Solving Intelligence, Solving Real-world Problems
As a smart technology entrepreneur with a machine intelligence research background and passionate about advancing the state-of-the-art in machine or artificial intelligence (AI) to help solve real-world problems, it is very encouraging and exciting to see the AI buzz in the tech industry right now, the progress made in the field to create an even stronger intelligence toolbox, and the ever increasing practical applications in all industries and enterprise functions. Machine intelligence is not only changing the way we use our computers and smartphones but the way we interact with the real world. It is also one of the key exponential technologies in the Fourth Industrial Revolution. Given how all the major technology companies are embracing machine intelligence as a core part of their business and the multitude of startups building their business on this technology, AI is clearly not a passing fad, but being pushed across the rest of the tech world too.
In this post I’m not only addressing some key topics about the current and future state of machine intelligence, but also practical steps we are taking here in Africa to not only use smart technology to solve problems, but also make a contribution towards advancing the-state-of-the-art in machine intelligence.
- Accurate Understanding of Reality with respect to Machine Intelligence
- Steering Machine Intelligence towards positive outcomes
- Advancing the state-of-the-art in Machine Intelligence
- The Machine Intelligence Landscape
- Cortex Logic in partnership with MIIA to advance the state-of-the-art and solve real-world problems in Africa
- Focus areas in advancing the state-of-the-art in Machine Intelligence
- Intelligent Virtual Assistants and Advisors
- Machine Intelligence Institute of Africa (MIIA)
Accurate Understanding of Reality with respect to the Machine Intelligence
“Truth, an accurate understanding of reality, is an essential foundation for producing good outcomes” – Ray Dalio, Bridgewater Associates
I’ve used this quote from Ray Dalio (founder of the world’s largest hedge fund, also with its own AI team) in a number of my talks on Machine Intelligence and the Fourth Industrial Revolution. We need to be realistic about what machine intelligence can do right now. That’s why I like a recent article by Andrew Ng, chief scientist at Baidu, about what AI can and can’t do right now. Understanding what AI can do and how it fits into a company’s strategy is only the beginning of the process. In today’s world of open source, the scares resources is indeed data and talent. As he correctly states, although AI is clearly transforming web search, advertising, e-commerce, finance, logistics, media, and most industries, today’s applications are predominantly using a form of supervised learning, with deep learning the most powerful, but also ‘data hungry’ technique in the machine intelligence toolbox. There are many areas of machine intelligence where breakthroughs are sorely needed. Unsupervised learning, for example, is to a large extent still a mystery – we currently only have limited forms focused on clustering and dimensionality reduction. Unsupervised learning should for example build a causal understanding of the sensory space with temporal correlations of concurrent and sequential sensory signals. Other powerful techniques include supervised recurrent neural networks (RNNs) and reinforcement learning RNNs. Having done my Ph.D. in RNNs, there is a clear need for improved supervised and unsupervised training algorithms. Going forward I expect massive RNNs on dedicated hardware to simultaneously perceive and analyze an huge number of multimodal data streams from many sources, learning to correlate all those inputs and use the extracted information in a range of commercial and non-commercial applications. Although Deep Mind has made significant progress with deep reinforcement learning (e.g. first computer program beating a top-level Go player and improvements with reinforcement learning with unsupervised auxiliary tasks), there are still so much to do in pushing the boundaries of AI, including the first serious applications of reinforcement learning RNNs to real world robots.
Steering Machine Intelligence towards positive outcomes
Can we advance machine intelligence in a way that is most likely going to benefit humanity as a whole and help solve some of our most pressing real-world problems? Can we shape machine intelligence to be an extension of individual human wills and as broadly and evenly distributed as possible? If society approaches machine intelligence with an open mind, the technologies emerging from the field could profoundly transform society for the better in the coming decades. Like other technologies, AI has the potential to be used for good or criminal purposes. A robust and knowledgeable debate about how to best steer machine intelligence in ways that enrich our lives and our society is an urgent and vital need. It is incumbent on all of us to make sure we are building a world in which every individual has an opportunity to thrive.
As mentioned in a post about how the future of AI will impact our everyday life the following were listed: automated transportation, cyborg technology, taking over dangerous jobs, solving climate change, robots as friends, and improved elder care. Stanford University’s The One Hundred Year Study on Artificial Intelligence highlights substantial increases in the future uses of AI applications, including more self-driving cars, healthcare diagnostics and targeted treatment, and physical assistance for elder care. Though quality education will always require active engagement by human teachers, machine intelligence promises to enhance education at all levels, especially by providing personalization at scale (I’m currently also involved in a digital education startup called The Student Hub with its innovative ERAOnline eLearning platform that enables personalization, which can be further enhanced with AI technology as it evolves – see Education section of previous LinkedIn post for more details). AI will increasingly enable entertainment that is more interactive, personalized, and engaging. Research should be directed toward understanding how to leverage these attributes for individuals’ and society’s benefit. With targeted incentives and funding priorities, machine intelligence could help address the needs of low resource communities. In the longer term, machine intelligence may be thought of as a radically different mechanism for wealth creation in which everyone should be entitled to a portion of the world’s AI-produced treasures. The measure of success for AI applications is the value they create for human lives. Misunderstandings about what AI is and is not could fuel opposition to technologies with the potential to benefit everyone. Poorly informed regulation that stifles innovation would be a terrible mistake. Going forward, the ease with which people use and adapt to AI applications will likewise largely determine their success. Society is now at a crucial juncture in determining how to deploy AI-based technologies in ways that promote rather than hinder democratic values such as freedom, equality, and transparency. Machine intelligence already pervade our lives and will likely replace tasks rather than jobs in the near term, and will also create new kinds of jobs. However, the new jobs that will emerge are harder to imagine in advance than the existing jobs that will likely be lost.
Advancing the state-of-the-art in Machine Intelligence
Apart from some of the many universities with advanced AI programs and research such as Stanford, CMU, MIT, UC Berkeley, CMU, Montreal, Toronto, New York, Washington, Oxford, etc. (see also a list of some of the universities world-wide with AI programs and some US-specific lists: Computer Science Degree Hub and U.S.News Education), the following table contains a list of some companies and organizations in various countries that are also working towards making significant progress in advancing the state-of-the-art in Machine Intelligence. Some of these includes OpenAI, Google Deep Mind, Google Brain, Microsoft’s AI & Research Group, IBM Research – AI & Cognitive Computing, Facebook’s FAIR, Baidu’s IDL, SVAIL & BDL, Allen Institute of AI, Vicarious, Numenta, Kimera Systems, Cogitai, NNAISENSE, and Element.AI. A number of these companies and non-profit organizations also have the stated goal of solving intelligence for a better world, in particular Open AI, Deep Mind, MIRI, as well as Partnership on AI formed by Google, Facebook, Amazon, IBM and Microsoft.
Google is deeply invested in furthering artificial intelligence capabilities as demonstrated by nine AI startups that were acquired over the last few years. Google has also just added more brainpower to AI research unit in Canada. Their main research focus is on machine learning which helps advance Google’s language, speech translation, visual processing, ranking and prediction capabilities.
IBM has been a leader in AI since the 1950s and has also made three recent acquisitions to further complement its AI service, Watson. IBM’s Cognitive Horizons Network is advancing the science behind AI and cognitive computing through collaborative research.
Microsoft has also made some major moves in AI and working towards embedding AI in to agents, applications, services and infrastructure. It has also merged Bing, Cortana and Research to make a 5,000-strong AI division. Microsoft has revealed that Altera FPGAs have been installed across every Azure cloud server, creating what the company is calling “the world’s first AI supercomputer”. OpenAI has also recently started to work with Microsoft with respect to running most of their large-scale deep learning and AI experiments on Azure.
The Amazon Machine Learning platform provides companies with the ability to predict and find patterns using data. Additionally, Amazon Echo brings AI into the home through the intelligent voice server, Alexa.
Apple has also further improved their machine intelligence capability by acquiring four AI startups within the past two years and hiring a CMU professor as director of Apple’s AI research to smarten up Siri, Apple’s virtual assistant. The latter has transformed over the years from being a fairly simple voice assistant to being a fully-fledged digital assistant.
Salesforce has also acquired three AI companies over the past two years, and recently announced Salesforce Einstein, their AI service. Their latest initiative, which includes a team of 175 data scientists, uses machine learning to help employees more efficiently perform tasks by simplifying and speeding them up.
Twitter has have acquired four AI companies to date (the latest being the AI tech startup, Magic Pony) and plan to harness the expertise gained through these acquisitions to become a key player in the video space.
General Electric (GE), a leader in the Industrial Internet, has also recently acquired two AI startups, Bit Stew Systems and Wise.IO. This is in addition to other AI acquisitions done a few years ago which included SmartSignal and CSense Systems, the latter being a company that I have founded. NEC also plans to contribute AI to GE’s Predix platform for the industrial Internet of Things (IoT).
The Machine Intelligence Landscape
As can be seen from the Bloomberg Beta’s most recent version of the current state of Machine Intelligence 3.0 and the enterprise implication of this diagram, a multitude of machine intelligence companies (with the majority being startups) are penetrating most industries as well as enterprise functions and intelligence. Apart from ground, aerial, and industrial autonomous systems, intelligent agents are clearly a key part of the next technology revolution. The machine intelligence technology stack is also maturing with respect to supporting big data analytics and data science related tasks, as well as a range of smart applications. The machine intelligence landscape also reflects some key open source libraries and some of the companies and organizations focused on advancing the state-of-the-art in Machine Intelligence as part of their research and development efforts (as discussed above). The race is on…
Cortex Logic in partnership with MIIA to advance the state-of-the-art and solve real-world problems in Africa
There are also many large corporations, startups, and organizations embracing machine intelligence technology on the African continent. Some of these companies and organizations that I’m involved in not only use smart technology to solve problems, but also make a contribution towards advancing the-state-of-the-art in machine intelligence. The Machine Intelligence Institute of Africa (MIIA) is a non-profit organization and innovative community and accelerator for machine intelligence and data science research and applications to help transform Africa. See also Artificial Intelligence and Data Science use cases in Africa for some examples discussed at a recent MIIA meetup. Another company is Cortex Logic, a machine intelligence software & solutions company that solves real-world problems through data science services and smart applications such as intelligent virtual assistants and advisors for finance, healthcare, education, retail, telecoms and other industries where the automation of tasks can lead to economic benefit, scalability and productivity. It also aims to solve intelligence through advancing the state-of-the-art in machine intelligence and building cognitive systems that are contextually aware, learn at scale, support unsupervised learning where possible, reason with purpose and interact with humans naturally.
Cortex Logic’s research and development (R&D) efforts are focused on solving intelligence to have a positive impact on the world and help solve real-world problems. Cortex Logic is specifically developing contextually aware cognitive systems that learn at scale, support unsupervised learning, reason with purposes and interact with humans naturally. In order to this we are building a world-class team of machine / artificial intelligence and data science experts, developers and out-of-the-box thinkers. To support this Cortex Logic is, amongst other partners, working closely with the Machine Intelligence Institute of Africa (MIIA), an innovative community and accelerator for machine intelligence and data science research and applications (MIIA has as of November 2016 already 340+ members and growing). Cortex Logic intends to also collaborate with other research/academic organizations and companies that’s advancing the state-of-the-art in Machine Intelligence and contribute to machine intelligence related open source projects. In order to accelerate our research and development efforts (as illustrated in the diagram below), Cortex Logic is enhancing its own proprietary software, algorithms, APIs, and services with a combination of best-of-breed open source and commercial APIs, services, libraries, tools, platforms, repositories and cloud infrastructure.
Focus areas in advancing the state-of-the-art in Machine Intelligence
Given Cortex Logic’s focus on solving intelligence and advancing the state-of-the-art in Machine Intelligence, herewith a brief list of some of the current research areas in this regard:
- Recurrent Neural Networks & Sequence-to-Sequence Machine Learning
- Unsupervised Learning by incorporating learnings from both neuroscience and machine learning (e.g., build causal understanding of sensory space with temporal correlations of concurrent & sequential sensory signals)
- Use of Tensor methods and novel techniques in non-convex optimization to solve complex highly dimensional problems
- Topological based analysis of data sets to uncover the shape of data sets
- Attention models for powerful learning algorithms that require ever less data to be successful on harder problems
- Knowledge representation architectures instantly malleable & shapeable (e.g., sensory learning using patterns & sequences of patterns in cause-effect vs probabilistic way)
- Application of evolutionary computation in applications such as robotics, software agents, design and web commerce
- Combining control principles with reinforcement learning for robust Machine Learning
- Probabilistic Graphical Models
See also the following post that describes five capability levels of Deep Learning Intelligence: 1. Classification only; 2. Classification with memory; 3. Classification with knowledge; 4. Classification with imperfect knowledge; and 5. Collaborative classification with imperfect knowledge. Even for level 2 “classification with memory” systems which effectively corresponds to recurrent neural networks, there is a clear need for improved supervised and unsupervised training algorithms. As mentioned above, unsupervised learning should for example build a causal understanding of the sensory space with temporal correlations of concurrent and sequential sensory signals. Much research still needs to be done with respect to knowledge representations and integrating this with deep learning recurrent neural network systems all the way from level 3 to level 5 systems (classification with knowledge, imperfect knowledge and collaborative systems with imperfect knowledge).
Intelligent Virtual Assistants and Advisors
Machine intelligence is shifting toward building intelligent systems that can collaborate effectively with people, including creative ways to develop interactive and scalable ways for people to teach robots. Many people have already grown accustomed to touching and talking to their smart phones. It is also becoming clear that people’s future relationships with machines will become ever more fluid, shaded, and personalized. There are also now a shift from simply building systems that are intelligent to building intelligent systems that are human-aware and trustworthy, with natural language processing becoming an active area of machine perception. As we have seen with intelligent virtual assistants and advisors, research is now shifting towards developing systems that are able to interact with people through dialog, not just react to stylized requests.
As illustrated in the following diagram some of the business use cases for intelligent virtual assistants and advisors include those in medicine and healthcare, finance, education, law, as well as consumer behavior analysis, personal shopping bots, customer support bots, travel agents, tutors, knowledge assistant, security, and diagnostics.
Bennit.AI, An Intelligent Production Assistant for Manufacturing
Bennit A.I. is another practical example of an intelligent virtual assistants and advisor that we are developing in collaboration with colleagues in the US. It is a self-learning, personalized virtual production assistant specifically designed for industry, making the time and efficiency of industrial users its most important mission throughout day. Bennit A.I. layers across existing technologies to drastically simplify and improve access to information, helping industrial users make better decisions and improve business outcomes.
Machine Intelligence Institute of Africa
In follow-up to the successful MIIA Meetup in October, herewith the agenda and corresponding links for the MIIA Meetup on 30 November 2016 (see MIIA Meetup and MIIA Events for details), sponsored by Cortex Logic.
Agenda for MIIA Meetup @ LaunchLab on 30 November 2016 at 6pm
- Solving Intelligence, Solving Real-world Problems for a Better World (Presentation PDF, YouTube [4:30])
- Bennit.AI, an Intelligent Virtual Production Assistant for Manufacturing (Presentation PDF, YouTube [24:35], Demo, Overview Presentation)
- Data Science Academy (Presentation PDF, YouTube)
- The Hardware for Humanity Project (Presentation PDF, YouTube [5:46])
The meetup format consists of presentations and Q&A (approximately an hour) followed by sufficient time for networking and discussions afterwards.
- MIIA Meetup: http://www.meetup.com/Machine-Intelligence-Institute-of-Africa/
- LaunchLab: http://www.launchlab.co.za/30-november-machine-intelligence-institute-africa-meet/
- MIIA-LaunchLab Collaboration: The Hardware for Humanity Project
Some Recent MIIA posts
- Artificial Intelligence and Data Science use cases in Africa
- Machine Intelligence Institute of Africa
- Artificial Intelligence in Finance, Education, Healthcare, Manufacturing, Agriculture, and Government
- Artificial Intelligence at the centre of MIIA partner activities with Silicon Cape, Rise Africa, and Insights2Impact
MIIA YouTube Channel – Playlist
Joining Machine Intelligence Institute of Africa (MIIA)
Anyone interested to join MIIA and/or participate in using smart technologies to help address African problems such as those in education, finance, healthcare, energy, agriculture and unemployment is welcome to do this here. See How to participate for more details on various ways to help MIIA execute its mission.