Much of the work undertaken by artificial intelligence involves a training process known as machine learning, where AI gets better at a task such as recognising a cat or mapping a route the more it does it. Now that same technique is being use to create new AI systems, without any human intervention.

For years, engineers at Google have been working on a freakishly smart machine learning system known as theAutoML system(or automatic machine learning system), which is already capable of creating AI that outperforms anything we've made.

Now, researchers have tweaked it to incorporate concepts of Darwinian evolution and shown it can build AI programs that continue to improve upon themselves faster than they would if humans were doing the coding.

The new system is called AutoML-Zero, and although it may sound a little alarming, it could lead to the rapid development of smarter systems - for example, neural networked designed to more accurately mimic the human brain with multiple layers and weightings, something human coders have struggled with.

"It is possible today to automatically discover complete machine learning algorithms just using basic mathematical operations as building blocks," write the researchers in their pre-print paper. "We demonstrate this by introducing a novel framework that significantly reduces human bias through a generic search space."

The original AutoML system is intended to make it easier for apps to leverage machine learning, and already includes plenty of automated features itself, but AutoML-Zero takes the required amount of human input way down.

Using a simple three-step process - setup, predict and learn - it can be thought of as machine learning from scratch.

The system starts off with a selection of 100 algorithms made by randomly combining simple mathematical operations. A sophisticated trial-and-error process then identifies the best performers, which are retained - with some tweaks - for another round of trials. In other words, the neural network is mutating as it goes.

When new code is produced, it's tested on AI tasks - like spotting the difference between a picture of a truck and a picture of a dog - and the best-performing algorithms are then kept for future iteration. Like survival of the fittest.

And it's fast too: the researchers reckon up to 10,000 possible algorithms can be searched through per second per processor (the more computer processors available for the task, the quicker it can work).

Eventually, this should see artificial intelligence systems become more widely used, and easier to access for programmers with no AI expertise. It might even help us eradicate human bias from AI, because humans are barely involved.

Work to improve AutoML-Zero continues, with the hope that it'll eventually be able to spit out algorithms that mere human programmers would never have thought of. Right now it's only capable of producing simple AI systems, but the researchers think the complexity can be scaled up rather rapidly.

"While most people were taking baby steps, [the researchers] took a giant leap into the unknown," computer scientist Risto Miikkulainen from the University of Texas, Austin, who was not involved in the work, told Edd Gent at Science. "This is one of those papers that could launch a lot of future research."

The research paper has yet to be published in a peer-reviewed journal, but can be viewed online at arXiv.org.

The rest is here:
Google Engineers 'Mutate' AI to Make It Evolve Systems Faster Than We Can Code Them - ScienceAlert

Artificial intelligence is a bit of a buzz term these days but what do people really mean when they say AI? And why should local governments care?

First of all, AI is extremely misunderstood. We arent talking about HAL from 2001: A Space Odyssey, necessarily; were talking about what Alan Turing speculated about thinking machines back in the 1950s. According to the Brookings Institute, AI is generally thought to refer to machines that respond to stimulation consistent with traditional responses from humans, given the human capacity for contemplation, judgment and intention. More simply put, AI uses algorithms to make decisions using real-time data. But unlike more traditional machines that can only respond in predetermined ways, AI can act on data it can analyze it and respond to it.

The concept has been evolving and the technology has become more sophisticated, but its still a little nebulous particularly for folks working in local government. It seems everyone kind of knows what AI is, but no one is exactly sure how they can apply it in their communities.

I spoke with Eyal Feder-Levy, the CEO and Founder of ZenCity, an AI-based tool that helps local government leaders listen to and synthesize conversations going on in their communities on social media, about the implications of AI for local governments, and how they can utilize these new tools in meaningful, beneficial ways. The following is a gently edited transcription of that discussion.

Derek Prall:So, when you say you use AI tools to analyze social media conversations crunch this data and make it meaningful can you tell me what that means? How does this work?

Eyal Feder-Levy:Lets use the current situation that local governments are facing as an example. First, I have to say I have nothing but admiration for local government leaders right now. Cities and counties are on the front lines of this global crisis that were facing they have to create the policies that will respond to this. They have to shape the information thats going out there. So in this current crisis, cities have a really important job to play. This means they have to constantly know whats working and what isnt working. They need to know if the messaging theyre putting out is resonating with people. They need to know if people are worried about child care or tax breaks for their businesses or are they worried about where to buy groceries. What are the things that they are prioritizing that we as local governments need to respond to in order for our communities to survive this crisis?

One of the only channels where we can still hear the population in this social distancing reality is online. People are talking more than ever on platforms like Facebook, Twitter and Instagram. Were talking about a massive amount of data. If we take a city like Dayton, Ohio, were seeing somewhere along the lines of tens of thousands of these online conversations in a week. Where AI comes into play is that no one in city hall has the time to go over 80,000 conversations a week and try to make sense out of them. We cant.So its amazing we have this information, its amazing we have this data, but we have to find a way to make sense out of it fast. This is where we as a city use AI. These are basically algorithms that break down the data in meaningful ways so it can be acted on.

Prall:Okay that definitely makes sense, but I want to take a bit of a step back. I think a lot of elected and appointed officials arent necessarily the most tech-savvy people. When you say artificial intelligence to someone who doesnt consider themselves to be good with technology how do you talk about what this is this as a concept. What is AI?

Feder-Levy:The first thing I want to say about AI is that its not robots coming to take our jobs its not something scary that only mathematicians can understand. Its actually part of our daily lives already. Its embedded in the technological and software tools we use every day. Its something that if we understand the basic concepts of it can be a very strong tool to help us automate a lot of things we dont have enough staff to do on a manageable level.

To read the complete article, visit American City & County.

Read the original post:
Artificial intelligence: The thinking machine - Urgent Communications

The corporate affairs ministry is betting on artificial intelligence and data analytics as key elements in the fight against the menace of shell companies as it works to put in place an ecosystem that will have "zero tolerance" for non-compliance with regulations.

Continuing efforts to have a robust corporate governance system and ensure high level of compliance, the ministry is also in the process of having an advanced MCA 21 portal.

The portal is used for submission of requisite filings under the companies law and is also a repository of data on corporates in the country.

Corporate Affairs Secretary Injeti Srinivas told PTI that once the third version of MCA 21 becomes fully operational, the portal would make it "almost impossible for a shell company to survive."

Generally, shell companies are those which are not complying with regulations and many such entities are allegedly used for money laundering and other illegal activities.

Noting that the third version of the portal might be fully operational in a year from now, the secretary said the ecosystem would have zero tolerance for non-compliance.

"Surveillance with respect to compliance will be on auto pilot mode with artificial intelligence (AI) and data analytics," he said.

MCA 21 system was started in 2006 and currently, the second version is operational.

There are nearly 12 lakh active companies in the country. Active companies are those that are in compliance with various regulatory requirements under the Companies Act.

Over the past two to three years, the ministry has been deregistering the names of companies from official records for prolonged non-compliance.

"From the trend I see, after 4.25 lakh shell companies having got struck off, the numbers getting added each year is reducing. This is a clear indication that the earlier scenario of shell companies openly indulging in accommodation entries has become a matter of past," Srinivas said.

Along with weeding out shell companies, the KYC (Know Your Client) drive for directors and companies has encouraged greater compliance.

"Now, more and more companies are becoming compliant. Compliance levels in terms of filings has crossed 80 per cent. The latest fresh start scheme for companies and settlement scheme for LLPs (Limted Liability Partnerships) are expected to further improve compliance levels it should soon cross 90 per cent," he noted.

At the end of February, there were around 19,89,777 registered companies in the country. Out of them, 7,44,014 companies were closed, 41,974 entities were in the process of being struck-off and 2,170 were assigned dormant status, as per data compiled by the ministry.

According to the ministry, there were 11,95,045 active companies as on February 29.

Read the original here:
Govt bets on artificial intelligence, data analytics to weed out shell cos - Business Standard

The technology sector is in an interesting place today. On the one hand, technology has historically tended to be cyclical, with technology demand fluctuating with GDP growth. On the other hand, technology is achieving more extraordinary feats by the day, and is helping to solve a lot of the problems caused by coronavirus. That may actually lead to a surge in demand for some tech products and services due to the stay-at-home economy.

These cross-currents came into focus in the first quarter earnings release of technology bell-weather Taiwan Semiconductor Manufacturing (NYSE:TSM). Taiwan Semi is the world's largest and most advanced outsourced foundry, making chips for tech giants including Apple (NASDAQ:AAPL), Qualcomm (NASDAQ:QCOM), and AMD (NASDAQ:AMD), among many others. Notably, Taiwan Semi leaped ahead of others in its ability to make leading-edge semiconductor chips on the 7nm node, with an eye toward 5nm production later this year.

Apparently, demand for leading-edge chips isn't seeing any slowdown from the COVID-19 outbreak.

Image source: Getty Images.

In the first quarter, Taiwan Semi saw explosive growth over the prior year.

Taiwan Semiconductor Manufacturing (NYSE:TSM)

Q1 2020

Revenue growth

42%

Gross margin

51.8% (+10.5 percentage points)

Net income growth

90.6%

Return on Equity

28.4%

Data source: Taiwan Semiconductor Q1 presentation. Table by author. YOY=year-over-year.

These are eye-popping growth numbers for sure, but don't expect the company to keep growing at this rate for the rest of 2020. The first quarter was lapping the first quarter of 2019, a recessionary quarter for tech due to the U.S.-China trade war. For the second quarter, Taiwan Semi's management basically predicts flat growth quarter over quarter.

Taiwan Semi management also anticipates a slowdown in the second half of this year amid the economic fallout from COVID-19. Management expects the overall semiconductor industry (ex-memory) to be flat to down for the year -- and 2019 wasn't exactly a great year for semiconductors.

However, for Taiwan Semiconductor specifically, the picture is much brighter. Management anticipates foundry growth in the high single digits or low teens this year, and that Taiwan Semi should outgrow even that, in the mid to high teens. That's pretty impressive as the rest of the world goes into recession.

Chalk up Taiwan Semiconductor's success to its lead in manufacturing chips on leading-edge nodes. Leading nodes are the smallest, densest, most advanced chips, with higher power and better battery efficiency. More powerful chips are needed in all the big megatrends today, from 5G communications to artificial intelligence applications in the data center.

For instance, Taiwan Semiconductor gets almost half of its revenue from smartphone chips. You might think this would cause Taiwan Semi's revenue to fall, since it expects smartphone units to decline in the "high single-digits" this year. However, because more and more 5G phones need leading-edge chips, TSM's content growth per smartphone will be over 20%, according to management, meaning overall smartphone revenue for Taiwan Semi should grow in the mid to high teens, even as units decline.

Meanwhile, high-performance computing, Taiwan Semi's other big sector, not only needs leading-edge chips, but is actually seeing a demand surge due to increased cloud use amid work-from-home streaming applications.

Last quarter, smartphones were 49% of TSM sales and high-performance computing was 30%. Leading-edge 7nm nodes made up 35% of revenue, the largest node segment for the company.

Basically, since Taiwan Semi has a lead on other foundries at the leading edge, it won't be nearly as affected as the rest of the semiconductor industry.

When asked about the company's 3.2%dividend on the conference call with analysts, management reiterated that the company will pay its current quarterly divided, with the intention of raising it in the future, and the dividend would not go below the current payout going forward. That's certainly refreshing in an environment when many companies are cutting their dividends instead.

When looking for dividend stocks in the midst of the coronavirus, it's probably best to stick with companies that:

Today, Taiwan Semiconductors fits all three criteria. That's why it's one of the safest dividendsout there, not only in tech, but also the entire market.

View original post here:
For This High-Yield Stock, 5G and Artificial Intelligence Outweigh Coronavirus - The Motley Fool

The banking industry has always seemed to be one of the most developed and willing to invest in new technologies.

It's no wonder that artificial intelligence has quickly become one of the technical pillars on which the entire modern financial market is built.

Not everyone is aware that AI is not only leading analytical solution, but also a way to change the way customers interact with services provided by the financial industry. Let's take a closer look at this extraordinary relationship, its impact on the way we use banks, and on issues such as fraud detection and compliance regulations.

Artificial Intelligence is used in many FinTech solutions. Its a cure for the daily challenges faced by many businesses like customer experience personalization and loyalty building, to strictly technical financial features such as anomaly detection or fraud prevention.

The beginnings of AI in the industry, however, were not so simple. The first attempts to improve the operation of banks using computers were made in the 1950s. The story started with the simplest and most obvious solutions: accountants wanted to use computers to make calculations much faster and more accurately than real people could.

However, it turned out that their use might not be so easy since the machines themselves were not as powerful as they are now. Despite this fact, Bayesian statistics, which is used in machine learning even today, was implemented to expand algorithms enabling processing actions such as stock market predictions, loan repayments or calculation of probabilities regarding auditing.

In the early 90s, AI and machine learning appeared on Wall Street along with the first hedge funds - but there was still no significant breakthrough. It appeared only with the increased availability of data, generally with the spread of the internet. Since then, there has been an extremely rapid evolution of operating systems, taking advantage of the increasing capabilities of machines.

Nowadays AI basically affects every area of a bank's operations as well as the work of departments that we often forget about in the context of using technology in the financial sector, such as corporate core aspects, including even human resource team work.

All the aspects in which AI is involved are perfectly summarized in the heat map below:

Source

Considering the multitude of AI applications in the industry, you could basically write a book about each of them. Let's focus on the most common solutions that are already (or are about to be) popularly used in FinTech and which, as consumers, we should be well aware of.

According to Accenture, there are some key trends in the industry which should be followed:

To simplify matters even more, the above-mentioned aspects can be divided into four groups:sales optimization, growth revenue driving, operational process improving, and credit risk management.

But its not only about the possibilities in the field of topics to cover. The scale of use of artificial intelligence across whole companies operating in the industry is enormous. The AI in Financial Services global study shows that 85% of all respondents currently use some form of AI. As a reason for implementing such solutions, respondents indicated the need to boost both speed and efficiency, and the demand for broader data-driven insights.

What's more, the declarative statistics contained in the report are even more optimistic: 77% of respondents said that AI will become the most important or one of the most important investment areas for their businesses by the end of 2020.

But what exactly do FinTech representatives want to invest in? Contrary to appearances, most of them (64%) primarily plan to invest in reaching mass users, thanks to the implementation of AI in aspects such as revenue generation, process automation, risk management, customer service, and client acquisition. This is an extremely significant increase. At present, only 16% of respondents declare the intention to invest in these areas.

Let's stop for a moment at the division placed at the end of the last paragraph and take a closer look at each of the areas mentioned:

Benefits resulting from the use of AI in FinTech, Synerise

They benefit with advantages valued by managers responsible for financial institutions strategies i.e. cost reduction, sales, and revenue improvement or business risk mitigation. However, its also worth mentioning that the AI options used by banks do not end there. The possibilities can go far behind typical expected features.

Theres a saying in the financial market thats very timely: people don't really need banks, they need banking.

In the age of smartphones and simplified login methods, theres a special way of entering a bank's mobile application: taking a selfie. This solution is relatively easy to deploy and implement. The identification procedure is speedy and does not require too many actions from the end user, which in itself is an encouraging way to drive adoption of the process.

The OCBC Bank from Singapore enabled clients to use such an AI-driven option; the only requirement needed to be logged in this way is to have an iPhone X.

Citing the bank's official statement, its users, thanks to facial recognition solution, can now: dispense with passwords or even fingerprints when doing their everyday banking on mobile apps.

But how does this technology actually work? As the name suggests, the success of the login process depends on the identifying or verifying the identity of a given bank user. AI captures, analyzes, and compares specific patterns that each of us has on our face.

Let's look at the instructions prepared by another bank using this technology, Polish bank PEKAO:

Instructions on how to take a selfie are needed to set up an account: gesture blink eyes, gesture turn head left, gesture turn head right

As the bank's CEO, Micha Krupiski, said, the company is very pleased with the results of the introduction of face recognition technology, and about 25% of the account entries were made outside the bank's business hours.

He emphasized that:

"We believe in our banking, our advantage is the strength of the mobile application, we've been growing by 50-60 percent year on year here. We will invest even more in the mobile app.

The results turned out to be so satisfying that selfie-verification functionality will probably also be adapted to the needs of micro-companies.

Maybe you associate virtual assistants with robots that look like people and will one day take over the world.

The accuracy about how reality looks, however, is completely different. It is true that chat assistants have come a long way from their humble beginnings, but in fact we are still developing machine learning and natural language processing, so assistants are just learning our human-like manner, and are really far from mastering the world.

One example is the twins created by Hang Seng Bank (China), Haro and Dori. They have extraordinary language skills; they communicate in Chinese, English, Cantonese, and a mixture of Chinese and English.

However, chat assistants shared their tasks. Haro focuses on general queries, such as products, services (with special emphasis on mortgage, personal loan, credit card and insurance services).

Dori, in turn, is a typical type of Facebook Messenger, using the opportunities offered by personalized recommendations based on customer preferences.

Of course, this is just one of many interesting examples. Another is Erica, a Bank of America employee and AI-driven chatbot who deals with card security updates and credit card debt reduction. In 2019, this virtual assistant processed over 50 million clients requests, regardless of their needs and age: 15% from Gen Z, 49% millennials, 20% Gen X and 16% percent from seniors, who are typically not the target group for such solutions.

Voice search is becoming more and more popular. In a recent report on this subject, Microsoft emphasizes that 69% of respondents by 2020 will regularly use voice assistants. Of course, such trends have not escaped the notice of banks, like Lloyds Bank, the Bank of Scotland, or Halifax UK.

These financial institutions have decided to simplify the lives of their customers by using "voice biometrics," i.e. confirmation of identity through AI-driven advanced analysis of the user's voice characteristics.

Of course it's hard to disagree that using an account through voice commands is easier and faster than traditional logging methods but is it completely safe? Some industry analysts point out that if there are recordings on the web that contain our voice (e.g. in the form of podcasts) - they can be used to log into our account by unauthorized persons.

Such solutions will likely blossom in the near future, although they already exist in limited forms in the industry. Current versions, especially in European and American markets, are somewhat more mundane and familiar solutions, focusing on mobile banking, fraud detection, and regulatory compliance.

Mobile is our future: it is predicted that by 2023 this device will be used over 7.33 billion people worldwide. By the same year, the mobile app market will generate revenue of 935.2 billion dollars, which of course also includes mobile banking applications. What makes us so willing to invest in them?

First off, mobile banking means improved security, which is often at a higher level than typical online counterparts. What's more, applications are eagerly used by banks for more prosaic reasons since they allow banks to cut operational costs. Thanks to mobile, expenditures on typical offline banking operations and human resources can be reduced and they are also cheaper than ATMs. What's more, they actually save not only money, but also time and paper commonly used to supplement "necessary" paperwork.

Also, mobile apps are always available. Its easy to analyze the data collected through this channel. Plus, mobile facilitates communication with the client thanks to the option of sending push notifications.

But what does AI have to do with this? At first glance, biometrics in mobile banking available thanks to the artificial intelligence solutions may look a bit like a part of a science fiction movie, especially aspects such as fingerprint scanning, facial recognition, iris scanning, and voice biometrics.

Lets stop for a minute and wonder if our smartphones are at all able to support such advanced technology. According to Juniper Research data, the availability of dedicated hardware will not be an obstacle to be used for these biometric purposes. The company predicts that by 2024, about 90% of phones will cope with these modern solutions.

The real question we should ask ourselves in this context is a bit different. Will people in the era of contactless cards really want to use this kind of mobile feature to be used to authenticate contactless payments? The forecasts mentioned above are not very promising - only 30% of respondents declared that they would gladly use this option.

The first seeds of AI fraud detection were implemented over 10 years ago, based on anomaly detection, a technique for identifying deviations from a norm, covering issues related to cybersecurity and anti-money laundering processes.

Nowadays, common fraud types in the financial sector include identity theft and extortion of loans using stolen documents or login details. As this McAfee report indicates, (which also includes data concerning financial fraud), cybercrime costs FinTech globally around $600 billion, equal to 0.8% of global GDP.

These events not only cause real financial losses, but also add to the problem of debt collection which in many legal systems is an extremely long-term process, but unfortunately not in all cases one hundred percent effective. Financial institutions are also harmed and as a consequence, they can lose their reputation in the market. This kind of damage can be fatal in financial markets.

Fortunately, AI, and solutions that automatically prevent financial fraud, also known as fraud detection/prevention systems (FDS), can help.

Detection and prevention systems differ primarily in the way they are implemented. Prevention is slightly more complicated and requires the bank to be authorized to intervene in the banking platform and transaction system; meanwhile, detection only requires access to data, without the need for direct intervention into the platform.

Regardless of which FDS system you choose, it should be able to detect and monitor all actions taken by the user, regardless of the channel he uses to complete the transaction. This means not only investments in caring for the web channel, but also protection of ATMs, some call center services, "offline" operations at the bank's branch or mobile payment orders.

According to the chart below, the size of the fraud prevention and detection market is constantly increasing. By 2022, it will be worth $41.50 billion, compared to $14.37 billion in 2016 a massive increase.

Regulations play a key role in the banking sector and this is another field where AI can help, facilitating (and accelerating) complex analyzes in the modern data-centric world. Lets take a look how it simplifies the whole process and makes it much more effective.

Let's start with the fact that AI can automate repetitive manual tasks. Regulatory compliance processes are based on collecting data from various source systems. Before these data can be forwarded for further decisions, they must be organized and carefully checked.

Without AI, all the work is labor-intensive and requires several manual interventions. Moreover, the whole procedure is time-consuming and prone to mistakes. Such a solution can be to some extent also be called robotic process automation (RPA). It can be done via automation, with webhooks, or APO integrations.

Thanks to the ability to process quickly and accurately, AI is definitely a better decision-maker. Algorithms will analyze all risks, including those related to financial crimes, money laundering and potential fraud (AML, MiFID II, FinCEN).

AI implemented in banking undoubtedly affects the optimization of sales and the operations of B2B and B2C sales. This is due to, among other things, improved customer service.

Artificial intelligence allows you to accurately reach the selected target group and personalize the message. Segmentation significantly shortens the entire purchasing process, and well-used knowledge of customer preferences affects the number of users of financial products.

AI is also able to carry out a detailed analysis of client decisions, and offer only those products that a given person really needs. It is worth emphasizing that recommendation models created for banks are much more complicated than those used in typical e-commerce.

Perhaps the extraordinary power of personalization provided by AI in the context of the above mentioned biometric examples seems quite "ordinary" - so let me show you specific numbers.

The Boston Consulting Group has estimated that only by personalizing customer interactions, a bank can garner up to $300 million in revenue growth for every $100 billion it has in assets.

Why is this happening? Consumers expect that, by definition, complex banking systems will be as accessible and easy to use as other services they use on a daily basis:

We are entering a completely new digitized era in which the possibilities of AI in FinTech are still developing. We don't know with what kind of new features artificial intelligence will surprise us in time but one thing is certain: brands need to take advantage of the power that it offers.

As you can see, banks are in possession of so many types data:

Some of them come in the real time, yet some are really scarce.

Lets take an advantage of AI and machine learning advancements in order to combine the information coming from these multiple possibilities into the lowest possible dimension.

Many of the functionalities are related to the most desired applications enabling revenue generation, such as recommending actions and offers based on true 360 degree customer profile or enhancing currently used statistical models by adding features allowing brands to evaluate and compare neighborhood of any entity banks are working with (debtors, creditors, merchants, individuals, enterprises).

AI will allow us not only to save time and money, but also to better protect savings and more easily access our money. What more could we want?

Learn more about all things AI by checking out G2's artificial intelligence hub.

Go here to see the original:
How Artificial Intelligence Is Influencing the Banking Sector - G2