Moo…
There is this very intriguing Enterprise Car Rental commercial on TV (I’m old school and still watch my commercials on a television) that shows a glimpse into future driving behaviors enabled by autonomous vehicles. The autonomous vehicles appear to be randomly crossing the intersection, with cars driving through the intersection without stopping and waiting for the other cars. Because these cars are literally talking to each other – without a human to slow down or confuse the conversation – they are able to synchronize their crossing of the intersection without today’s ineffective human-necessitated process of negotiating stop lights.
And we know that humans need strict controls to govern their driving process because many (most?) humans suck at complex tasks such as driving a car. Read More
Monthly Archives: May 2019
Big Data and the Rise of Augmented Intelligence: Sean Gourley at TEDxAuckland
3 Things You Need To Know About Augmented Intelligence
Steve Jobs was fond of saying that the key to Apple’s success has been a marriage of technology and the liberal arts. What he meant by this is that innovation emerges at the intersection of art and technology– rather than either alone.
In fact, it was Douglas Engelbart who first understood the importance of computer technologies in bootstrapping human capabilities and augmenting human creativity. Building on Engelbart’s thinking, we need to begin to better comprehend the challenges of machine intelligence in the context of human creativity and innovation— particularly with regard to our systems of learning and education. Read More
Artificial intelligence can now make art. Artists, don’t panic.
Music lovers gathered in a London theater one night in March to take part in an unusual event: half classical concert, half futuristic experiment. Their task was to listen to music that had been composed partly by Bach and partly by artificial intelligence— and try to guess which parts were which. Throughout the performance, the audience members voted by holding up cards with a blue human face on one side and a red robot face on the other.
“It was quite shocking,” Marcus du Sautoy, an Oxford mathematician who masterminded the event, told me. “There were moments when I think Bach would have turned in his grave! Moments when Bach was playing and people were saying it was the AI.” Read More
Revisiting Unreasonable Effectiveness of Data in Deep Learning Era
The success of deep learning in vision can be attributed to: (a) models with high capacity; (b) increased computational power; and (c) availability of large-scale labeled data. Since 2012, there have been significant advances in representation capabilities of the models and computational capabilities of GPUs. But the size of the biggest dataset has surprisingly remained constant. What will happen if we increase the dataset size by 10× or 100×? This paper takes a step towards clearing the clouds of mystery surrounding the relationship between ‘enormous data’ and visual deep learning. By exploiting the JFT-300M dataset which has more than 375M noisy labels for 300M images, we investigate how the performance of current vision tasks would change if this data was used for representation learning. Our paper delivers some surprising (and some expected) findings. First, we find that the performance on vision tasks increases logarithmically based on volume of training data size. Second, we show that representation learning (or pretraining) still holds a lot of promise. One can improve performance on many vision tasks by just training a better base model. Finally, as expected, we present new state-of-the art results for different vision tasks including image classification, object detection, semantic segmentation and human pose estimation. Our sincere hope is that this inspires vision community to not undervalue the data and develop collective efforts in building larger datasets. Read More
The Unreasonable Effectiveness of Data
Eugene Wigner’s article “The Unreasonable Effectiveness of Mathematics in the Natural Sciences”1 examines why so much of physics can be neatly explained with simple mathematical formulas such as f = ma or e = mc2. Meanwhile, sciences that involve human beings rather than elementary particles have proven more resistant to elegant mathematics. Economists suffer from physics envy over their inability to neatly model human behavior. An informal, incomplete grammar of the English language runs over 1,700 pages.2 Perhaps when it comes to natural language processing and related fields, we’re doomed to complex theories that will never have the elegance of physics equations. But if that’s so, we should stop acting as if our goal is to author extremely elegant theories, and instead embrace complexity and make use of the best ally we have: the unreasonable effectiveness of data.
One of us, as an undergraduate at Brown University, remembers the excitement of having access to the Brown Corpus, containing one million English words.3 Since then, our fi eld has seen several notable corpora that are about 100 times larger, and in 2006, Google released a trillion-word corpus with frequency counts for all sequences up to five words long.4 In some ways this corpus is a step backwards from the Brown Corpus: it’s taken from unfiltered Web pages and thus contains incomplete sentences, spelling errors, grammatical errors, and all sorts of other errors. It’s not annotated with carefully hand-corrected part-of-speech tags. But the fact that it’s a million times larger than the Brown Corpus outweighs these drawbacks. A trillion-word corpus—along with other Web-derived corpora of millions, billions, or trillions of links, videos, images, tables, and user interactions—captures even very rare aspects of human behavior. So, this corpus could serve as the basis of a complete model for certain tasks—if only we knew how to extract the model from the data. Read More
A robot named Heliograf got hundreds of stories published last year
Robots are taking our jobs, no doubt about it. Just in the past year we’ve seen barista robots, fast-food robots, pizza delivery robots, and even a robot conducting a symphony orchestra. But robots can’t replace journalists, right? The dogged reporters, members of the vaunted Fourth Estate, the men and women who bring us the news stories we read every day?
Think again. It’s happening, and odds are that you’ve been reading stories created by artificial intelligence in local and world news already.
A year ago, the Washington Post introduced Heliograf, an in-house program that automatically generates short reports for their live blog. It was first used during the Rio Olympics to provide information such as the results of medal events for services like Alexa. At that time Sam Han, engineering director of data science, said, “The next challenge is to broaden the subjects covered, deepen the kind of analysis possible and identify potential stories for our newsroom.” Read More
The Washington Post’s robot reporter has published 850 articles in the past year
It’s been a year since The Washington Post started using its homegrown artificial intelligence technology, Heliograf, to spit out around 300 short reports and alerts on the Rio Olympics. Since then, it’s used Heliograf to cover congressional and gubernatorial races on Election Day and D.C.-area high school football games, producing stories like this one.
The Associated Press has used robots to automate earnings coverage, while USA Today has used video software to create short videos. But media executives are more excited about AI’s potential to go beyond rote reporting. Jeremy Gilbert, director of strategic initiatives at the Post, shared what the paper has learned so far from robo reporting and what it’s still trying to figure out.
In its first year, the Post has produced around 850 articles using Heliograf. That included 500 articles around the election that generated more than 500,000 clicks — not a ton in the scheme of things, but most of these were stories the Post wasn’t going to dedicate staff to anyway. For the 2012 election, for example, the Post did just 15 percent of what it generated in 2016. Read More
Meet Bertie, Heliograf And Cyborg, The New Journalists On The Block
An article in The New York Times, “The rise of the robot reporter”, discusses the new machine learning tools such as Forbes’ Content Management System called Bertie, The Washington Post’s Heliograf, Bloomberg’s Cyborg, and others used by Reuters, the Associated Press and The Guardian for a range of tasks in their newsrooms, providing further insight into how robots are taking over a growing number of jobs.
As the article points out, a number of media, notably those with a financial focus, now use algorithms to analyze data such as quarterly earnings reports and that can then be used to chart their progress over time, detect anomalies, as well as writing up stories. Machine learning can now be put to a growing number of uses in a newsroom. I have seen journalists use powerful analytical tools to analyze graphs, write and document their articles, adding information from different sources, as well as identifying information that may come from suspicious sources. It is hardly surprising that Bloomberg and Reuters now compete to provide real time financial information with the same hedge funds that were previously their clients but that now have access to the same means and resources. Read More
The “smarter” wall: how drones, sensors, and AI are patrolling the border
In an era of increasingly polarized politics, there are few issues as divisive as President Trump’s proposal to build a physical wall across part of the 2,000-mile US-Mexico border.
The Trump administration has argued that the border wall is a necessary deterrent to drug smugglers and immigrants seeking to enter the country unlawfully. It says unchecked immigration is a national security crisis, and one that needs to be addressed.
Critics, meanwhile, argue that the wall is a wildly expensive, ineffective, and misdirected effort. The actual crisis, they say, is a humanitarian one worsened by Trump’s restrictive immigration policies — about refugees seeking lawful entry into the US to flee violence and poverty in their home countries.
But there is another kind of border wall increasingly being talked about — one that proponents pitch as being less costly, less disruptive, and less politically controversial than a physical barrier: a so-called “smart wall.” Read More
Rick's Cafe AI 