How GANs can turn AI into a massive force

ai-robot-face

 

Deep learning models can already achieve state-of-the-art results in some applications, but their capabilities are still limited. Unlike humans, deep learning models are unable to handle minor changes, and hence can only be applied for specific and narrowly defined tasks.

Consider this conversation of what might be the most sophisticated negotiation software on the planet, which occurred between two AI agents developed at Facebook:

Bob: “I can can I I everything else.”

Alice: “Balls have zero to me to me to me to me to me to me to me to me to.”

At first, they were speaking in plain old English, but researchers realized they forgot to include a reward for sticking to the language. So, the AI agents began to diverge, eventually rearranging legible words into seemingly nonsensical (but, in their perspective, highly efficient) sentences. They invented their own codewords, abbreviations, and structures.

This phenomenon is observed again and again and again.

A vanguard AI technology that can learn, recognize, and generate information on a nearly human level doesn’t exist yet, but we have taken steps toward that direction.

What are generative adversarial networks (GANs)?

Generally intelligent systems must be able to generalize from limited data and learning causal relationships. In 2016, Ian Goodfellow, a fellow at Google Brain, suggested using generative adversarial networks (GANs) as an alternative unsupervised machine learning method. This aimed to address many of the ailing points of the existing methods.

GANs consist of two deep neural networks: generator and discriminator. The generator’s goal is to create data samples that are so indistinguishable to the real ones. The discriminator’s goal is to identify which of the generator’s data samples are real and which are fake.

These two networks compete against each other in a zero-sum game (i.e. one’s loss implies another’s win). Both networks would then become stronger in a relatively short period of time.

gan-chart

Backpropagation is used to update the model parameters and train the neural networks. Over time, the networks learn many features of the provided data. To create realistic forged samples, the generator needs to learn the data’s features and patterns, while the discriminator does the same to correctly distinguish between real and fake samples.

GANs are thus able to overcome the above weaknesses by training (i.e. playing) neural networks against each other, thus learning from each other (which necessitates less data) and eventually performing better in a broader range of problems.

Applications of GANs

There are several types of GANs, and some of its most obvious applications include high-resolution or interactive image generation/blendingimage inpaintingimage-to-image translation, abstract reasoning, semantic segmentation, video generation, and text-to-image synthesis, among others.

The video game industry is the first area of entertainment to start seriously experimentingusing AI to generate raw content. There’s a huge cost incentive to invest in video game development automation given the US$300 million+ budget of modern AAA video games.

GANs have also been used for text, with less success⏤a bot developed to speak like Friedrich Nietzsche started to speak in a manner similar to the philosopher, but the sentences did not make sense. GANs for voice applications are able to reproduce a given text string to life-like voices with approximately 20 minutes of voice samples, such as these popular impersonations of American presidents Donald Trump and Barack Obama. In the near future, videos can likely be generated just by providing a script.

Goodfellow and his colleagues used GANs for image generation, recognition, and classification by teaching one of the networks to create images of handwritten digits (humans were not able to distinguish real handwritten digits). They also trained a neural network to create images of objects, which humans could only differentiate (from real ones) 78.7 percent of the time. Below are some sample images of faces created entirely by deep convolutional GANs.
face-samples-gan

Despite all the above achievements, GANs still have weaknesses:

  • Instability (the generator and the discriminator losses keep oscillating) and non-convergence (to optimum) of the objective function in GANs
  • Mode collapse (this happens when the generator doesn’t produce diverse images or information)
  • The possibility that either the generator or the discriminator becomes too strong as compared to the others during training
  • The possibility that either the generator or the discriminator never learns beyond a certain point

An existential threat

Do GANs and AI in general pose an existential threat to humanity? Elon Musk thinks so. Since 2014, he has been advocating adoption of AI regulations by authorities around the world. Recently, he reiterated the urgent need to be proactive in regulation.

“AI is a fundamental risk to the existence of human civilization,” Musk tells US politicians recently.

His concerns stem from the rapid developments related to GANs, which might push humanity toward the inception of artificial general intelligence. While AI regulations may serve as safeguards, AI is still far from the fictitious depictions seen frequently in Hollywood sci-fi movies.

(By the way, Facebook ultimately opted to require its negotiation bots to speak in plain old English.)

Here are some recommended resources for GAN:

This article originally appeared on Tech in Asia.

Failure parallels between Mao Tse-Tung and Facebook

What do an alleged 70 million death of Chinese population in peacetime and an online social network boasting 955 million members have in common? While Mao’s life and work were considered inspirational at the time, so is modern phenomenon Facebook sweeping domains and redefining frontiers of science, culture and technology. And there are some parallels that can’t and shouldn’t be ignored.

1) Put appearance and rhetoric of doing things for public good while in reality following his/its own agenda

  • Mao: Mao got into the Long March in 1934 not due to any ideological reasons. He was sick and was going to be left behind by Communist forces under the pursuit of Nationalists because of his activities undermining communist goals and his disobedience to orders from the Soviet Union. A combination of putting an appearance of a patriot fighting for the communist cause and a trickery of positioning himself at the point where the Communist army was to exit from, he managed to get himself picked up by the army, soon passing through a bridge, which bore him into the history.
  • Facebook:  The most valuable asset on Internet is data. User data, to be more specific, as it can be used (in some cases sold) for conducting all kinds of research – it’s invaluable for any organization that has part or all of its customer base online. Facebook has been introducing number of usually unpopular privacy setting changes that, for most cases, used its member information for its own agenda. A telling sign is its recent change of replacing the word “privacy” with “data use.”

2) Decisions are taken unilaterally and without concern for those affected – and when it goes out of control, some measures are taken in guise of public appearances and apologies

  • Mao: Mao, in the framework of his “Great Leap Forward.” ordered the nation to make steel. He made a target of 10.7 million tons of steel (as the previous year’s output was 5.3 millon tons, Mao asked his metallurgy minister whether that could be doubled for current year. The “yes-man” obliged). Steel mills and related industries like coal mines were ordered to go flat out to speed up the production. Rules and common sense were cast aside. Equipment overworked to the point of breakdown. In first few months alone, 30,000 workers were killed in various accidents. Realizing, the existing mills wouldn’t be able to meet the target, Mao ordered the nation to build “backyard furnaces,” which some 90 million people were forced to build. To feed those furnaces, the nation was coerced into donating virtually every scrap of metal they had including farm tools and even kitchen utensils. By the end of 1958, the target of 10.7 million tons was reached, but only 40% of which was good steel (all produced by mills). Estimates put death toll of starvation and overwork during the Leap (which lasted 4 years) to about 38 million.
  • Facebook: Not only it has been playing with privacy settings of its members in order to gain and retain user information, Facebook has also been, rather quietly, granting access to all kinds of user information to 3rd party applications. As its business becomes more and more app-oriented – unsurprisingly so, as apps account for 30% of its revenues for the moment – Facebook has been making changes, that would grant a more seamless, easy access of user info to apps. Check this post for some details. With subtle design changes, it went, for example, from a two-button procedure of granting (or not) access to that of one-button. You click an invitation from a friend and, unbeknownst to you, end up giving access to the app to gain your information. Or that in old design Facebook explained what sort of information an app could use. No longer so. Facebook, considering much past opposition to measures less significant for usage of member information, unilaterally decided to introduce changes that would benefit its bottom line. And last but not least, “We’re going public for our employees and our investors,” as Zuckerberg wrote in his IPO letter. Translation: we don’t care about our customers/members – we go public so our employees get the share of our riches and our investors feel satisfied.

3) No care or thought about needs and aspirations of subjects/members

  • Mao: On Feb 27, 1957, Mao delivered a 4-hour speech to the rubber-stamp Supreme Council announcing that he was inviting criticism of the Communist Party. The Party, he said, needed to be accountable and ‘under supervision.’ He sounded reasonable criticizing Stalin for his ‘excessive’ purges, and giving the impression there wouldn’t be any of those in China. In this context he cited the adage, “Let the hundred flowers bloom.” Few guessed it was a trap setup to invite people to speak, so that he could use what they said as an excuse ti victimize them. His main target were intellectuals and educated, those mostly likely to speak out. Most of criticism never reached public. Instead, Mao drafted a secret circular in which he branded between 1% and 10% of ‘intellectuals’ (those who spoke out amounting to about 5 million) as ‘Rightists.’ Denunciations, labor camps and executions followed.
  • Facebook:  Among other unilateral and unsolicited changes, Facebook decided to not only introduce its own email system, but silently replace members’ emails with its own @facebook.com email (only noticed by bloggers during a weekend). The step could have been well accepted, it Facebook took care of showing consideration of its members’ opinion or warned in advance. Backlash followed. Or its previous changes to News Feed, which ushered in a plethora of groups/videos with a theme “Change Facebook Back to Normal.”  As an example, Facebook group and a video demonstrating how to change it to normal are still up.

4) Everything can be and is sacrificed for achievement of set goals – no morality or etiquete or qualms

  • Mao: In 1953, Mao proposed a program to industrialize China. His obsession was to enter the league with Soviet Union and the US in economic terms. At that time, he was told that the program could be completed in 10-15 years. In 1957, celebrating the 40th anniversary of October, Khrushchev presented a target of surpassing the US in 15 tears. Mao was inspired and claimed China could surpass the UK in 15 years. And as China was already getting nuclear other cutting-edge technological help and knowhow from Soviets, Mao was told in 1958 that his industrialization program could be completed in 5 years instead. He fancied he could achieve this in one big leap, calling it “Great Leap Forward.” It boiled down to excessive food production and extraction from the people. Mao was squeezing his nation to produce fantastic amounts of food (grain, rice, pork, etc), which was exported in exchange for technology. Food was also used as a raw material for the nuclear program. Grain, for example, was turned into pure alcohol and used in missile tests, each of which consumed about 10 million kg of grain. Mao thought to industrialize form within by ordering construction of irrigation systems (dams, reservoirs, canals). Starting in 1958 for over 4 years, about 100 million people were coerced into such projects, and as he wanted instant results his slogan was “”Survey, Design and Execute Simultaneously.” Results are not hard to guess.
  • Facebook:  As Facebook moved ahead with its fresh design, it didn’t forget to, without asking opinions of members, make it easier to access the past activities with introduction of its Timeline. Timeline is conceived with ads, sponsored stories and marketing data gathering to be much easier, i.e. playing a hand for Facebook’s bottom line.  Launched in September 2011, Facebook’s Timeline claimed to be the online book of members’ lives showing events on a yearly bases. Initially, Timeline was an optional feature. But once accepted, there was no going back, which suggested the imminent (and permanent) nature of the change. By the end of January 2012, not content with the fact that many of its members still preferred the older design – Sophos’ polls showed about 52% worried about and only 8% liked Timeline – Facebook simply stated that in a week time everyone will have a Timeline. A flurry of petitions, anti-Timeline groups and a lawsuit ensued.

Let me be clear. I am not a hater of Facebook. I have a special admiration for Mark Zuckerberg and for Facebook. I can’t however discount all the above developments nor the logic of how Facebook advances towards its claimed goals and objectives.

I only wish that Facebook becomes more open, transparent and forthcoming, and what’s more important, more considerate towards its ever-increasing member base. If not, then, as the saying goes, “the higher you climb, the further you fall.

I bet many would hate to see Facebook fail/fall – as Mao fell.

historic diplomacy and modern e-diplomacy in America

Modern diplomacy origins trace to 13th-century Milan. Sforza’s Milan saw many modern diplomatic traditions begin (ambassador credential presentation,…) and was first to practice balance of power.

One of first American diplomats was Franklin, who secured French support against England, leading to Treaty of Alliance, a turning point in Revolutionary War.

But nowadays, Obama’s America is less popular in MENA than Bush’s. In Turkey, its popularity plummeted from 52% in 2000 to 10% in 2011.

Obama-style solution? E-diplomacy: US State Department operates 230 Facebook accounts, 80 Twitter feeds, 55 YouTube channels, according to this research, leveraging tools like Klout, Facebook Insights,… for influence and measurement.

Will Google fail… again?

If you work just for money, you’ll never make it, but if you love what you’re doing and you always put the customer first, success will be yours.

Ray Kroc, founder of McDonald’s said. Google founders loved their search engine. Today Google seems to love money more than anything.

Google Catalogs, Google Answers, Google Wave and even the most recent, hyped up Google Buzz feature on a growing list of Google flops, with, on average, 200 projects that are being worked upon at any time at Google.

A seeming common denominator of all its failed attempts is its chase of existing and successful business models or competitors. Google Wave was to reinvent email; Google Buzz was to be a direct response to Twitter; Google Answers was to counter Yahoo! Answers and so on.

Since 2001, Google embarked on acquisition of middle and small size companies, in its bid to enhance the range of its services both vertically and horizontally. More than 80 acquired companies and 10 years later, yet its strategy, approach and mentality have hardly changed.

In a certain sense, Google stopped innovating. Many of its failures could be somewhat explained away by looking at how it tries to go about chasing others’ success. An enlightening interview with a Google exec revealed some crucial points –  scope of work, team size and usage of infrastructure, etc. – of how Google cannot, for example, build an Instagr.am equivalent.

And now Groupon. Google’s unsuccessful story of trying to buy it for USD 6 billion did not finish there. It now decided to come up with its own answer. A déjà vu?

Of course, Google did and does good things. Adwords, Adsense, Analytics, Android and its transformation of online advertising using acquired DoubleClick technology, have done and will continue bringing value to businesses and end users.

But, is the value offered by Google justified by the growing number and impact of its flops?

It all boils down to a company’s DNA. Google’s DNA is search, and it built around it, growing and becoming successful. Now it tries to “go out” of its DNA and diversify, not an unusual drive for a company of its size and track record. It needs t keep in mind, however, that similar attitude brought down other big and successful companies in the past. Instead, what it could do is to make its own search smarter and richer (in relevancy, targetting and search result contents).

The usual question that Google and other failed/successful entrepreneurs/businesses ask themselves daily is, “Should we Innovate or Copy?” The word “Innovation” became a cliché, despite the fact that innovation wars lead nowhere and hurt everyone.

Perhaps it is time we start to mInnovate.

Does your business need a model to be successful?

What is business model?

Many people assume that as long as they have a great product or amazing service, success is a guarantee. This is an illusion, especially in the rat-race of the 21st century. Without a  solid and well-thought business model, organizations/firms struggle to grow or even to survive.

Let’s define few terms, before continuing. Traditional business model has six components: value proposition, market segment, value chain structure, revenue generation/margins, position in value network and competitive strategy. The successful ones, however, tend to have the following three common features: offer a unique value, are difficult to imitate, and meet/create an (untapped) market demand.  For the Internet, online business models became essential for online companies and traditional ones expanding their reach into the online. Online business models form an ecosystem comprising of four parts: companycustomerproduct and experience.

Some,  Dot-com bust exposed an armada of good “look-and-feel” online companies, which had no working/sustainable business model and were a mere fluff in the market.  But it is not only dot-coms that fail to conceive a deeper value in their offer. Some traditional businesses (such as newspapers), which branch out into online, blur their existing business models (by charging for access to obituaries while charging fee from those who choose to put them in the paper).

Is there a “right” business model for a company?

Some companies, both purely online or traditional, finding a business model that works, try to protect it as as Netflix did when it filed and was granted a patent for its business models (subsequently suing Blockbuster for patent infringement). An example of a very successful business model was at Xerox. It created the world’s first automatic paper copier. Xerox saw that with the high expense for each machine, there was no room for a substantial profit by simply selling machines. Their initial business model was based on the idea of leasing the machines and charging a fee per copy made. Because of its innovative and strong marketing activities, Xerox brand-name gradually became a verb “Xerox” which stands for copying paper (like Google stands for search). Recently, by incorporating environment-friendly concepts into their business model, Xerox started producing earth friendly designs that helped it save USD 2 billions in last 2 years.

On the Internet, where some social networks take a gigantic chunk of users’ online “time” are not, generally speaking, a commercial success and the investors are generally apprehensive about the business model they pursue – Facebook (which makes money from brand ads, partnerships, paid applications like virtual goods and performance ads) and Twitter that are still with us contrary to countless others that did not survive. Some suggested Twitter – it recently debuted its promoted tweets to build business models around data mining and performing trend analysis, feed advertising, SMS ads, and subscriptions, integrating contextual ads, apps, and pages or even selling/suggesting friends. Lastly, few envision a tiered/freemium model whereby “basic” Twitter remains free but a premium service (a version of promoted tweets with more flexibility, options and cost structures) is offered for online brand builders and commercial businesses. Indeed, the whole business model/monetization topic for Twitter (and for Facebook) are so hot a topic that there are few dedicated topical channels as this one on Business Exchange.

Society + business = ?

Few days ago, the US state of Maryland has become the first state in America to recognize a new type of socially responsible corporation that can consider the public good in addition to shareholder obligations in business decisions. Until now the term “corporate social responsibility” was a badge for innovative, brave and socially conscious businesses such as Ben & Jerry’s whose social activism is well-known. There was no law supporting what they did. From now on, businesses that become (by amending their charters) “benefit corporations” may consider factors like employee interests, the environment and promoting arts and sciences, as well as shareholder interest.

And last but not least, what have you heard of a company called TOMS? Even if you have not, you will, very soon because they just completed giving away a million shoes. Their business model is simple: for every pair of shoes bought, they give one pair away. TOMS calls their model – unsurprisingly – One for One. Their name, TOMS, is taken from the word “tomorrow,” being part of the idea that if you buy a pair of TOMS shoes today, a pair is given away tomorrow.

What is the business model you currently operate in? Is your business model in-line with trends and values of your society, your target market and in the 21st century?