Shulou(Shulou.com)06/03 Report--

Produced by big data Digest

Compiled by Yang Wei, Zhou Jiaxu, Zhang Nanxing, Aileen

2018 is coming to an end and 2019 is about to begin. We will take you through the experts' analysis and prediction of AI. We put a question to the experts: what do you think are the major developments in data science and analysis in 2018? What will happen in 2019?

Although we are talking about data science and data analysis, AI is the main content in most of the answers. The topics mentioned by these experts include advances in artificial intelligence (real situation and hype are mixed); democratization of data science and data analysis, such as self-help analysis; and "everything is automated", such as data science, GDPR,AI risk, real-time analysis, and so on.

The list of experts is as follows: Meta Brown,Tom Davenport,Carla Gentry,Bob E Hayes,Cassie Kozyrkov,Doug Laney,Bill Schmarzo,Kate Strachnyi,Ronald van Loon,Favio Vazquez and Jen Underwood. Here are their profiles, Twitter ID and answers to this question.

Major advances in data Science and data Analysis in 2018 and Development Trends in 2019

Meta Brown (@ metabrown312), author of "Data Mining for Dummies" (getting started with data mining) and president of A4A Brown.Inc, a company dedicated to enhancing communication between management and technicians.

The hot topic of analysis in 2018 is artificial intelligence (AI). In my impression, artificial intelligence has caused more discussion than other analytical applications. But regrettably, most of the discussions did not produce much value.

Alan Turing, a pioneer in computer science, once envisioned that computer intelligence would rival human intelligence, that is, people could not tell whether they were talking to computers or humans.

Think about how we interact with today's artificial intelligence applications. Personal assistants such as Siri or Alexa are useful, but the feeling of communicating with them is far from being able to communicate with real people. Online self-help chat robots are even more disappointing. If you try to ask a real-life question, you will realize that there is no real "brain" behind it.

According to Turing's definition, artificial intelligence does not yet exist. Gary Marcus, a professor of psychology and neuroscience at New York University, says the biggest misunderstanding about artificial intelligence is that "we are not far from artificial intelligence."

Indeed, computers have been used to perform logical operations in our real life. Although they don't think like people, they are fast enough and continuous enough, which are valuable advantages. These applications enable machines to perform practical tasks, such as marking potentially fraudulent transactions, operating cars, and so on.

Although the limitations of AI technology are obvious, the public, and even the scientific and technological community, are full of unrealistic claims and expectations for artificial intelligence. These distorted views arouse fear in the crowd and frustrate some people's expectations, while the reality we see is even more disappointing.

Tom Davenport (@ tdav) is a distinguished professor of presidential information technology and management at Babson College, co-founder of the Institute for International Analysis, a researcher at the MIT Institute for Digital Economics and a senior mentor at Deloitte Analytics.

We have made a forecast and analysis of the annual development trend of technology at the International Institute of Analytics. here are some of my points:

Organizations are increasingly concerned about the model deployment rate-according to the Rex data Science (Rexer Data Science Survey) survey, only 10% of 15% of companies deploy "almost always", the other 50% of companies deploy only "often", and the remaining 35%-40% of companies have only an "occasional or rare" chance of successfully deploying analytical models. I have met some organizations that say their successful deployment rate is less than 10%. Of course, undeployed analytical models have no economic value. Companies need to evaluate and increase deployment rates in 2019.

Private data scientists and business analysts will always exist, and there will be more and more. The rise of visualization and search-based analysis, as well as more automated machine learning at the front end of data science, means that amateurs will produce a large number of analysis results. There is no doubt that the fight against this trend will fail, so it is better to turn to empowerment to help its development. This also means that professionals engaged in quantitative work can focus on complex and difficult modeling tasks, or turn to understanding the business and solving organizational change problems.

Carla Gentry (@ dat_nerd), consult data scientists and Analytical-Solution owners.

2018 is a brilliant year for the development of data analysis and data science, but we have also seen the outbreak of artificial intelligence, neural networks and machine learning, and not all propositions are supported by talent and experience. We are also seeing an increase in the use of artificial intelligence in health care and policing, but again, the risks of prejudice and the support of talent and experience may not be fully taken into account, and I think some people may have forgotten a big principle: in these situations, with wearables and the Internet of things (Google Home,Alexa, etc.), data is everything, and this trend will continue.

In 2019, the industry will continue to discuss these buzzwords, and companies will begin to implement what they have learned from tens of thousands or even billions of examples using neural networks, and to make matters worse, whenever you want neural networks to identify new objects, you have to start from scratch (at least time consumption is necessary)-and talent is another problem. Apart from Geoffrey Hinton,Yejin Choi or Yann LeCun, you are really not a neural network expert, so don't count on a talent pool.

Data science focuses on gathering all kinds of data and gaining insights from it, and in some cases, it is incorrect and unnecessary to want to be an expert in AL, machine learning or neural networks, so we must study the differences carefully, and newcomers will have to relearn new technologies in order to be some competitive in the new field. My concern is that if we don't really understand how machines learn and how to apply artificial intelligence correctly, the weaknesses of some companies / algorithms / enterprises will continue to be exposed.

Let's enjoy these new technologies with joy, but understand that there will be serious consequences if you screw up!

Bob E. Hayes (@ bobehayes), researcher, author and consultant, publisher of Business over Broadway, holds a doctorate in industrial organization psychology.

The field of data science and data analysis has experienced an upsurge in all content related to machine learning, including reinforcement learning, chatbots and social impact research.

In 2019, I expect people to pay more and more attention to the ethics of artificial intelligence, including privacy and security issues. More and more emphasis is placed on the understanding of the algorithm's decision-making process; we need to know not only that machine learning helps us make decisions, but also how it works (how it makes decisions). In addition, American companies will focus on how to use consumers' personal data. California has passed the California Consumer Privacy Act (which will take effect in January 2020), and I expect and hope that other states will follow.

But at the same time, I am also worried that artificial intelligence / mechanics will lead to an increase in the creation and dissemination of fake news. Deep fakes (artificial intelligence's deep face-changing technology) shows that people can fabricate video content to show things they have never said or done without feeling guilty. As Max Tegmack (Max Tegmark) said, fully recognize that the potential threat of artificial intelligence is not Niezo, but only "security engineering".

Now there are many ways for data practitioners to learn data science, such as training camps, MOOC, and college courses, but I also hope that non-data professionals (for example, managers and front-line employees) will try to take data analysis courses more often.

Cassie Kozyrkov (@ quaesita), Chief decision engineer of Google Cloud. Statistics enthusiasts, artificial intelligence, data, puns, art, science fiction, drama, decision science.

One of the major developments in 2018 is the democratization of data science. With cloud technology, people can provide continuous power for resource-intensive big data and AI applications without the need to build Kubeflow and other tools to build data centers, thus providing scalable data science for those who do not have architectural expertise. This trend of empowering the public and making data scientific tools accessible to all will accelerate in 2019.

Doug Laney (@ Doug_Laney) is vice president, distinguished analyst and chief data officer researcher at Gartner, and author of Infonomics.

Gartner has just released a strategic forecast for data and analysis in 2019. It mentions a point of view: it is clearly stated in the enterprise strategy that information is an important asset of the enterprise, and analysis is an indispensable ability. Not only does the IT strategy mention this, but corporate strategies and plans also mention similar content.

In addition, we expect that data literacy programs will become the norm to help business people and data analysis professionals communicate better, especially in the context of increasingly complex data analysis needs. As principles and practices in information economics spread widely, we expect chief data officers to work more frequently with chief financial officers to formally assess the organization's information assets. Doing so can bring important information management and business benefits to many customers. But analysis and digital ethics are still an issue, and we believe organizations will begin to introduce a professional code of conduct for their data science teams.

In addition, we expect that in the next 3-5 years, most new business systems will achieve sustained intelligent applications by making full use of real-time environmental data; the POC project of quantum computing (proof-of-concept project) will make computers much more capable of analysis than it is now; intelligent enhanced and automated internal analysis insights will replace most of the forecast reports; and the utilization rate of location analysis will increase nearly tenfold. Machine learning will also ease the competition for talent for data scientists.

President of Gregory Piatetsky,@kdnuggets,KDnuggets, data scientist, co-founder of KDD Conference and SIGKDD, and top speaker of LinkedIn 2018 in the field of data science and analytics.

I think the main developments in 2018 are:

The General data Protection regulations (GDPR), which came into effect in May 2018, is an important milestone in data science, not only in Europe, but also in the United States and elsewhere, and many companies have updated their privacy policies under its influence. But at the same time, it remains to be seen whether consumer privacy protection under the new regulations can really be improved, or whether consumer privacy data can be used as usual to create value under the new protection policy.

Democratization of data science: with the growth of access tools and channels, the democratization of data science continues. The new tool released by AWS reinvent is one of them.

Artificial intelligence risk: the first fatal accident in a self-driving car is that the machine fails to distinguish between pedestrians pushing bicycles. This incident has attracted social attention to the inevitable risks of artificial intelligence. At the same time, it should be noted that it is unrealistic to use zero error as the criterion for evaluating autopilot (and automated AI). The correct way to look at it is that it makes driving risk much lower than it is now. For example, the accident rate of human driving is very high, with 37000 people killed in the United States in 2017 alone.

Main trends in 2019:

The automation of data science will accelerate, but at least in recent years, the work of data scientists will not be replaced by complete automation.

The progress and hype of artificial intelligence: although the development of artificial intelligence is real, the hype of artificial intelligence is growing more rapidly.

China has become a major player in artificial intelligence, and many Chinese companies are starting to carry out independent innovation and research and development, rather than just copying from the United States.

Reinforcement learning will play a more and more important role in the progress of artificial intelligence. For example, reinforcement learning algorithms have made amazing progress in solving Montezuma's Revenge Atari game, reaching an unprecedented level of 100, surpassing the best previous human or computing records.

Chief Technical Officer, Internet of things and Analytics, Bill Schmarzo, @ schmarzo, Hitachi Vantara

Big data, main developments in data science and analysis in 2018:

Through the publication of more and more successful cases, stakeholders have a deeper understanding of the application potential of machine learning and deep learning.

There are still investment misunderstandings in the data lake. Too many organizations see the data Lake as a way to expel expensive data warehouses and reduce ETL costs, but do not fully understand its potential as a platform for value creation for stakeholders and data science teams to access and drive business value.

Main trends in 2019:

For industry leaders, the main motivation for the development of big data and data science should come from the business side, not the IT side. Leading enterprises will enable big data, the Internet of things and data science (machine learning, deep learning, artificial intelligence) to drive business development by identifying, verifying, reviewing, evaluating and optimizing business results.

In addition to using data science to optimize key business and operational processes, which is still a good place to start getting an attractive return on investment, leading companies will realize that insights into customers, products, and operations hidden in the data will be the drivers of future opportunities.

Kate Strachnyi (@ StorybyData), data visualization expert, author of "saboteur: leaders in data Science" and "how to be a data Scientist"; video podcast on human data science.

Major developments in data science and analysis in 2018:

General data Protection regulations (GDPR): EU regulations, which came into force in May 2018, provide a set of personal data protection regulations for EU citizens. This practice promotes the setting of similar standards elsewhere. For example, California has passed its own Digital Privacy Act, which allows consumers to know what information they are collecting, why they are collecting it, and with whom they share it.

Self-service business intelligence (BI) tools: business intelligence analysis tools will become more common among data analysts and business analysts. However, it is not clear whether the users of these tools really understand the true meaning of these analysis scenarios. There is still a gap between the speed at which people learn to use drag-and-drop tools and make charts, and their understanding of real business scenarios.

Main trends in 2019:

Data morality and privacy: in every aspect of data science, the ethics and privacy of using data will be paid more and more attention. People who dance with data need to have a clear understanding of the power they have and the broad impact that their work can have. With the development of digitization around the world, these issues should be paid attention to at the individual, corporate and government levels.

Process automation: the company will continue to promote process automation to reduce costs and improve efficiency. This automation may also lead to unemployment for individuals who automate the process. Therefore, in the future, people need to focus on constantly learning new skills in order to adapt to the rapidly changing environment.

Ronald vanLoon (@ Ronald_vanLoon), director of Adversitement (a company that helps data-driven companies succeed), big data, one of the top ten influencers of data science, Internet of things, and artificial intelligence.

In 2018, as end-to-end data management grows, companies increase the maturity of their analytical capabilities while fully tapping all data resources to gain trusted insights and build infrastructure and business models adapted to the current digital economy. Machine learning is widely accepted by the public, and software vendors use a lot of machine learning technology in their product applications to solve specific problems.

In 2019, to support more complex and high-level deep learning applications, more integrated hardware and software frameworks will emerge to promote a wider range of innovation. Deep learning applications require fully optimized hardware and software stacks to advance the new modern AI architecture. We will see suppliers in a variety of industries begin to use this full-stack solution to meet the need for in-depth learning of the best performance and functionality.

With the growth of Internet of things devices, real-time edge analysis will grow exponentially, which makes real-time analysis easier and will help to obtain immediate feedback based on real-time insights.

Favio Vazquez (@ FavioVaz), data scientist. Physicist and Computing engineer, founder of Ciencia y Datos

2018 is an amazing year for data Science (DS), both in theory and in practice. The industry has put forward several research methods of data science, which can help to transform data science into practical science. I have been involved in the discussion for more than a year, and recently I have seen more people participate in the discussion. Compared with machine learning, the content of automated machine learning is very complicated, including automated deep learning.

Main trends in 2019:

AutoX: we will see more and more companies using automated machine learning and deep learning technologies in their stacking technologies and function libraries. X here means that this automated tool will be extended to data intake, data integration, data cleaning, exploration and deployment. Automation will exist here.

Semantic technology: my most interesting discovery this year is the correlation between data science and semantics. It is not a new field in the data world, but I have found that more and more people are interested in semantics, ontology, knowledge graph and their relevance to data science and machine learning.

Programming complexity into simplicity: this is a difficult thing to define, but as every process of data analysis is automated, our daily programming work will be less and less. We will have tools for creating code that will understand our need for natural language processing and then convert it into queries, commands, and complete programs. I think programming is still a very important part of learning, but it will soon become simpler and easier.

Digital education: digital education is growing every year, but next year we will see more people participating in MOOC, digital courses or online courses than ever before. Some people call it "democratization of education", and I agree with that to a large extent, but I want to say to everyone: pay attention to the content and way you learn, and investigate the courses you have learned before investing time and money. A good course will change your life, and on the contrary, it is also very dangerous.

Jen Underwood (@ idigdata), senior director of DataRobot, founder of Impact Analytix.LLC.

The hype and transformation of artificial intelligence will be everywhere in 2018. A few years ago, big data was all the rage, then cloud technology, and now machine learning is dominant. Artificial intelligence has gone too far in mobile applications, robotics and business intelligence solutions. Today, even beer is driven by artificial intelligence.

This year, we have also seen a rapid expansion of the automation market. Today, many machine learning solutions are promoting the transformation from manually guided, automated data analysis to automated machine learning throughout the life cycle. From simple dragging and clicking buttons to create basic models, to complex feature engineering, model retrieval, parameter adjustment, deployment, model management and monitoring, automated machine learning has a wide range of applications-with uneven results.

In 2019, public management, privacy, prejudice, moral issues and deeper false questions about data science will challenge our confidence in artificial intelligence. Innovative technologies such as blockchain will begin to change the way we store, share, and track data. I also expect that artificial intelligence will become fairer, more transparent and more responsible, so that non-data scientists can understand, interpret and trust artificial intelligence. At present, there is still a huge gap in translating data science terms into languages that ordinary people can understand. Many organizations apply artificial intelligence technology in our imperfect world, while providing a hotbed for the emergence of private data scientists, more people need to become data experts to ensure that the direction of AI is not deviated.

Related reports:

Https://www.kdnuggets.com/2018/12/predictions-data-science-analytics-2019.htm

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.