Some people have spent their quarantine downtime bakingsourdough bread. Others experiment with tie-dye. But others namely Toronto-based artist Daniel Voshart have createdpainstaking portraits of all 54 Roman emperors of the Principate period, which spanned from 27 BC to 285 AD.

The portraits help people visualize what the Roman emperors would have looked like when they were alive.

Included are Vosharts best artistic guesses of the faces of emperors Augustus, Nero, Caligula, Marcus Aurelius and Claudius, among others. They dont look particularly heroic or epic rather, they look like regular people, with craggy foreheads, receding hairlines and bags under their eyes.

To make the portraits, Voshart used a design software called Artbreeder, which relies on a kind of artificial intelligence called generative adversarial networks (GANs).

Voshart starts by feeding the GANs hundreds of images of the emperors collected from ancient sculpted busts, coins and statues. Then he gets a composite image, which he tweaks in Photoshop. To choose characteristics such as hair color and eye color, Voshart researches the emperors backgrounds and lineages.

It was a bit of a challenge, he says. About a quarter of the project was doing research, trying to figure out if theres something written about their appearance.

He also needed to find good images to feed the GANs.

Another quarter of the research was finding the bust, finding when it was carved because a lot of these busts are recarvings or carved hundreds of years later, he says.

In a statement posted on Medium, Voshartwrites: My goal was not to romanticize emperors or make them seem heroic. In choosing bust/sculptures, my approach was to favor the bust that was made when the emperor was alive. Otherwise, I favored the bust made with the greatest craftsmanship and where the emperor was stereotypically uglier my pet theory being that artists were likely trying to flatter their subjects.

Related:Battle of the bums: Museums complete over best artistic behinds

Voshart is not a Rome expert. His background is in architecture and design, and by day he works in the art department of the TV show "Star Trek: Discovery," where he designs virtual reality walkthroughs of the sets before they'rebuilt.

But when the coronavirus pandemic hit, Voshart was furloughed. He used the extra time on his hands to learn how to use the Artbreeder software.The idea for the Roman emperor project came from a Reddit threadwhere people were posting realistic-looking images theyd created on Artbreeder using photos of Roman busts. Voshart gave it a try and went into exacting detail with his research and design process, doing multiple iterations of the images.

Voshart says he made some mistakes along the way. For example, Voshart initially based his portrait of Caligula, a notoriously sadistic emperor, on a beautifully preserved bust in the Metropolitan Museum of Art. But the bust was too perfect-looking, Voshart says.

Multiple people told me he was disfigured, and another bust was more accurate, he says.

So, for the second iteration of the portrait, Voshart favored a different bust where one eye was lower than the other.

People have been telling me my first depiction of Caligula was hot, he says. Now, no ones telling me that.

Voshart says people who see his portraits on Twitter and Reddit often approach them like theyd approachTinder profiles.

I get maybe a few too many comments, like such-and-such is hot. But a lot of these emperors are such awful people!

I get maybe a few too many comments, like such-and-such is hot. But a lot of these emperors are such awful people! Voshart says.

Voshart keeps a list on his computer of all the funny comparisons people have made to present-day celebrities and public figures.

Ive heard Nero looks like a football player. Augustus looks like Daniel Craigmy early depiction of Marcus Aurelius looks like the Dude from 'The Big Lebowski.'

But the No. 1 comment? Augustus looks like Putin.

Related:UNESCO says scammers are using its logo to defraudartcollectors

No one knows for sure whether Augustus actually looked like Vladimir Putin in real life.Voshart says his portraits are speculative.

Its definitely an artistic interpretation, he says. Im sure if you time-traveled, youd be very angry at me."

Visit link:
This artist used machine learning to create realistic portraits of Roman emperors - The World

Machine Learning Chips Market 2018: Global Industry Insights by Global Players, Regional Segmentation, Growth, Applications, Major Drivers, Value and Foreseen till 2024

The recent published research report sheds light on critical aspects of the global Machine Learning Chips market such as vendor landscape, competitive strategies, market drivers and challenges along with the regional analysis. The report helps the readers to draw a suitable conclusion and clearly understand the current and future scenario and trends of global Machine Learning Chips market. The research study comes out as a compilation of useful guidelines for players to understand and define their strategies more efficiently in order to keep themselves ahead of their competitors. The report profiles leading companies of the global Machine Learning Chips market along with the emerging new ventures who are creating an impact on the global market with their latest innovations and technologies.

Request Sample Report @ https://www.marketresearchhub.com/enquiry.php?type=S&repid=2632983&source=atm

The recent published study includes information on key segmentation of the global Machine Learning Chips market on the basis of type/product, application and geography (country/region). Each of the segments included in the report is studies in relations to different factors such as market size, market share, value, growth rate and other quantitate information.

The competitive analysis included in the global Machine Learning Chips market study allows their readers to understand the difference between players and how they are operating amounts themselves on global scale. The research study gives a deep insight on the current and future trends of the market along with the opportunities for the new players who are in process of entering global Machine Learning Chips market. Market dynamic analysis such as market drivers, market restraints are explained thoroughly in the most detailed and easiest possible manner. The companies can also find several recommendations improve their business on the global scale.

The readers of the Machine Learning Chips Market report can also extract several key insights such as market size of varies products and application along with their market share and growth rate. The report also includes information for next five years as forested data and past five years as historical data and the market share of the several key information.

Make An EnquiryAbout This Report @ https://www.marketresearchhub.com/enquiry.php?type=E&repid=2632983&source=atm

Global Machine Learning Chips Market by Companies:

The company profile section of the report offers great insights such as market revenue and market share of global Machine Learning Chips market. Key companies listed in the report are:

Market Segment AnalysisThe research report includes specific segments by Type and by Application. Each type provides information about the production during the forecast period of 2015 to 2026. Application segment also provides consumption during the forecast period of 2015 to 2026. Understanding the segments helps in identifying the importance of different factors that aid the market growth.Segment by TypeNeuromorphic ChipGraphics Processing Unit (GPU) ChipFlash Based ChipField Programmable Gate Array (FPGA) ChipOther

Segment by ApplicationRobotics IndustryConsumer ElectronicsAutomotiveHealthcareOther

Global Machine Learning Chips Market: Regional AnalysisThe report offers in-depth assessment of the growth and other aspects of the Machine Learning Chips market in important regions, including the U.S., Canada, Germany, France, U.K., Italy, Russia, China, Japan, South Korea, Taiwan, Southeast Asia, Mexico, and Brazil, etc. Key regions covered in the report are North America, Europe, Asia-Pacific and Latin America.The report has been curated after observing and studying various factors that determine regional growth such as economic, environmental, social, technological, and political status of the particular region. Analysts have studied the data of revenue, production, and manufacturers of each region. This section analyses region-wise revenue and volume for the forecast period of 2015 to 2026. These analyses will help the reader to understand the potential worth of investment in a particular region.Global Machine Learning Chips Market: Competitive LandscapeThis section of the report identifies various key manufacturers of the market. It helps the reader understand the strategies and collaborations that players are focusing on combat competition in the market. The comprehensive report provides a significant microscopic look at the market. The reader can identify the footprints of the manufacturers by knowing about the global revenue of manufacturers, the global price of manufacturers, and production by manufacturers during the forecast period of 2015 to 2019.The major players in the market include Wave Computing, Graphcore, Google Inc, Intel Corporation, IBM Corporation, Nvidia Corporation, Qualcomm, Taiwan Semiconductor Manufacturing, etc.

Global Machine Learning Chips Market by Geography:

You can Buy This Report from Here @ https://www.marketresearchhub.com/checkout?rep_id=2632983&licType=S&source=atm

Some of the Major Highlights of TOC covers in Machine Learning Chips Market Report:

Chapter 1: Methodology & Scope of Machine Learning Chips Market

Chapter 2: Executive Summary of Machine Learning Chips Market

Chapter 3: Machine Learning Chips Industry Insights

Chapter 4: Machine Learning Chips Market, By Region

Chapter 5: Company Profile

And Continue

Read more from the original source:
Machine Learning Chips Market Dynamics Analysis to Grow at Cagr with Major Companies and Forecast 2026 - The Scarlet

Machine learning era has reached the stage of interpretability where developing models and making predictions is simply not enough any more. To make a powerful impact and get good results on the data it is important to investigate and probe the dataset and the models. A good model investigation involves digging deep into the understanding of the model to find insights and inconsistencies in the developed model. This task usually involves writing a lot of custom functions. But, with tools like What-If, it makes the probing task very easy and saves time and efforts for programmers.

In this article we will learn about:

What-If tool is a visualization tool that is designed to interactively probe the machine learning models. WIT allows users to understand machine learning models like classification, regression and deep neural networks by providing methods to evaluate, analyse and compare the model. It is user friendly and can be used not only by developers but also by researchers and non-programmers very easily.

WIT was developed by Google under the People+AI research (PAIR) program. It is open-source and brings together researchers across Google to study and redesign the ways people interact with AI systems.

This tool provides multiple features and advantages for users to investigate the model.

Some of the features of using this are:

WIT can be used with a Google Colab notebook or Jupyter notebook. It can also be used with Tensorflow Board.

Let us take a sample dataset to understand the different features of WIT. I will choose the forest fire dataset available for download on Kaggle. You can click here for downloading the dataset. The goal here is to predict the areas affected by forest fires given the temperature, month, amount of rain etc.

I will implement this tool on google collaboratory. Before we load the dataset and perform the processing, we will first install the WIT. To install this tool use,

!pip install witwidget

Once we have split the data, we can convert the columns month and day to categorical values using label encoder.

Now we can build our model. I will use sklearn ensemble model and implement the gradient boosting regression model.

Now that we have the model trained, we will write a function to predict the data since we need to use this for the widget.

Next, we will write the code to call the widget.

This opens an interactive widget with two panels.

To the left, there is a panel for selecting multiple techniques to perform on the data and to the right is the data points.

As you can see on the right panel we have options to select features in the dataset along X-axis and Y-axis. I will set these values and check the graphs.

Here I have set FFMC along the X-axis and area as the target. Keep in mind that these points are displayed after the regression is performed.

Let us now explore each of the options provided to us.

You can select a random data point and highlight the point selected. You can also change the value of the datapoint and observe how the predictions change dynamically and immediately.

As you can see, changing the values changes the predicted outcomes. You can change multiple values and experiment with the model behaviour.

Another way to understand the behaviour of a model is to use counterfactuals. Counterfactuals are slight changes made that can cause a model to flip its decision.

By clicking on the slide button shown below we can identify the counterfactual which gets highlighted in green.

This plot shows the effects that the features have on the trained machine learning model.

As shown below, we can see the inference of all the features with the target value.

This tab allows us to look at the overall model performance. You can evaluate the model performance with respect to one feature or more than the one feature. There are multiple options available for analysis of the performance.

I have selected two features FFMC and temp against the area to understand performance using mean error.

If multiple training models are used their performance can be evaluated here.

The features tab is used to get the statistics of each feature in the dataset. It displays the data in the form of histograms or quantile charts.

The tab also enables us to look into the distribution of values for each feature in the dataset.

It also highlights the features that are most non-uniform in comparison to the other features in the dataset.

Identifying non-uniformity is a good way to reduce bias in the model.

WIT is a very useful tool for analysis of model performance. Ability to inspect models in a simple no-code environment will be of great help especially in the business perspective.

It also gives insights to factors beyond training the model like understanding why and how that model was created and how the dataset is fitting in the model.

comments

Read the original:
Demonstration Of What-If Tool For Machine Learning Model Investigation - Analytics India Magazine

Fort Collins, Colorado The Global Machine Learning & Big Data Analytics Education Market research report offers insightful information on the Global Machine Learning & Big Data Analytics Education market for the base year 2019 and is forecast between 2020 and 2027. Market value, market share, market size, and sales have been estimated based on product types, application prospects, and regional industry segmentation. Important industry segments were analyzed for the global and regional markets.

The effects of the COVID-19 pandemic have been observed across all sectors of all industries. The economic landscape has changed dynamically due to the crisis, and a change in requirements and trends has also been observed. The report studies the impact of COVID-19 on the market and analyzes key changes in trends and growth patterns. It also includes an estimate of the current and future impact of COVID-19 on overall industry growth.

Get a sample of the report @ https://reportsglobe.com/download-sample/?rid=64357

The report has a complete analysis of the Global Machine Learning & Big Data Analytics Education Market on a global as well as regional level. The forecast has been presented in terms of value and price for the 8 year period from 2020 to 2027. The report provides an in-depth study of market drivers and restraints on a global level, and provides an impact analysis of these market drivers and restraints on the relationship of supply and demand for the Global Machine Learning & Big Data Analytics Education Market throughout the forecast period.

The report provides an in-depth analysis of the major market players along with their business overview, expansion plans, and strategies. The main actors examined in the report are:

The Global Machine Learning & Big Data Analytics Education Market Report offers a deeper understanding and a comprehensive overview of the Global Machine Learning & Big Data Analytics Education division. Porters Five Forces Analysis and SWOT Analysis have been addressed in the report to provide insightful data on the competitive landscape. The study also covers the market analysis and provides an in-depth analysis of the application segment based on the market size, growth rate and trends.

Request a discount on the report @ https://reportsglobe.com/ask-for-discount/?rid=64357

The research report is an investigative study that provides a conclusive overview of the Global Machine Learning & Big Data Analytics Education business division through in-depth market segmentation into key applications, types, and regions. These segments are analyzed based on current, emerging and future trends. Regional segmentation provides current and demand estimates for the Global Machine Learning & Big Data Analytics Education industry in key regions in North America, Europe, Asia Pacific, Latin America, and the Middle East and Africa.

Global Machine Learning & Big Data Analytics Education Market Segmentation:

In market segmentation by types of Global Machine Learning & Big Data Analytics Education , the report covers-

In market segmentation by applications of the Global Machine Learning & Big Data Analytics Education , the report covers the following uses-

Request customization of the report @https://reportsglobe.com/need-customization/?rid=64357

Overview of the table of contents of the report:

To learn more about the report, visit @ https://reportsglobe.com/product/global-machine-learning-big-data-analytics-education-assessment/

Thank you for reading our report. To learn more about report details or for customization information, please contact us. Our team will ensure that the report is customized according to your requirements.

How Reports Globe is different than other Market Research Providers

The inception of Reports Globe has been backed by providing clients with a holistic view of market conditions and future possibilities/opportunities to reap maximum profits out of their businesses and assist in decision making. Our team of in-house analysts and consultants works tirelessly to understand your needs and suggest the best possible solutions to fulfill your research requirements.

Our team at Reports Globe follows a rigorous process of data validation, which allows us to publish reports from publishers with minimum or no deviations. Reports Globe collects, segregates, and publishes more than 500 reports annually that cater to products and services across numerous domains.

Contact us:

Mr. Mark Willams

Account Manager

US: +1-970-672-0390

Email:[emailprotected]

Web:reportsglobe.com

See the original post:
Machine Learning & Big Data Analytics Education Market Size is Thriving Worldwide 2020 | Growth and Profit Analysis, Forecast by 2027 - The Daily...

The algorithm could be used to quickly identify social media posts related to a disaster.

Getty Images

There has been a significant increase in the use of social media to share updates, seek help and report emergencies during a disaster. Algorithms keeping track of social media posts that signal the occurrence of natural disasters must be swift so that relief operations can be mobilized immediately.

A team of researchers led by Ruihong Huang, assistant professor in the Department of Computer Science and Engineering at Texas A&M University, has developed a novel weakly supervised approach that can train machine learning algorithms quickly to recognize tweets related to disasters.

Because of the sudden nature of disasters, theres not much time available to build an event recognition system, Huang said. Our goal is to be able to detect life-threatening events using individual social media messages and recognize similar events in the affected areas.

Text on social media platforms, like Twitter, can be categorized using standard algorithms called classifiers. This sorting algorithm separates data into labeled classes or categories, similar to how spam filters in email service providers scan incoming emails and classify them as either spam or not spam based on its prior knowledge of spam messages.

Most classifiers are an integral part of machine learning algorithms that make predictions based on carefully labeled sets of data. In the past, machine learning algorithms have been used for event detection based on tweets or a burst of words within tweets. To ensure a reliable classifier for the machine learning algorithms, human annotators have to manually label large amounts of data instances one by one, which usually takes several days, sometimes even weeks or months.

The researchers also found that it is essentially impossible to find a keyword that does not have more than one meaning on social media depending on the context of the tweet. For example, if the word dead is used as a keyword, it will pull in tweets talking about a variety of topics such as a phone battery being dead or the television series The Walking Dead.

We have to be able to know which tweets that contain the predetermined keywords are relevant to the disaster and separate them from the tweets that contain the correct keywords but are not relevant, Huang said.

To build more reliable labeled datasets, the researchers first used an automatic clustering algorithm to put them into small groups. Next, a domain expert looked at the context of the tweets in each group to identify if it was relevant to the disaster. The labeled tweets were then used to train the classifier how to recognize the relevant tweets.

Using data gathered from the most impacted time periods for Hurricane Harvey and Hurricane Florence, the researchers found that their data labeling method and overall weakly-supervised system took one to two person-hours instead of the 50 person-hours that were required to go through thousands of carefully annotated tweets using the supervised approach.

Despite the classifiers overall good performance, they also observed that the system still missed several tweets that were relevant but used a different vocabulary than the predetermined keywords.

Users can be very creative when discussing a particular type of event using the predefined keywords, so the classifier would have to be able to handle those types of tweets, Huang said. Theres room to further improve the systems coverage.

In the future, the researchers will look to explore how to extract information about the users location so first responders will know exactly where to dispatch their resources.

Other contributors to this research include Wenlin Yao, a doctoral student supervised by Huang from the computer science and engineering department; Ali Mostafavi and Cheng Zhang from the Zachry Department of Civil and Environmental Engineering; and Shiva Saravanan, former intern of the Natural Language Processing Lab at Texas A&M.

The researchers described their findings in the proceedings from the Association for the Advancement of Artificial Intelligences 34th Conference on Artificial Intelligence.

This work is supported by funds from the National Science Foundation.

Follow this link:
Improving The Use Of Social Media For Disaster Management - Texas A&M University Today

Aims

Prediction models for outcomes in atrial fibrillation (AF) are used to guide treatment. While regression models have been the analytic standard for prediction modelling, machine learning (ML) has been promoted as a potentially superior methodology. We compared the performance of ML and regression models in predicting outcomes in AF patients.

The Outcomes Registry for Better Informed Treatment of Atrial Fibrillation (ORBIT-AF) and Global Anticoagulant Registry in the FIELD (GARFIELD-AF) are population-based registries that include 74 792 AF patients. Models were generated from potential predictors using stepwise logistic regression (STEP), random forests (RF), gradient boosting (GB), and two neural networks (NNs). Discriminatory power was highest for death [STEP area under the curve (AUC) = 0.80 in ORBIT-AF, 0.75 in GARFIELD-AF] and lowest for stroke in all models (STEP AUC = 0.67 in ORBIT-AF, 0.66 in GARFIELD-AF). The discriminatory power of the ML models was similar or lower than the STEP models for most outcomes. The GB model had a higher AUC than STEP for death in GARFIELD-AF (0.76 vs. 0.75), but only nominally, and both performed similarly in ORBIT-AF. The multilayer NN had the lowest discriminatory power for all outcomes. The calibration of the STEP modelswere more aligned with the observed events for all outcomes. In the cross-registry models, the discriminatory power of the ML models was similar or lower than the STEP for most cases.

When developed from two large, community-based AF registries, ML techniques did not improve prediction modelling of death, major bleeding, or stroke.

View post:
Machine Learning Does Not Improve Upon Traditional Regression in Predicting Outcomes in Atrial Fibrillation: An Analysis of the ORBIT-AF and...

Global Machine Learning in Medical Imaging Industry: with growing significant CAGR during Forecast 2020-2025

Latest Research Report on Machine Learning in Medical Imaging Market which covers Market Overview, Future Economic Impact, Competition by Manufacturers, Supply (Production), and Consumption Analysis

Understand the influence of COVID-19 on the Machine Learning in Medical Imaging Market with our analysts monitoring the situation across the globe. Request Now

The market research report on the global Machine Learning in Medical Imaging industry provides a comprehensive study of the various techniques and materials used in the production of Machine Learning in Medical Imaging market products. Starting from industry chain analysis to cost structure analysis, the report analyzes multiple aspects, including the production and end-use segments of the Machine Learning in Medical Imaging market products. The latest trends in the pharmaceutical industry have been detailed in the report to measure their impact on the production of Machine Learning in Medical Imaging market products.

Leading key players in the Machine Learning in Medical Imaging market are Zebra, Arterys, Aidoc, MaxQ AI, Google, Tencent, Alibaba

Get sample of this report @ https://grandviewreport.com/sample/21159

Product Types:, Supervised Learning, Unsupervised Learning, Semi Supervised Learning, Reinforced Leaning

By Application/ End-user:, Breast, Lung, Neurology, Cardiovascular, Liver

Regional Analysis For Machine Learning in Medical ImagingMarket

North America(the United States, Canada, and Mexico)Europe(Germany, France, UK, Russia, and Italy)Asia-Pacific(China, Japan, Korea, India, and Southeast Asia)South America(Brazil, Argentina, Colombia, etc.)The Middle East and Africa(Saudi Arabia, UAE, Egypt, Nigeria, and South Africa)

Get Discount on Machine Learning in Medical Imaging report @ https://grandviewreport.com/discount/21159

This report comes along with an added Excel data-sheet suite taking quantitative data from all numeric forecasts presented in the report.

Research Methodology:The Machine Learning in Medical Imagingmarket has been analyzed using an optimum mix of secondary sources and benchmark methodology besides a unique blend of primary insights. The contemporary valuation of the market is an integral part of our market sizing and forecasting methodology. Our industry experts and panel of primary members have helped in compiling appropriate aspects with realistic parametric assessments for a comprehensive study.

Whats in the offering: The report provides in-depth knowledge about the utilization and adoption of Machine Learning in Medical Imaging Industries in various applications, types, and regions/countries. Furthermore, the key stakeholders can ascertain the major trends, investments, drivers, vertical players initiatives, government pursuits towards the product acceptance in the upcoming years, and insights of commercial products present in the market.

Full Report Link @ https://grandviewreport.com/industry-growth/Machine-Learning-in-Medical-Imaging-Market-21159

Lastly, the Machine Learning in Medical Imaging Market study provides essential information about the major challenges that are going to influence market growth. The report additionally provides overall details about the business opportunities to key stakeholders to expand their business and capture revenues in the precise verticals. The report will help the existing or upcoming companies in this market to examine the various aspects of this domain before investing or expanding their business in the Machine Learning in Medical Imaging market.

Contact Us:Grand View Report(UK) +44-208-133-9198(APAC) +91-73789-80300Email : [emailprotected]

See the rest here:
Machine Learning in Medical Imaging Market 2020 : Analysis by Geographical Regions, Type and Application Till 2025 | Zebra, Arterys, Aidoc, MaxQ AI -...

Machine learning algorithms will learn patterns from the past data and predict trends and best price. These algorithms can predict the best price, discount price and promotional price based on competition, macroeconomic variables, seasonality etc.

To find out the correct pricing in real-time retailers follow the following steps:

Gather input data

In order to build a machine learning algorithm, retailers collect various data points from the customers. These are:

Transactional data

This includes the sales history of each customer and the products, which they have bought in the past.

Product description

The brands, product category, style, photos and the selling price of the previously sold products are collected. Past promotions and campaigns are also analysed to find the effect of price changes on each category.

Customer details

Demographic details and customer feedback are gathered.

Competition and inventory

Retailers also try to find the data regarding the price of products sold by their competitors and supply chain and inventory data.

Depending on the set of key performance indicators defined by the retailers, the relevant data is filtered.

For every industry, pricing would involve different goals and constraints. In terms of the dynamic nature, the retail industry can be compared to the casino industry where machine learning is involved inonline live dealer casino games too.

Like casinos, retail also has the target of profit maximisation and retention of customer loyalty. Each of these goals and constraints can be fed to a machine learning algorithm to generate dynamic prices of products.

Visit link:
news and analysis for omnichannel retailers - Retail Technology Innovation Hub

This is a problem if we want AIs to be trustworthy. Thats why Diffbot takes a different approach. It is building an AI that reads every page on the entire public web, in multiple languages, and extracts as many facts from those pages as it can.

Like GPT-3, Diffbots system learns by vacuuming up vast amounts of human-written text found online. But instead of using that data to train a language model, Diffbot turns what it reads into a series of three-part factoids that relate one thing to another: subject, verb, object.

Pointed at my bio, for example, Diffbot learns that Will Douglas Heaven is a journalist; Will Douglas Heaven works at MIT Technology Review; MIT Technology Review is a media company; and so on. Each of these factoids gets joined up with billions of others in a sprawling, interconnected network of facts. This is known as a knowledge graph.

Knowledge graphs are not new. They have been around for decades, and were a fundamental concept in early AI research. But constructing and maintaining knowledge graphs has typically been done by hand, which is hard. This also stopped Tim Berners-Lee from realizing what he called the semantic web, which would have included information for machines as well as humans, so that bots could book our flights, do our shopping, or give smarter answers to questions than search engines.

A few years ago, Google started using knowledge graphs too. Search for Katy Perry and you will get a box next to the main search results telling you that Katy Perry is an American singer-songwriter with music available on YouTube, Spotify, and Deezer. You can see at a glance that she is married to Orlando Bloom, shes 35 and worth $125 million, and so on. Instead of giving you a list of links to pages about Katy Perry, Google gives you a set of facts about her drawn from its knowledge graph.

But Google only does this for its most popular search terms. Diffbot wants to do it for everything. By fully automating the construction process, Diffbot has been able to build what may be the largest knowledge graph ever.

Alongside Google and Microsoft, it is one of only three US companies that crawl the entire public web. It definitely makes sense to crawl the web, says Victoria Lin, a research scientist at Salesforce who works on natural-language processing and knowledge representation. A lot of human effort can otherwise go into making a large knowledge base. Heiko Paulheim at the University of Mannheim in Germany agrees: Automation is the only way to build large-scale knowledge graphs.

To collect its facts, Diffbots AI reads the web as a human wouldbut much faster. Using a super-charged version of the Chrome browser, the AI views the raw pixels of a web page and uses image-recognition algorithms to categorize the page as one of 20 different types, including video, image, article, event, and discussion thread. It then identifies key elements on the page, such as headline, author, product description, or price, and uses NLP to extract facts from any text.

Every three-part factoid gets added to the knowledge graph. Diffbot extracts facts from pages written in any language, which means that it can answer queries about Katy Perry, say, using facts taken from articles in Chinese or Arabic even if they do not contain the term Katy Perry.

Browsing the web like a human lets the AI see the same facts that we see. It also means it has had to learn to navigate the web like us. The AI must scroll down, switch between tabs, and click away pop-ups. The AI has to play the web like a video game just to experience the pages, says Tung.

Diffbot crawls the web nonstop and rebuilds its knowledge graph every four to five days. According to Tung, the AI adds 100 million to 150 million entities each month as new people pop up online, companies are created, and products are launched. It uses more machine-learning algorithms to fuse new facts with old, creating new connections or overwriting out-of-date ones. Diffbot has to add new hardware to its data center as the knowledge graph grows.

Researchers can access Diffbots knowledge graph for free. But Diffbot also has around 400 paying customers. The search engine DuckDuckGo uses it to generate its own Google-like boxes. Snapchat uses it to extract highlights from news pages. The popular wedding-planner app Zola uses it to help people make wedding lists, pulling in images and prices. NASDAQ, which provides information about the stock market, uses it for financial research.

Adidas and Nike even use it to search the web for counterfeit shoes. A search engine will return a long list of sites that mention Nike trainers. But Diffbot lets these companies look for sites that are actually selling their shoes, rather just talking about them.

For now, these companies must interact with Diffbot using code. But Tung plans to add a natural-language interface. Ultimately, he wants to build what he calls a universal factoid question answering system: an AI that could answer almost anything you asked it, with sources to back up its response.

Tung and Lin agree that this kind of AI cannot be built with language models alone. But better yet would be to combine the technologies, using a language model like GPT-3 to craft a human-like front end for a know-it-all bot.

Still, even an AI that has its facts straight is not necessarily smart. Were not trying to define what intelligence is, or anything like that, says Tung. Were just trying to build something useful.

Go here to see the original:
This know-it-all AI learns by reading the entire web nonstop - MIT Technology Review

Python was released in the 1990s as a general-purpose programming language. Despite its clean syntax, the exposure Python got in its first decade wasntencouraging, and itdidnt really find inroads into the developers workspace. Perl was the first choice scripting language and Java had established itself as the go-to in the object-oriented programming arena. Of course, any language takes time to mature and only gets adopted when its better suited to a task than the existing tools.

For Python, that time first arrived during the early 2000s when people started realizing it has an easier learning curve than Perl and offers interoperability with other languages. This realization led to a larger number of developers incorporating Python into their applications. The emergence of Django eventually led to the doom of Perl, and Python started gaining more momentum. Still, it wasnt even close in popularity to Java and JavaScript, both of which were newer than Python.

Fast forward to the present, and Python has trumped Java to become the second-most-popular language according to the StackOverflow Developer Survey 2019. It was also the fastest-growing programming language of the previous decade. Pythons rise in popularity has a lot to do with the emergence of big data in the 2010s as well asdevelopments in machine learning and artificial intelligence. Businesses urgently required a language for quick development with low barriers of entry that could help manage large-scale data and scientific computing tasks. Python was well-suited to all these challenges.

Besides having those factors in its favor, Python was an interpreted language with dynamic typing support. More importantly, it had the backing of Google, whod invested in Python for Tensorflow, which led to its emergence as the preferred language for data analysis, visualization, and machine learning.

Yet, despite the growing demand for machine learning and AI at the turn of this decade, Python wont stay around for long. Like every programming language, it has its own set of weaknesses. Those weaknesses make it vulnerable to replacement by languages more suited to the common tasks businesses ask of them. Despite the presence of R, the emergence of newer languages such as Swift, Julia, and Rust actually poses a bigger threat to the current king of data science.

Rust is still trying to catch up with the machine learning community, and so I believe Swift and Julia are the languages that will dethrone Python and eventually rule data science. Lets see why odds are against Python.

All good things come at a cost, and Pythons dynamically typed nature is no exception. It hampers developers, especially when running the code in production. Dynamic typing that makes it easy to write code quickly without defining types increases the risk of running into runtime issues, especially when the codebase size increases. Bugs that a compiler would easily figure out could go unidentified in Python, causing hindrances in production and ultimately slowing the development process in large-scale applications.

Worse, unlike compiled code, Pythons interpreter analyzes every line of code at execution time. This leads to an overhead that causes a significantly slower performance when compared to other languages.

Julia allows you to avoid some of these problems. Despite being dynamically typed, it has a just-in-time compiler. The JIT compiler either generates the machine code right before its executed or uses previously stored, cached compilations, which makes it as performant as statically typed languages. More importantly, it has a key feature known as multiple dispatch thatis like function overloading of OOPs, albeit at runtime. The power of multiple dispatch lies in its ability to handle different argument types without the need to create separate function names or nested if statements. This helps in writing compact code, which is a big win in numeric computations since unlike Python, you can easily scale solutions to deal with all types of arguments.

Even better, Swift is a statically typed language and is highly optimized due to its LLVM (Low-Level Virtual Machine) compiler. The LLVM makes it possible to quickly compile into assembly code, making Swift super-efficient and almost as fast as C. Also, Swift boasts better memory safety and management tools known as Automatic Reference Counting. Unlike garbage collectors, ARC is a lot more deterministic as it reclaims memory whenever the reference count hits zero.

As compiled languages that offer type annotations, Swift and Julia are a lot faster and robust for development than Python. That alone might be enough to recommend them over the older language, but there are other factors to consider as well.

If slowness was not the most obvious drawback of Python, the language also has limitations with parallel computing.

In short, Python uses GIL (Global Interpreter Lock), which prevents multiple threads from executing at the same time in order to boost the performance of single threads. This process is a big hindrance because it means that developers cannot use multiple CPU cores for intensive computing.

I agree withthe commonplace notion that were currently doing finewhen leveraging Pythons interoperability with C/C++ libraries like Tensorflow and PyTorch. But a Python wrapper doesnt solve all debugging issues. Ultimately, when inspecting the underlying low-level code, were falling back on C and C++. Essentially, we cant leverage the strengths of Python at the low level,which puts it out of the picture.

This factor will soon play a decisive role in the fall of Python and rise of Julia and Swift. Julia is a language exclusively designed to address the shortcomings of Python. It primarily offers three features: coroutines (asynchronous tasks), multi-threading, and distributed computing all of which only show the immense possibilities for concurrent and parallel programming. This structure makes Julia capable of performing scientific computations and solving big data problems at a far greater speed than Python.

On the other hand, Swift possesses all the tools required for developing mobile apps and has no problems with parallel computing.

Despite the disadvantages it has with respect to speed, multi-threading, and type-safety, Python still has a huge ecosystem that boasts an enormous set of libraries and packages. Understandably, Swift and Julia are still infants in the field of machine learning and possess only a limited number of libraries. Yet, their interoperability with Python more than compensates for the lack of library support in Julia and Swift.

Julia not only lets programmers use Python code (and vice-versa), but also supports interoperability with C, R, Java, and almost every major programming language. This versatility would certainly give the language a good boost and increase its chances of a quick adoption among data scientists.

Swift, on the other hand, provides interoperability with Python with the PythonKit library.The biggest selling point for Swift (which has an Apple origin) is a strong support its been getting from Google, who fully backed Python decades ago. See how the tables have turned!

Also, the fact that the creator of Swift, Chris Lattner, is now working on Googles AI brain team just shows that Swift is being seriously groomed as Pythons replacement in the machine learning field.The Tensorflow team investing in Swift with their S4TF project only further proves that the language isnt merely regarded as a wrapper over Python. Instead Swift, thanks to its differential programming support and ability to work at a low level like C, will potentially be used to replace the underlying deep learning tools.

As the size of data continues to increase, Pythons Achilles heel will be soon found out. Gone are the days when ease of use and ability to write code quickly mattered. Speed and parallel computing are the name of the game and Python, which is a more general-purpose language, will no longer solve that problem. Inevitably, it will fade away while Julia and Swift seem like the candidates to take over the reins.

Its important to note that Python as a programming language wont disappear any time soon. Instead, itll only take a backseat in data science as the languages that are more specifically designed for deep learning will rule.

View post:
Python Is About to Get the Squeeze - Built In