepMind CEO Demis Hassabis.
The framework could lay the groundwork for artificial general intelligence (AGI), which is the holy grail of AI — decision-making AI system that automatically completes tasks like data entry, but which reasons about its environment. That’s the long-term goal of other research institutions, like OpenAI.
Exploitability is the level of performance against players. For example, one variant of Texas Hold’em — Heads-Up Limit Texas Hold’em — has roughly 1014 decision points, while Go has approximately 10170. One way to get around this is with a policy that can exploit a player to be evaluated, using reinforcement learning — an AI training technique that spurs software agents to complete goals via a system rewards.
The framework is called Approximate Best Response Information State Monte Carlo Tree Search.