On 2019(And previous years), we asked many top experts about 2020

Prediction.Some of the trends predicted last year have been realized:

  •  Pay more attention to ethics in AI
  •  Democratization of data science
  •  Progress in reinforcement learning
  •  China's growing success in AI

There are also surprises in 2019-last year's experts did not predictNLPBreakthroughs (such as GPT-2 and other versionsBERTAnd Transformers).

We ask experts again this year:

What are the main developments in AI, data science, deep learning and machine learning in 2019? What are the main trends you expect in 2020?

We received about 20 responses, Part XNUMX published earlier concerns about research.

This is the second part, focusing more on technology, industry and deployment. Some common topics include: AI Hype, AutoML, Cloud, Data, Interpretable AI, AI Ethics.

Here are the answers from Meta Brown, Tom Davenport, Carla Gentry, Nikita Johnson, Doug Laney, Bill Schmarzo, Kate Strachnyi, Ronald van Loon, Fabio Vazquez and Jen Underwood.

2020 Forecasting Technology Expert

Meta brown.@ metabown312Is the author of fool data mining and the president of A4A Brown

In 2018, the term "artificial intelligence" was used to describe a surge in usage from truly complex applications and increasingly successful applications such as self-driving cars to widespread use of propensity scores in direct marketing. I predict that by 2019, people will find that it's all math. I'm right.

On the one hand, more and more people are beginning to see the limitations now labelled "AI". The public is aware that facial recognition technology can be frustrated by Juggalo's makeup, that there is no smart life behind these customer service chatbots, and that you may spend millions of dollars trying to make software smarter than a doctor, but still fail.

However, "artificial intelligence" is still a buzzword, and venture capital funding is still pouring in. 2019

In the first nine months, more than $9 billion went to AI startups.In 130, the gap between these two prospects is growing. Artificial intelligence: the public is increasingly suspicious of the limitations of AI, skeptical and conscious of the image, and continue to invest hope, dreams and money in the business of AI promises And investment communities.

Tom Davenport(@tdav) Is President's Distinguished Professor of Information Technology and Management at Babson College, Co-founder of the International Institute for Analytics, Fellow of the Massachusetts Institute of Technology's Digital Economy Program, and Senior Advisor to Deloitte Analytics.

Major developments in 2019:

  •  Widespread deployment of automated machine learning tools for more structured aspects of data science.
  •  It is widely recognized that analytics and artificial intelligence have ethical dimensions and require conscious resolution
  •  There is increasing recognition that most analytics and AI models are not deployed and therefore have no value for the organization that created them

Upcoming developments in 2020:

  •  Provide tools to create, manage, and monitor an organization's suite of machine learning models, and continuously retrain drift models, with a focus on model inventory management.
  •  Improved status and recognition of analytics and AI converters, who work with business users and leaders to translate business requirements into high-level specifications for models
  •  Recognizing whether the model fits the data is just one consideration of whether the model is useful.

Carla Gentry.@data_nerdIs a consulting data scientist and owner of analytics solutions.

Regarding another year of hype and buzz about what artificial intelligence, machine learning, and data science can't do, I cringe for unskilled professionals entering the field, while universities issue so-called Certificates and degrees are not eligible to teach these courses.

Data science and machine learning rely on a large amount of data, but we are facing a year of misunderstanding about deviations, and the data that needs to be explained always faces the risk of deviations. Unbiased data are independent and need no explanation, for example-Mary has increased her sales return on investment by 10%, because Mary is a hardworking person, this is a view that cannot be measured.

A few days ago, the title of the article caught my attention: "Is data science dying?" Even before reading, my initial thought was: "No, but all the topics and hype I want to do will definitely not help our field. -Data science is more than just writing code".The misunderstanding of technology coupled with the lack of data and necessary infrastructure will continue to haunt us in 2020, but at least some people realize that the sexiest jobs in the 21st century are not so sexy after all, because we spend most of the cleaning and preparing data in Before we gather insights and answer business questions.

In 2020, let all of us remember that it is about data and ensure that we can advance our field in a complete and transparent manner. The era of artificial intelligence "black box" must pass before we can continue to move in a positive direction. development of. Remember that the algorithms, models, chatbots, etc. you build may affect someone’s life, and the data points in the database correspond to someone’s life, so please eliminate prejudice and let the facts speak for yourself... as always responsibly Play and entertainment data.

Founder of RE.WORK Deep Learning and AINikita Johnson.@teamrework.

In 2019, we have witnessed breakthroughs in many fields, which have made AI widely used as never before. Advanced software technologies such as transfer learning and reinforcement learning have also helped advance the development of AI breakthroughs and adoption, helping us separate system improvements under the constraints of human knowledge.

Next year, by 2020, we will move towards "interpretable AI" to increase the transparency, responsibility, and repeatability of AI models and technologies. We need to increase our understanding of the limitations and advantages and disadvantages of each tool. Enhanced learning will enhance our ability to build trust in the products we use and allow AI to make more informed decisions!

Doug Laney.@Doug_Laney, Chief Data Strategist, Caserta, Bestseller of Information Economics, Visiting Professor, School of Business, University of Illinois, Illinois

In the early 90s, the resurgence of artificial intelligence from a peaceful era, coupled with the mainstream of data science, nothing more than promoted the development of data. Today, big data is just data. Even if it continues to expand, its size will no longer overwhelm storage or computing power. At least there is no longer any excuse to say that any organization is tied to the sheer volume of data. (Hint: cloud.) Indeed, there have been gradual improvements in technology, but the surge of data from social media platforms, exchanges between partners, collection from websites, and availability on connected devices has made it impossible to Anticipated insights, automation and optimization. It has also spawned new data-centric business models.

I envision expanding the emergence of the information ecosystem in 2020 (not a pun, isn't it?), Thereby further enabling digital coordination among business partners driven by AI and data science. Some organizations may choose to build their own data exchange solutions to monetize their and their information assets. Other companies will enhance their advanced analytics capabilities through blockchain-enabled data exchange platforms and / or data aggregators that provide a range of alternative data.

Bill Schmarzo.@schmarzoIs CTO, IoT and Analytics Hitachi Vantara.

Major developments in 2019

  •  "Consumer proof points" about integrating AI into our daily lives through smartphones, websites, home devices, and vehicles is growing.
  •  Formal recognition of the DataOps category, an increasingly important recognition of the role of data engineering
  •  In the executive suite, there is increasing respect for the business potential of data science.
  • CIOs continue to work to realize their promise of data monetization. Disillusionment of the data lake leads to the "second operation" of the data lake

Major trends in 2020

  • Industrial companies are making more use of real-world examples, using sensors, edge analytics and AI to create products that become smarter through use; they appreciate rather than devalue use value
  •  Unable to provide reasonable financial or operational impact, the grand smart space project continues to struggle to surpass the initial pilot.
  •  For organizations that use data and analytics to drive meaningful business outcomes, the economic downturn will create a gap between "yes" and "no"

Kate Strachnyi.@StorybyData, DATAcated to tell the data story | Runner | 2's mom | The top voice of data science and analysis.

In 2019, we saw the integration of data visualization/business intelligence software; Salesforce acquired Tableau Software, and Google acquired Looker.This investment in business intelligence tools proves the company's value in data democratization and makes it easier for users to view and analyze their data.

What we can expect to see in 2020 is the continued shift towards automated data analysis / data science tasks. Data scientists and engineers need tools that can scale and solve more problems. This need will lead to the development of automation tools at multiple stages of the data science process. For example, some data preparation and erasure tasks are partially automated; however, they are difficult to fully automate due to the unique needs of a company. Other candidates for automation include feature engineering, model selection, and more.

Ronald Van Loon.@Ronald_vanLoon, Director of Adversitement, helping data-driven companies generate success. Top 10 big data, data science, IoT, artificial intelligence influencers

In 2019, the industry has witnessed the increasing popularity of interpretable artificial intelligence and enhanced analytics, which has enabled companies to bridge the gap between the potential that AI can provide and the complexity of decision-based technology to justify AI results. The full-stack AI approach is another development for organizations in 2019 to help accelerate the path to innovation and support AI growth, while improving integration and communication between different teams and individuals.

By 2020, due to the ease of use of conversational AI and the intuitive interface, we will see some trends in customer experience improvement. This automated solution enables the company to expand and transform the customer experience, while providing customers with 24/7 service, and provides opportunities for rapid problem resolution and reliable self-service. In addition, as we integrate AI into existing processes and work to change the questions we ask about AI, Narrow Intelligence will continue to support how we can most effectively use the power of people and machines.

Favio Vazquez.@FavioVaz.ClosterCEO

In 2019, we have seen amazing developments in artificial intelligence technology, mainly in terms of deep learning. Data science can use these advances to solve more difficult problems and shape the world we live in. Data science is the engine that uses science to catalyze change and turn paper into products. Our field is no longer just "hype", it is becoming a serious field. We will see more and more important online and offline education about data science and its friends. We hope that we are more confident in how and how we work. Semantic technology, decision intelligence and knowledge data science will become our companions in the coming years, so I suggest that people start exploring graph databases, ontology and knowledge representation systems.

Jen Underwood.@idigdata, The forces of nature move organizations faster

In 2019, we reached a critical point where organizations take seriously the competition in the algorithmic economy. Instead of launching a one-time project, market-leading companies are increasing the visibility of data science by planning enterprise-wide AI strategies. At the same time, established data science organizations have launched ethics, governance and ML Ops programs. Unfortunately, although adoption of machine learning has increased success rates, most have not yet.

From a technical perspective, we have witnessed the rise of hybrid distributed computing and serverless architectures. At the same time, algorithms, frameworks and AutoML solutions have evolved rapidly from innovation to commoditization.

By 2020, I expect personal data security, regulations, algorithmic biases, and deep fake topics will make headlines. From a brighter perspective, the advancement of interpretable AI and the enhanced understanding of natural language generation and optimization technologies will help bridge the gap between data science and business. With the further rise of data literacy and citizen data science programs, machine learning practitioners should continue to thrive.

This is a word cloud based on their predictions

2020 technology prediction word cloud


This article is adapted from kdnuggets,Original address