Google Cloud Platform cuts the price of GPUs by up to 36 percent

Google has announced lower prices for the use of Nvidia’s Tesla GPUs through its Compute Engine by up to 36 percent. In U.S. regions, using the somewhat older K80 GPUs will now cost $0.45 per hour while using the newer and more powerful P100 machines will cost $1.46 per hour (all with per-second billing).
The company is also dropping the prices for preemptible local SSDs by almost 40 percent. “Preemptible local SSDs” refers to local SSDs attached to Google’s preemptible VMs. You can’t attach GPUs to preemptible instances, though, so this is a nice little bonus announcement — but it isn’t going to directly benefit GPU users.
As for the new GPU pricing, it’s clear that Google is aiming this feature at developers who want to run their own machine learning workloads on its cloud, though there also are a number of other applications — including physical simulations and molecular modeling — that greatly benefit from the hundreds of cores that are now available on these GPUs. The P100, which is officially still in beta on the Google Cloud Platform, features 3594 cores, for example.
Developers can attach up to four P100 and eight K80 dies to each instance. Like regular VMs, GPU users will also receive sustained-use discounts, though most users probably don’t keep their GPUs running for a full month.
It’s hard not to see this announcement in the light of AWS’s upcoming annual developer conference, which will take over most of Las Vegas’s hotel conference space next week. AWS is expected to make a number of AI and machine learning announcements, and chances are we’ll see some price cuts from AWS, too.
Read the source article at TechCrunch.
Source: AI Trends


This artificial intelligence is designed to be mentally unstable

DABUS is a new type of AI that paves the way for smarter machines. Nothing new there – but this machine-learning system is different. DABUS, short for “device for the autonomous bootstrapping of unified Sentience”, is deliberately created to be mentally unstable.
Since the 90s, computer scientist Stephen Thaler has been injecting noise into a special neural network to generate novel ideas. Thaler, CEO of Missouri-based Imagination Engines, calls this module an Imagitron. The stream of creativity is assessed by a second network called a Perceptron, which provides feedback to improve ideas. This AI approach of neural networks playing off each other has been adopted by Facebook, Google and others – under the name “generative adversarial networks” – as a method of creating images that look authentic to human observers.

Now Thaler has introduced a technique to assess ideas according to how they resonate with existing knowledge – the AI equivalent of art or music that triggers happy or unhappy associations. Another process boosts the slow and tentative rhythm characteristic of creative activity in a neural network. The system swings between extremes of unimaginative plodding and novel thinking. It can also exceed the bounds of sanity.

“At one end, we see all the characteristic symptoms of mental illness, hallucinations, attention deficit and mania,” Thaler says. “At the other, we have reduced cognitive flow and depression.” This process is illustrated by DABUS’s artistic output, which combines and mutates images in a progressively more surreal stream of consciousness.
Thaler is also using the technology for stock-market prediction and to help autonomous robots find creative ways of tackling obstacles. He believes that with larger networks the approach will offer human-like problem-solving and genius-level ideas, impacting on scientific discovery, economics and more. But developers will need to restrict creativity to sensible limits. “The AI systems of the future will have their bouts of mental illness,” Thaler says. “Especially if they aspire to create more than what they know.”
Read the source article at Wired.

Source: AI Trends


Rocket Man Human Drivers and AI Self-Driving Cars: Outrunning Them

By Dr. Lance Eliot,the AI Trends Insider
I was on the freeway the other day and traffic was moving along rather smoothly. Even though there were quite a number of cars on the road, we were all doing a steady 55 miles per hour. That’s an accomplishment in the crowded freeways of Southern California. Some days I get maybe an average speed of 10 miles per hour during commute times, and often it drops to around 5 miles per hour on the average. Whenever the freeway moves along at a fast clip, I look around wondering if maybe the end of the earth is nearly upon us, or some other miracle has occurred that no one bothered to tell me about.
Well, there I am, enjoying my speedy 55 miles per hour, when in my rear view mirror I spot a Rocket Man. When I say Rocket Man, I am generically referring to any human driver that decides they are going to rocket through traffic. It could be a man or a woman, and I just use “Rocket Man” because it is a handy and catchy term for the behavior. Please think of it as Rocket Person.
I could see the car about a quarter mile behind me. It was moving forward at a fast pace, probably doing at least 20 to 30 miles an hour faster speed than the prevailing traffic. I would guess that the car was going around 85 to 90 miles per hour. This might be Okay if the driver was on a straightaway that had no traffic, but instead this driver was doing that kind of speed while weaving into and around other cars. The offending car would zip ahead in the fast lane, come upon a “slow” car that was doing the 55 miles per hour of the rest of us, and then dive into the lane to the right when there was an opening.
The driver would then zip forward in that lane, and was looking to make the next jump to another lane, since the driver had reached the bumper of a car in the existing lane and was now blocked from going at the 90 miles per hour clip. Into and out of the other lanes of traffic and narrowly missing hitting other cars was the pattern of driving behavior. The wild driver would almost always come right up to the bumper of a car in whatever lane the wild driver was in and then in a semi-panic mode desperately try to switch lanes. It was a very dangerous effort. The speeding driver would muscle into another lane and cause the cars in that lane to slow down to let in the wild driver.
I am sure that in the mind of the wild driver that he or she perceived the rest of us as sheep. We were just abiding by traffic flow, and this other driver figured why be so sheepish. Instead, this aggressive driver figured that they would try to tie together any possible openings and jump from one to the other. I am further guessing that the wild driver didn’t think this was particularly dangerous. They probably thought it was perfectly fine as a driving strategy. I’ve spoken to such drivers and they claim too that they are actually helping traffic. They think that they are optimizing the available driving space by using it in this fashion.  No sense in leaving any gaps or openings, they figure, and instead maximize traffic flow by having cars in all available roadway space.
Of course, that’s a crock.
This Rocket Man driver is putting all other drivers at a heightened risk of injury or death. Their wild antics can easily cause an accident to occur. This can happen by their direct actions such as they ram into a car and cause the accident to happen. It can also occur by their indirect actions, such that if the rest of traffic is trying to adjust to this nutty behavior, you can have innocent cars that get caught into a domino effect that leads to car crashes. In other words, if car W, the wild driver, cuts off a driver X, and then driver X hits their brakes, but driver Y behind driver X wasn’t expecting it, and then driver X and Y hit each other, the driver W can pretend they had nothing to do with it.
Some of these knuckle head drivers will even insist that if other cars get into an accident because of their rude behavior, it merely shows that those other drivers are bad drivers and should not be on the freeways. The wild driver believes that other drivers should be watching out for them, and it is not the wild driver that needs to be on alert. All I’m doing is weaving into and out of traffic, the Rocket Man says, and if other drivers are so stupid that they can’t handle it, they deserve to get into a wreck.
I often wish that I had some instant means to inform the police about such drivers. You look around and hopelessly wish that a highway patrol car will be on the freeway and spot such a driver. I almost never see these drivers getting caught. I’d like to pretend that they do. I’d like to pretend that their behavior is so frequent that they ultimately are getting caught. But, unfortunately, I would realistically guess that it is somewhat rare that they get caught. Getting caught is not especially likely since these wild drivers are usually watching for the cops, and they try to turn innocent when they see the police. Once the police are no longer around, they continue their wild driving.
Furthermore, it would be difficult for a highway patrol car to realize what is going on. If the wild driver was on a straightaway and doing 90 miles per hour, a highway patrol car could easily see that the driver was speeding and observe as such over a clear-cut distance. Instead, by weaving in and out of traffic, the wild driver is actually somewhat obscured and hidden from view. Only if you were observing from above, such as being in a police helicopter or plane, could you readily see the pattern of the driver and realize they are driving recklessly and at high speeds.
Some of these wild drivers do other illegal acts too. They will often dart into and out of the car pool lanes, even though they are not legally allowed to do so. They will often make use of the emergency lane as though it is a conventional lane, committing another illegal act. They will often intimidate other cars and pretty much chase them, and take other actions that are totally abusive of the privileges of being able to drive a car. You might say that their weaving into and out of traffic is really just a microcosm on their overall bad behavior as a driver. They are likely drivers that have disdain for the civilized rules of driving, and we merely happen to witness more clearly their disdain when they act in these aggressive manners.
That being said, there are certainly some instances of the normal everyday driver that opts to drive in a Rocket Man like way. If you are late for work and worried about losing your job, you’re likely to adopt that same bad practice that day or that moment. Or, if you are maybe in a joyful playful mood, you might think it is fun to momentarily try to weave in and out of traffic. It becomes almost like a personal game of Frogger. Though, you are at risk of bodily harm due to driving a motor vehicle at high speeds, and others around you are also at heightened risk, you nonetheless treat driving like a game at times. When behind the wheel of a car, we all sometimes forget that we are in a killing machine and can become distanced from reality by acting as though we are in some gigantic video game.
What does this have to do with AI self-driving cars?
At the Cybernetic Self-Driving Car Institute, we are developing AI that takes into account the Rocket Man drivers and can undertake various evasive actions accordingly. Furthermore, the self-driving car can become a Rocket Man, if needed, by using the same techniques that we generally don’t want cars to do (more on this in a moment).
First, let’s discuss how to detect that a car is driving in a Rocket Man way.
The sensory devices of the self-driving car should be looking behind the self-driving car and up ahead in front of the driving car. These might be a combination of LIDAR, cameras, ultra-sonic, etc. By analyzing the sensory data, what we are looking for is car behavior outside the norm. If the norm of the traffic at the time of analysis is that most of the cars are doing speed A, and if a car is doing a much faster speed B, we want to flag internally that as a car to be further closely observed. It stands out among the rest of the traffic.
This car B could be making a one-time sudden lane change and thus it is not truly a Rocket Man. Thus, we need to watch over time to see whether a car is consistently acting as a Rocket Man. It usually doesn’t take much time to ascertain the behavior. The biggest problem is usually being able to track the Rocket Man candidate. There are likely other cars and trucks on the freeway that obscure the wild driver. This is especially the case due to the wild driver weaving into and out of lanes. Imagine it is like playing a game of hide-and-seek. One moment, you can see the wild car, the next moment is seems to disappear.
Our main criteria is the speed differential in comparison to the other traffic, combined with the rapid lane changes, combined with the narrow gaps between cars. The Rocket Man has a tendency to run right up to the bumper of other cars. In their little minds, they think that this is the optimum way to make forward progress. If they were to study simulations, which we’ve done extensively, they would find out that their bumper nearing antics is actually not the fastest way to skirt through traffic. Had they a more open mind, they might realize that a more optimal path is possible, but most of the time they are just doing a monkey-see monkey-do kind of driving practice.
In fact, in some of our testing on the roads, we’ve been able to move ahead at a faster pace than the average Rocket Man, by adopting the same principles but performing the actions in a more studied manner. The act of changing lanes and weaving can be done with grace and aplomb, while the crasser approach is not only more dangerous but not even necessarily as successful. In essence, sometimes weaving across all four lanes of traffic, though it might seem like a faster way to proceed, can be beat by for example weaving only within two lanes of traffic. It involves lining up the weaving opportunities and timing them just right.
You might be wondering why I am describing this driving behavior as wild and yet at the same time touting there are better ways to do it. Am I being inconsistent?  Nope. The reason why it is useful to do this kind of driving in a better way is related to the future of driving. Eventually, some believe that the roadways will have only self-driving cars (there won’t be any human driven cars). Though it is questionable that this will happen, let’s go with the assumption for the moment.
If all cars on the freeway are self-driving cars, they can potentially coordinate their movements. It will be like a herd or a swam of animals that work in unison with each other. As such, the overall pace for all of the self-driving cars can be heightened by working together. This might also involve allowing some cars to do the Rocket Man like behavior. Suppose that the freeway is filled with self-driving cars, all moving along at some normal speed, and then one of the self-driving cars has an emergency, such as a human occupant that has had a heart attack. We might want that self-driving car to then perform Rocket Man maneuvers to allow it to proceed ahead at a faster pace than the rest of the traffic.
I realize that some of you that are cynics will say that you’d want your self-driving car to always be the Rocket Man, and thus move faster than the rest of the traffic. But, of course, if all the self-driving cars did this, we’d not really end-up moving any faster. All in all, we will likely ultimately have new driving regulations that will indicate when self-driving cars are to behave like the rest of traffic, and when they can do individualized acts such as a Rocket Man (such as the heart attack example of an occupant).
It is for these above reasons that we not only are developing AI to cope with the human driven Rocket Man behavior, but also want the AI of the self-driving car to be able to drive like a Rocket Man. The Rocket Man skill is worthwhile for the AI to have available. This does not mean that it is skill that should be used, and presumably would only be used when appropriate.
Let’s get back to the detection of human driven Rocket Man behavior.
Once the sensory data has been examined and we’ve detected a potential Rocket Man, the virtual model of the driving world is then updated to flag that car. We can then begin to predict what that car will do next. Based on the prediction, the self-driving car AI can take a defensive posture.
For example, if the Rocket Man is coming up from behind the self-driving car, the AI can opt to switch lanes if that will help avoid having the Rocket Man get onto the bumper of the self-driving car. The AI might opt to slow down, or speed-up, depending upon which approach is best for the circumstance. Generally, the AI is trying to avoid the Rocket Man from ramming into the self-driving car.
I’ve seen some human drivers that have either unintentionally or intentionally cut-off a Rocket Man driver. This can anger the Rocket Man driver and make them do even worse things. The Rocket Man will sometimes purposely get in front of a car that cut them off, and then play a dangerous game of braking to make the other car get scared. This of course actually slows down the progress of the Rocket Man, but they often seem to be of such a mired mind that seeking revenge is apparently more important than speeding ahead.
If necessary, the AI of the self-driving car might even opt to take the self-driving car off the freeway entirely, for the moment, and allow the Rocket Man to proceed on their way. It all depends on how desperately the Rocket Man seems to be driving and the predicted danger to the self-driving car and its occupants.
Speaking of the occupants of the self-driving car, we also have the AI inform the occupants about the Rocket Man, if appropriate to do so. This is based on whether the occupants have indicated to the AI that they want to be kept informed about the traffic conditions. On the one hand, the AI does not want to needlessly panic the occupants, while at the same time if the AI is going to be taking evasive maneuvers then the occupants might be wondering and concerned as to why the self-driving car is taking such actions.
Once we have self-driving cars that are communicating with each other via V2V (vehicle to vehicle) communications, the AI could inform other cars to be watchful of the Rocket Man. This would aid other self-driving cars that have not yet detected the Rocket Man, or that have software that is not as advanced that is able to detect Rocket Man behavior. Likewise, if self-driving cars are able to communicate externally, they could potentially alert the police – which then takes us to my earlier point about wanting to let the cops know when a driver of this ilk is on the roadway.
I have been describing the Rocket Man as a lone wolf driver. There is nothing that precludes there being multiple Rocket Man drivers at the same time. Indeed, I see this during my daily commute. There are often several Rocket Man drivers all vying to get ahead in the traffic. The AI therefore needs to be able to handle the one-at-time circumstance and the gaggle of Rocket Man drivers too.
This content is originally posted on AI Trends.
Source: AI Trends


AI & Academic Luminaries Appointed to Glasswing’s Connect Council

Glasswing Ventures, an early-stage venture capital firm investing in the next generation of AI-powered startups, this week unveiled its Connect Council. The Connect Council consists of two working groups: the AI & Academic Group, and the Business Leadership Group. It is the first of three advisory councils to support and extend Glasswing Ventures’ investment strategy. Collectively, these councils bring together 40 renowned entrepreneurs and technologists, AI visionaries, and world-leading executives to exclusively advise and support the firm and its portfolio companies.
Members of Glasswing’s Connect Council AI & Academic Group include:
⦁ Sir Tim Berners-Lee, inventor of the World Wide Web, Professor at MIT and Oxford University and winner of the ACM A.M. Turing Prize;
⦁ Dr. Brad Berens, Chief Strategy Officer at the Center for the Digital Future at USC Annenberg;
⦁ Dr. Cynthia Breazeal, Associate Professor of Media Arts and Sciences at MIT, Founder and Chief Scientist of Jibo, Inc.;
⦁ Dr. Thomas R. Eisenmann, Howard H. Stevenson Professor of Business Administration at the Harvard Business School, Faculty Co-Chair of the HBS Rock Center for Entrepreneurship;
⦁ Dr. Alex ‘Sandy’ Pentland, MIT Professor and Media Lab Entrepreneurship Program Director;
⦁ Dr. Manuela Veloso, Herbert A. Simon University Professor and Head of Machine Learning Department at Carnegie Mellon University;
⦁ Dr. Peter Weinstock, Executive Director and Anesthesia Endowed Chair of the Boston Children’s Hospital Simulator Program and Associate Professor of Anesthesia at Harvard Medical School
Shaping the Next Generation of AI: The Connected World
A critical part of the Glasswing Ventures’ DNA, the Connect Council extends the firm’s strength in providing AI expertise and advice exponentially amplifying Glasswing’s and its portfolio companies’ competitive edge.
“The Connect Council brings tremendous scale to Glasswing, as we help harness the positive potential of AI across industries and markets,” said Rudina Seseri, Managing Partner of Glasswing Ventures. “The Connect Council is a collaborative and vibrant body composed of the most influential thought leaders and innovators in academia and AI technology today. Our team, our founders and portfolio companies, gain access to a brilliant collective of luminaries at the forefront of AI and innovation, who are committed to fueling its success and growth.”
Sir Tim Berners-Lee, MIT Professor and inventor of the World Wide Web noted: “We are standing on the threshold of another major tech disruption and the Connect Council’s work to guide the brightest AI founders and startups as they work to shape the future of AI will be vital.”
Glasswing Ventures’ Connect Council presents a unique opportunity for all members to collaborate, share and pool resources, helping accelerate discovery and development of connected AI consumer and enterprise technology companies.
Glasswing Ventures is an early-stage venture capital firm dedicated to investing in the next generation of AI-powered technology companies that connect consumers and enterprises and secure the ecosystem.
The company was launched 18 months ago. “Our core strategy was to have advisory councils that would help us scale as a firm. But we wanted people who would work with us beyond just lending their names and faces on our website,” said Seseri. The advisors have been working with Glasswing and are just now being announced. Seseri said they each agree to work with Glasswing exclusively. “That is uncommon and difficult to attain; that’s a big achievement,” Seseri said.
Much work in AI is being done in labs and research groups, with commercialization coming rapidly from research in many cases. Having the councils “gives us a look forward to all that is coming out, and the advisors who work with us get to see cutting edge work put into practice.”
For more information, visit Glasswing Ventures.
Source: AI Trends


Meet SAM, the artificially intelligent politician

A Wellington, New Zealand-based technology group has unveiled a pilot digital platform: ‘SAM – the Virtual Politician’ at the launch of the  Wellington CIT.AI chapter, part of a global network of cities showcasing their artificial intelligence (AI) capability and talent.
The group’s spokesperson Nick Gerritsen, said the group aimed to test whether the ways in which people engage with politics and debate the big issues could be improved.
“We’re asking whether an AI politician could provide the facts rather than push a party line,” he said. “We believe it’s time to consider whether technology, and in this case AI, can help us get better information to inform decision-making on the major issues like water quality, housing, or climate change. We need better outcomes.”
Gerritsen said the current political system was reliant on politicians staying on top of all the major issues, being well-informed and coming up with good decisions. “We’ve seen in the US, UK, and Spain recently, however, that politicians may be wildly out of touch with what people actually think and want. Perhaps it’s time to see whether technology can produce better results for the people than politicians.”
According to Gerritsen, the technology proposed would be better than traditional polling because it would be akin to having a continuous conversation, and could give the ‘silent majority’ a voice.

“Natural language processing technology and sentiment analysis algorithms have come a long way in recent times and are now at the level where they have practical application in day-to-day interactions with people,” Gerritsen said.
Initially, the project will conduct research into whether people would engage with a virtual politician and then move on to the viability of building the software. If it proves to be viable the aim would be to have the virtual politician up and running for the next general election.
Gerritsen said the group wanted to have a positive impact on political discussion and democracy and did not have a political agenda or bias as such.
“We might be surprised at the outcome. And that might be a good thing if it’s more in tune with the voting public. The technology will enable greater people power,” he said.
Read the source article at Computerworld New Zealand.
Source: AI Trends


The future of getting dressed: AI, VR and smart fabrics

Cher Horowitz’s closet from the film “Clueless” had a futuristic computer system that helped her put together outfits. Back in 1995, the concept teased what it might be like to get dressed in the future.
Technology has evolved a lot since then, but closets have been largely untouched by innovation.
Now, that’s starting to change.
“If algorithms do their job well, people will spend less time thinking about what to wear,” said Ranjitha Kumar, an assistant professor in the Department of Computer Science at the University of Illinois at Urbana-Champaign.
From artificial intelligence and gadgets to smart fabrics and virtual reality, technology is poised to breathe innovation into not only how we dress but how we shop.
The most recognizable example is Amazon’s Echo Look, which received significant buzz when it was announced earlier this year. The gadget ($200) serves as a style assistant to help you decide what to wear.
Like Amazon’s other smart speakers, the Echo Look will tell you the weather or play music. But the oval-shaped product also has a voice-controlled camera for taking photos of you in various outfits. It works alongside an app.
After snapping photos of you in two outfits in front of the device, its built-in Style Check tool decides which one is best. It leans on a combination of machine learning technology and human opinion.
Amazon’s “fashion specialists” train the software to be a judge of style. The automated results consider “fit, color, styling, seasons and current trends.” It’ll also suggest similar styles to buy from various brands. Through testing, we found that the suggestions can be hit or miss.
“The brand selection is pretty limited, and while the Echo Look may help you decide between two looks, it can’t take into account the total context of where you’re going,” said personal stylist and creative director Taylor Okata.
Okata, whose clients include E! and SELF Magazine, doesn’t consider the technology a threat to his work: “There’s just that interpersonal communication that it just doesn’t have.”
Meanwhile, retail experts say the Echo Look’s success will depend on if it adds more value than just asking a friend for fashion advice.
Sucharita Mulpuru, an analyst at research firm Forrester, said those buying the device are early adopters — and it lacks widespread appeal.
“It’s such a foreign concept to rely on a device to tell you what to wear,” she told CNN Tech.
Read the source article at
Source: AI Trends


AI Trends Weekly Brief: AI in Fintech

AI Pushing Financial Sector to Innovate,
While Fintech Startups Offer New Banking Models
Fintech, or Financial Technologies, represents many technologies that have disrupted traditional financial services. Fintech is advancing mobile payments, money transfers, loans, fundraising and asset management. Innovations in Fintech are being driven by the advancements made in AI.
Fintech investment got off to a quiet start in 2017. While Fintech deal volumes and deal value held relatively stable quarter-over-quarter at $3.2 billion invested globally, results remained below the levels seen in 2015 and early 2016, according to an analysis by KPMG..
In other points, KPMG finds:

While the US led Fintech investment in Q1’17, with $1.5 billion across VC, PE and M&A, one of the strongest elements of the global Fintech market is in the wide variety of Fintech hubs that have developed around the world. This quarter’s top ten global deals accurately portray the diversity of the global Fintech market, with deals in the US, Canada, India, China, Sweden and the UK making the list. Even within the US, Fintechs have succeeded in growing outside of Silicon Valley, with companies based in Delaware and Ohio making the top deals list.
With the exception of a few jurisdictions, there has been exponential growth in Fintech over the past few years. With Fintech companies and technologies now maturing in leading jurisdictions, investors appear to be looking for early investments to prove and show one’s ability to achieve scale.
Q1’17 saw a number of mature Fintech companies and Fintech investors focusing on expansion as a means to fuel growth, either geographically or through product or service expansion. Unicorn company, SoFi, is a great example of this. During Q1’17, SoFi acquired Zenbanx gaining the ability to provide more functions of a traditional bank, including customer deposits. Other companies mentioned included Ant Financial and Kakao Pay.
Despite the results of Q4’16, a limited number of banks have left the R3 consortium in Q1’17, including Goldman Sachs and Santander. The consortium model continues to evolve as a way to develop blockchain. In Q1’17, a number of new consortia focused on a more limited set of use case areas were announced, including the State Bank of India’s National Bank Blockchain Consortium.

New words to Learn – from Investopedia
Insurtech – Refers to the use of technology innovations designed to squeeze out savings and efficiency from the current insurance industry model. Insurtech is a portmanteau of “insurance” and “technology” that was inspired by the term Fintech. The belief driving insurtech companies is that the insurance industry is ripe for innovation and disruption.
Regtech – A blend word of ‘regulatory technology’ that was created to address regulatory challenges in the financial services sector through innovative technology. Regtech consists of a group of companies that use technology to help businesses comply with regulations efficiently and inexpensively.
In more highlights from the KPMG Fintech analysis:

Q1’17 saw VC-specific insurtech investment drop to $243 million across 43 deals globally as the sector experienced a pause following strong growth in 2016. This lull is not expected to last, particularly as insurance companies around the world seem to have begun to feel the pressure to embrace insurtech innovation.
Investors are showing continued interest in regtech, with strong early investment in Q1’17 following 2016’s peak. In addition to reducing compliance costs through automation, regtech solutions are increasingly supporting a broader remit, delivering capabilities to support the growth agenda.
On a technology level, AI is expected to be a key area of focus for many investors, in addition to smart data and predictive analytics. With PSD2 approaching in Europe, open banking and API offerings are also expected to gain significant attention.

Where AI is Being Applied in Fintech
AI is having a major impact on where Fintech is headed. An account in Due cites nine categories where AI is being applied in Fintech:

Credit Scoring/Direct Lending

Companies in this category use AI for credit scoring and lending applications.
2. Assistants / Personal Finance
Companies in this category rely on AI chatbots and mobile app assistant applications in order to monitor personal finances.
3. Quantitative & Asset Management
Companies in this category employ AI algorithmic trading and investment strategies or tools.
4. Insurance
Companies in this category use AI to quote and insure.
5. Market Research / Sentiment Analysis
Companies within this category use AI for research and to measure sentiment.
6. Debt Collection
Companies in this category use AI to improve creditor collection of outstanding debt through personalized and automated communication.
7. Business Finance & Expense Reporting
Companies in this category use AI to improve basic business accounting.
8. General Purpose / Predictive Analytics
Companies in this category use AI for general purpose semantic and natural language applications as well as for broadly applied predictive analytics.
9. Regulatory, Compliance, & Fraud Detection
Companies in this category use AI to detect fraudulent and abnormal financial behavior, and/or to improve general regulatory compliance matters and workflows. While this technology may sound revolutionary, it’s actually been around for years in the trading industry.
“There are two types of algorithms used in trading today,” says Juergen Schmidhuber, an AI researcher of more than 20 years, quoted in the Due account. The New York Times called Mr. Schmidhuber the father of AI in November 2016.
“There are simple programs pre-wired to do certain things that the traders have identified, for instance. Little tricks that enable it to propose a price for a certain share. Then, of course there are little [systems] that are just going through a bunch of rules. Depending on the risk profile of the client, they make certain decisions according to not very intelligent self-learning mechanisms.”
“On the other hand, you have systems that have been used since the 1990s that learn from experience to become better prediction machines,” adds Schmidhuber.
“These use neural networks to predict behavior, financial indicators and so on. The hope is you have a system that works better than those of your competitors and detects patterns that the others don’t see.”
Fintech is expanding around the world, with major investments in China, Japan and Europe. “Smarter computers, algorithms and dedicated AI systems are a Fintech dream. Faster decision-making and deeper learning (recognising, for example, predictors of financial turbulence) are obvious and huge boons to financial organisations,” writes Rich Wordsworth for Wired UK.
“The drive to eliminate human fallibility has also made artificial intelligence (AI) driven to the forefront of research and development. It’s also expected to have a major impact in Fintech due to potential of game changing insights that can be derived from the sheer volume of data that humanity is generating,” adds Nikolai Kuznetsov for The Next Web, quoted in Due.
Speaking at a WIRED Money 2017 event, the CEO of online investment management company Nutmeg, Nick Hungerford, made some observations on where Fintech is headed: “It’s a combination of big data and artificial intelligence. We’re going to be able to be more intelligent about people’s spending habits, their health, their lifestyles. [We’re] going to get more effective at predicting what they’re going to need for different scenarios of spending and saving.”
“AI predicts when people are likely to get married, when you are likely to have a baby, etc. So in five years, we should be really good at giving people financial advice before they even realise they need it,” says Hungerford.
Traditional financial institutions are moving to experiment with AI as well.
“AI driven workflows will be the only way for traditional banks to leapfrog the competition from new, nimble breed of banks built around innovative technology such as Blockchain and business models such as peer-to-peer payments,” says Ramesh Mahalingam, CEO and Founder at Vizru, quoted in Due.
“The emerging new shareconomy demands banks to reassess their role where products and services need to be increasingly personalized. Using real-time data can only be delivered by AI-driven digital ecosystems. AI systems dynamically and continuously learn, reason and solve problems in real-time,” Mahalingam said. “AI driven workflows will play an instrumental role for banks in the future to deliver immersive customer experience.”
Simply put, innovations in Fintech are being driven by the advancements made in AI.
Banks Historically Resistant to Change
Banks historically have been slow to innovate, either because they are too big to quickly adapt or because they don’t know how to truly change. That has changed in the last five years, out of necessity and not a voluntary push by the banking sector, suggests Francesco Corea, Ph.D., complexity scientist, data science consultant and AI advisor, in a recent article in Hackathon.
Financial innovation is often characterized by a product innovation. Technology advances today have tightened the relationship between innovation and growth, making the drive to innovate in the financial sector today a drive to survive. Still the financial industry is not experimenting and pushing to create new models that spread innovation risk, Corea suggests.
Better innovation models are coming forward in Fintech. AI is creating a strong pressure to innovate for the financial sector, and AI is risky. Corea noted these AI development cycle characteristics: it requires a long time to be created, implemented and correctly deployed; it is highly technical and requires highly specialized talents; it is highly uncertain, because you need to experiment a lot before finding something that works; it is expensive, both in terms of time as well as monetary investments (talents, hardware, and data are really expensive); it is risky and the risk lies in the initial development phase, with a very high-payout but a high likelihood to fail as well.
AI is introducing a completely new speed and degree of trust in the financial industry. However AI introduces biased data and a lack of transparency, as demonstrated in some consumer applications.
AI in financial services is important not for the specific innovation or product it is introducing, “but rather because it is revolutionizing a centuries-old industry innovation flow from the ground,” Corea suggests. AI is using structured and unstructured data in financial services to improve the customer experience and engagement, to detect outliers and anomalies, to increase revenues, reduce costs, find predictability in patterns and increase forecast reliability.
Financial services is an industry full of data. Instead of the data being concentrated in big financial institutions’ hands, most of it is actually public and thanks to the new EU payment directive PSD2 (Revised Payment Service Directive, to take effect in 2018), larger datasets are available to smaller players as well. AI can then be easily developed and applied because the barriers to entry are lower with respect to other sectors, Corea suggests.
Moreover, many of the processes underlying financial services can be automated rather easily, while others can be improved by either brute force computation or speed. And historically, financial services is a sector that has needed this type of innovation the most, is incredibly competitive and is always looking for some new source of ROI. Bottom line: the marginal impact of AI is greater than in other sectors.
In addition, the transfer of wealth across different generations makes the field really fertile for AI, Corea suggests. AI needs a lot of innovative data and above all feedback to improve, and millennials are happy to use AI, provide feedback and apparently be less concerned about privacy and giving away their data.
AI does face some challenges in the financial sector that limit a smooth and rapid implementation: legacy systems that do not talk to each other; data silos; poor data quality control; lack of expertise; lack of management vision; and a lack of cultural mindset to adopt this technology.
Dr. Corea credits CB Insights with offering good framework and classifications of AI Fintech startups, and he offered his personal framework for the AI Fintech landscape:

Financial Wellness: this category is about making the end-client life better and easier and it includes personalized financial services; credit scoring; automated financial advisors and planners that assist the users in making financial decisions (robo-advisor, virtual assistants, and chatbots;  smart wallets that coach users differently based on their habits and needs. Examples include [robo-advisors and conversational interfaces] Kasisto; Trim; Penny; Cleo; Acorns; Fingenius; Wealthfront; SigFig; Betterment; LearnVest; Jemstep; [credit scoring] Aire; TypeScore; CreditVidya; ZestFinance; Applied Data Finance; Wecash;
Blockchain: I think that, given the importance of this instrument, it deserves a separate category regardless of the specific application is being used for (which may be payments, compliance, trading, etc.). Examples include: Euklid; Paxos; Ripple; Digital Asset;
Financial Security: this can be divided into identification (payment security and physical identification — biometrics and KYC) and detection(looking for fraudulent and abnormal financial behavior —AML and fraud detection). Examples include, respectively: EyeVerify; Bionym; FaceFirst; Onfido; and Feedzai; Kount, APEX Analytics;
Money Transfer: this category includes payments, peer-to-peer lending, and debt collection. Examples include: TrueAccord; LendUp; Kabbage; LendingClub;
Capital Markets: this is a big section, and I tend to divide it into five main subsections:

i) Trading (either algotrading or trading/exchange platforms). Examples include: Euclidean; Quantestein; Renaissance Technologies, Walnut Algorithms; EmmaAI; Aidyia; Binatix; Kimerick Technologies; ;Sentient Technologies; Tickermachine; Walnut Algorithm ; Clone Algo; Algoriz; Alpaca; Portfolio123; Sigopt;
ii) Do-It-Yourself Funds (either crowdsource funds or home-trading). Examples include: Sentifi; Numerai; Quantopian; Quantiacs; QuantConnect; Inovance;

iii) Markets Intelligence (information extraction or insights generation). Examples include: Indico Data Solutions; Acuity Trading; Lucena Research; Dataminr; Alphasense; Kensho Technologies; Aylien; I Know First; Alpha Modus; ArtQuant;

iv) Alternative Data (most of the alternative data applications are in capital markets rather than broader financial sector so it makes sense to put it here). Examples include: Cape Analytics; Metabiota; Eagle Alpha;
v) Risk Management (this section is more a residual subcategory because most of the time startups in this group fall within other groups as well). Examples include: Ablemarkets; Financial Network Analysis.

Dr. Corea concludes, “AI is making the industry more digitalized than ever before. Its final goal will be to create the frictionless bank of the future: no branches, no credit cards, no frauds, no menial reporting activities. A bank-as-a-platform with modular components that increases our financial literacy and has no physical products or spaces. It would definitely be a great world to live in. Can’t wait for it.”

Compiled and written by John P. Desmond, AI Trends Editor

Source: AI Trends

Welcome to Fusion Informatics
(AI, Mobility, Blockchain, IoT Solution Providers)

How Can I Help You?