AI sees more VC Investments, but what are these startups promising to deliver?

AI startups received $1.8bn from investors in the first half of 2017, according to CB Insights. The second half of the year has seen a continuation of the trend, as Graphcore and SenseTime secured large capital financing rounds. This said, AI remains a trend with question marks hanging over it, with many of the industry’s leading researchers still feeling that we are a long way off solving critical issues with AI – like the frame problem, and inference.
Typically, when most consider AI, they think of the Googles and Baidus of this world, who spent between $20 to $30bn internally on the area in 2016, according to McKinsey. Much of this effort has been focused on a race to achieving patents and intellectual property, in the hope that they will enable these companies to own the future of software and hardware.
Machine-learning has already been implemented by Facebook, which uses its own internally developed algorithms to read user posts, or Netflix’s algorithms to make better recommendations to its subscribers – increasing the time spent in their ecosystems.
The sector is yet to really face down challenges from researchers involved with the well-financed AI efforts of ‘big tech.’ Google’s Ray Kurzweil, makes the point that today “machine-learning is today very brittle, requiring a lot of preparation by humans in the form of special-purpose coding, special-purpose sets of training data, and a custom learning structure. Today’s machine-learning really fails to imitate anything like the sponge-like learning structure that humans engage in.”
This form of intuitive learning still presents significant market opportunities, in the form of autonomous driving, voice recognition, video recognition, and data processing in business applications. On the back of this potential, investments have continued to flow.
SenseTime announced it will collaborate with Qualcomm on developing proprietary AI algorithms, to be deployed in smart devices. Although the two companies did not disclose the size of the investment, SenseTime is currently trying to raise $500m in a new funding round, in what would be the biggest ever fundraising by an AI startup – which values the company at $2bn.
SenseTime focuses on developing facial recognition technology, and is just one of a number of startup efforts that are competing to develop an AI to quickly identify and analyze identities using cameras. SenseTime’s algorithms have to date been used in limited tests by Chinese authorities to track and capture suspects in public spaces such as airports and festivals.
The startup already has 40 Chinese local authorities as clients, and is now seeking to expand overseas. SenseTime raised $410m in July, in a funding round led by its main backer, Chinese firm CDH Investments, and China’s state-backed fund Sailing Capital.
Elsewhere, Graphcore has raised $50m in new funding from Sequoia Capital. Graphcore had reportedly been chasing a $1bn valuation, but couldn’t quite convince investors of that potential. The funding comes on top of the $60m the group has already raised in its first two funding rounds – taking total investments in the group to $110m.
Graphcore was founded in 2016 in the UK, and focuses on building intelligence processing units (IPU) – chips that are specifically designed to assist programmers in creating machine-learning systems, that can be used in fields such as autonomous cars, data centers, and medical detection devices. The target applications for the IPU are those that already run machine-learning algorithms on standard graphics and processing units – hoping to entice them onto the dedicated silicon.
The company claims its IPU accelerators and Poplar software framework deliver “the fastest and most flexible platform for current and future machine intelligence applications, lowering the cost of AI in cloud data centers and improving performance by between 10x to 100x.”
Graphcore is underpinned by the belief that Efficient AI processing power is rapidly becoming the most sought-after resource in the world, and that the current CPU and GPU market can’t serve machine-learning type applications anywhere near as effectively as a dedicated processing unit.
However, it’s still unclear as to which use cases are really going to drive this AI dedicated processor market – given that most autonomous driving programs are currently partnered with Nvidia, to use its GPU, and that other applications are still quite nascent in their development.
Graphcore CEO, Nigel Toon, stresses that the companies IPU processor will fulfil two critical elements – high computing power, so that a high number of calculations can be achieved at relevant speeds, and interconnect functions. As more processors are required for the high computing capacity, the number of interconnections between processor chips becomes increasingly important, to network both processing cycles and data.
Toon, claims Graphcore is also focusing on the software to achieve this interconnect function in its IPU chips, through its Poplar software programing framework and application libraries. The libraries can be ported to Google’s TensorFlow machine learning software framework. In 2015, Google open-sourced the TensorFlow software library, aiming to set a de facto standard for ML systems.
One of the software features that Graphcore is working on is the ability to generate noise or random numbers, to improve its effective decision making. For instance, there may be multiple valid answers or responses to a given situation, like when driving, and the addition of noise in the processing unit apparently helps a better decision to be made.
Graphcore has stiff competition in the AI chip market, which although still an emerging area has seen a lot of activity. Google launched Tensor Processing Units (TPUs) and Intel revealed its Nervana Neural Network Processor (NNP) family.
Graphcore investor Sequoia Capital has been the early private investment partner behind Apple, Oracle, Nvidia, Yahoo, Google, YouTube, PayPal, Instagram, WhatsApp, and Airbnb. Capital Group has experience in the processor market, having worked with Nvidia. Graphcore has already received investments in previous rounds from Samsung, Bosch, DeepMind’s (now part of Google) co-founder Demis Hassabis and Amadeus Capital (a venture capital group ran by ARM co-founder Herman Hauser).
CEO Toon sold Icrea, his previous company and a maker of 3G and 4G baseband chips, to Nvidia for $367m in 2011. Before Graphcore, the electrical engineering graduate ran Picochip, a Bath-based semiconductor group, which is now owned by Intel, and XMOS, another Bristol-based chip company. Although this form in the chip market suggests that Toon is a man chasing trends, he has stressed that he believes Graphcore has the potential to go public.
Source: copyright 2017 Rethink Research, Inc.
Source: AI Trends


60 Cybersecurity Predictions For 2018

Like death and taxes, there are only two safe predictions about cybersecurity in 2018: There will be more spectacular data breaches and the EU General Data Protection Regulation (GDPR) will go into effect on May 25. But as the continuing digital transformation of our lives entails the ongoing digital transformation of crime, vandalism and warfare, 2018 could also bring a lot of new takes on old vulnerabilities, some completely new types of cyberattacks, and successful new defenses.
The following list of 60 predictions starts with three general observations and moves to a wide range of cybersecurity topics: Attacks on the US government and critical infrastructure, determining authenticity in the age of fake news, consumer privacy and the GDPR, the Internet of Things (IoT), Artificial Intelligence (AI) as a new tool in the hands of both attackers and defenders, cryptocurrencies and biometrics, the deployment of enterprise IT and cybersecurity, and the persistent cybersecurity skills shortage.
IoT vulnerabilities will get more critical and more dangerous. Despite this, there will be no real changes in US law to regulate these devices. This isn’t a very risky prediction; Congress is currently incapable of passing even uncontroversial laws, and any IoT regulation faces powerful industry lobbies that are fundamentally opposed to government involvement. More interesting is what’s happening in Europe. GDPR takes effect next year, and European regulators will begin to enforce it. The regulation has provisions on security as well as privacy, but it remains to be seen how they will be enforced. If Europe starts enforcing Internet security regulations with penalties that make a difference, we might start seeing IoT security improve. If not, the risks will continue to increase—Bruce Schneier, Schneier on Security
 Sophisticated adversaries will leverage the granular metadata stolen from breaches like Equifax, OPM, and Anthem, in precision targeted attacks that rely on demographic and psychographic Big Data algorithms powered by machine-learning and artificial intelligence. Attackers will deploy armies of bots to propagate the false narratives used to weaponize malicious fake news, inflate partisan debates, and undermine democratic institutions; meanwhile, they will launch multi-vector DDoS, ransomware, and malware campaigns to impede critical infrastructure cybersecurity and national security. The demographic and psychographic metadata will enable advanced spear-phishing operations against privileged critical infrastructure executives and pervasive Influence Operations against populations—James Scott, Senior Fellow, Institute for Critical Infrastructure Technology

We’re going to see more attacks that attempt to subvert two-factor authentication, as sophisticated attackers set their sights on two factor authentication-protected accounts and use flaws in SS7 to redirect SMS text messages. In addition, software supply chain attacks like the MEDocs compromise with NotPetya will be more prominent—Paul Roberts, The Security Ledger
Read the source article at
Source: AI Trends


AI Avoidance of Car Insurance Scams for Self-Driving Cars

By Dr. Lance B. Eliot, the AI Insider for AI Trends
I drive a somewhat exotic luxury car. Driving around, I am at times especially aware of the fact that it is a pricey car and either get accolades from other drivers and pedestrians, or at times receive ire from those that think it is wrong to drive such a car (ecologically because it is a gas guzzler, or because it seems boastful and a brag). Some places that I drive are equally filled with such exotic cars, and sometimes even more elaborate ones. In other cases, I drive in areas where the car stands out because the other cars in the area tend to be less expensive. In those areas, it instantly draws attention.
Some drivers of such exotic cars relish the attention, wherever and whenever they drive. For me, I am not that keen on the attention, especially in places that seem suspect. When I park the car on the streets in downtown Los Angeles, I never know whether when I come back if the car will be still there (might be stolen), or might be marred (graffiti or worse).  One of my previous cars had actually got stolen (I am a statistic now).
Driving this kind of a car has other potential consequences too.  
The other day, I was on the freeway and driving along without incident. Suddenly, a rather ragged car came up from behind me, switched into the lane to my left, zipped ahead, and then opted to unexpectedly jump into my lane directly in front of me.  There didn’t seem to be any obvious reason for this driving behavior. If the driver was trying to get ahead in traffic, the act of getting into my lane at that moment actually slowed down the progress of the other driver. Given the seemingly frantic movements, the driver should have stayed in the lane it had gotten into, or upon getting into my lane the driver should have accelerated further forward since there was empty space ahead.
As Spiderman might say, my spidey sense was tingling. Things weren’t adding up.
I next saw another ragged car coming up behind me, and it suddenly switched into the lane to my left. As it drove past me, I gave the driver a hard look. The driver seemed to be acting like they didn’t see me, but I am sure they must have. The driver was focused straight ahead. I noticed though that the driver ahead of me seemed to be studying his rearview mirror. For whatever reason, he suddenly seemed quite interested in what was happening behind him.
If I had not been paying attention, I would have just continued forward and not given much thought to what was occurring around me. Nothing explicit had yet happened. A car was ahead of me, and a car was to my left. They were both driving quickly, faster than the surrounding cars. They were both cars of a bit ragged in nature. They were presumably completely independent of each other. But, I just felt that maybe they were somehow connected to each other.
I opted to switch over to the rightmost lane. There was no real need to do so, but it seemed like it might be handy to change lanes and see what else would happen with the other two cars. The car that had been ahead of me tried to follow me over to the rightmost lane. He was blocked though by other cars. The cars in my lane then passed him. Meanwhile, the other car that had been to my left opted to slow down and keep pace with the other car. Why didn’t that driver zoom forward, which seemed like what he was earlier trying to accomplish?
The whole situation smelled. I knew that an upcoming freeway exit could be used to get off the freeway and just a block afterward would be an entrance that I could use to get back onto the freeway. At the last possible moment, I veered into the exit and got off the freeway. When I then drove ahead and got back onto the freeway, the other two cars were no longer to be seen (assuming they continued at the speed of the freeway, they would have been a distance ahead by then).
I might have just avoided a swoop and squat.
Are you familiar with a swoop and squat? If not, welcome to the vocabulary known to those that deal with car insurance fraud. The swoop and squat is the name given to a series of maneuvers by criminals trying to force a car accident.
Here’s how it works. Two vehicles (or more) work together to execute the swoop and squat. A driver in a lead car (the squatter) will get ahead of a target car. The target car is usually an expensive vehicle, which has been identified while driving along as a good candidate to be involved in an insurance fraud. The second criminal car moves ahead of the now lead criminal car and the target car. The front most criminal car then swoops into the front of the squatter. The squatter jams on their brakes. The target car driver then also presumably tries to quickly brake, but with the short distance between them and the squatter, they are going to rear-end the squatter car.
The swoop car then darts away and does not stop for the accident that it has now apparently caused. Meanwhile, the squatter car and the target car usually agree to stop and exchange insurance information. The squatter car might even have more than just the driver in it, perhaps several occupants. This allows the squatter car to potentially make multiple insurance claims, including that the occupants of the squatter car claim various injuries.
The beauty of this “accident” will be that the target car driver is usually held responsible for hitting the squatter car. You can of course try to profess that there was another car that cut-off the squatter car and that the squatter car messed-up. But, you would be held accountable for not allowing sufficient driving distance between you and the car ahead of you. It’s a mess. The scammers have staged the whole thing, and any savvy insurance adjuster is going to recognize it. Unfortunately, the odds are that the scammers will probably get away with the scam.
You might be thinking that this kind of scamming rarely occurs. You’d be wrong! I am either proud or disappointed to let you know that Los Angeles is considered the capital of auto insurance fraud. The California Department of Insurance (CDI) has about one hundred detectives devoted to auto insurance fraud. They are widely overworked and undermanned for the volume of auto insurance fraud occurring. Some of the auto insurance fraud is well organized and accomplished by gangs or other criminal enterprises.
The payoff can be high for those that commit auto insurance fraud. Insurance companies have deep pockets. They need to weigh the payout versus the effort to prove some kind of auto insurance fraud. Los Angeles is attractive to scammers because it has a goodly percentage of high-value vehicles, it has a tremendous amount of daily traffic, and lots of non-scam accidents that happen all the time. Thus, the scam car accidents are easier to pull off and can more readily hide among the many other non-scam car accidents that occur.  If you were a criminal and tried to pull the scam in some other locale, it might be more suspicious and standout to police and the insurance companies.
The scammers will try to make as much money of a scam as they can. They will often take their damaged squatter car to an auto-body shop that is also involved in the scam. The car body shop will make the damage appear to be more extensive than it really was. Either they will file false indications about the damage, or in some instances they will even do more damage to the car to make sure that it really does appear to have the extensive damage claimed. The occupants of the squatter car will potentially claim personal injuries due to the accident. They might have a physician that’s also involved in the fraud ring. The physician will substantiate the false injuries and then get a part of the loot for the scam.
If convicted, the scammers could face some serious prison time since this kind of fraud is considered a felony. They could also be financially penalized too. In one sense, this though is a type of fraud that is one of the least likely to be spotted. It is a low likelihood that it will be investigated. It is a low likelihood that it will be prosecuted. Sadly, the amount of money to be made by the scam, versus the chances of getting caught and getting penalized, means that auto insurance fraud continues to be a budding business.
Was I faced with a potential swoop and squat when I was on the freeway? I don’t know for sure. It certainly had the right ingredients. I was driving a high-value car. I was on a crowded freeway. The potential squat car had purposely maneuvered in front of me, when there didn’t seem to be any reasonable reason to do so. It was a ragged car. The second car, the potential swoop car, appeared to be working in conjunction with the other car. It was a ragged car. They were both positioning themselves into a classic swoop and squat situation. It might have been only in my mind, but I figured it was worth taking a mild evasive action to avoid the chances of getting mired in an auto insurance fraud case.
What does this have to do with self-driving cars?
At the Cybernetic Self-Driving Car Institute, we are developing AI for self-driving cars that detects these kinds of potential auto insurance fraud scam maneuvers and then seeks to avoid getting mired in them.
When I give presentations about our work at autonomous vehicle conferences, one of the first objections that I get is that there will not be a need for detecting auto insurance fraud cases, which purportedly is because once we have all self-driving cars on the roads there will no longer be any such scams. In other words, if we have all self-driving cars on the road, these self-driving cars would not act in such a nefarious manner. In this nirvana world, all self-driving cars are respectful of each other and we won’t have scam accidents.
Wake-up! We are going to have a mixture of human driven cars and self-driving cars for many, many, many years to come. This idea that by some magic act that suddenly all of the human driven cars disappear and are entirely replaced by self-driving cars is not realistic. It is a crazy dream. Therefore, self-driving cars must be prepared to interact with and deal with human driven cars. Period.
I would also like to add an aside. These same dreamers think that self-driving cars will always be respectful of the laws of driving and that they will always be respectful to other self-driving cars. Why will this be the case? It assumes again some kind of idealized world. We can pretty much anticipate that self-driving cars are going to be varying from this all-respect approach. We might even see scammers that hack a self-driving car to participate in scams. There are likely even going to be new kinds of scams involving self-driving cars that we aren’t even thinking about as yet (some self-driving cars will be targets, some will be perpetrators).
Another question that I sometimes get involves the aspect of whether self-driving cars will have car insurance. Normally, the driver of the car is the one that has the auto insurance. But, if the driver is AI, who then has the car insurance? Will the AI have the car insurance?
We pretty much can reject the notion that AI will be considered the equivalent of a human and be getting car insurance. The auto maker that made the self-driving car might be the one that has the auto insurance for the car, or someone else such as the tech firm that made the AI, or others. I think we can all agree that one way or another, self-driving cars are going to have car insurance.  I don’t think we’re going to have uninsured self-driving cars driving around on our public roadways (well, at least not legally doing so).
When self-driving cars first get mired in auto insurance scams, it will be a highly visible issue. The scammers will probably try to claim that the AI was mistaken and that it caused the accident. This though is something that today’s scammers are generally not sophisticated enough to try and pull off. Plus, it would make them overly visible. Nonetheless, I am sure that self-driving cars will be an attractive target. These first self-driving cars will be high-value cars and probably have high-value occupants.
In fact, you could suggest that self-driving cars are going to be ripe and easy targets. Most self-driving cars today are being developed without the kinds of defensive driving tactics that human drivers know and use. Self-driving cars tend to act like a novice driver. They are easy to fool. You might be aware of the famous case of the self-driving car that came to a four-way stop. The other human driven cars were able to roll through the stop signs and the self-driving car kept waiting its turn. In a similar manner, I am sure that scammers will be aware of the limitations of the self-driving cars in the marketplace and be able to exploit those limitations to undertake an auto insurance scam.
Another form of today’s auto insurance frauds involves bicycle riders that intentionally ram into a car. These bike riders are willing to get hit by a car, in order to file an insurance claim. Usually, though, these scams are dealt with immediately in that the bike rider asks the human driver for cash to make the case go away. Asking for say $200 cash is an easy scam and the driver will often want to avoid the insurance paperwork, so they give the cash to the scammer and continue along on their driving journey.
Anyway, let’s get back to the AI of self-driving cars and how it needs to be prepared to cope with potential auto insurance scams.
We are developing and testing AI that recognizes the swoop and squat. Similar to how I noticed the actions of other cars around me, there is a module in the AI of a self-driving car that is watching for signs of a potential scam. In the case of the swoop and squat, it sits aside of the rest of the AI driving the car, and tries to see if there is something suspicious about the other cars around the self-driving car. If it spots something potentially amiss, it notifies the strategic and tactical AI components that are driving the car. If the suspicion has a high enough probability, and if an avoidance effort can be done without undue risk, the self-driving car will take appropriate evasive action.
The self-driving car can also let the occupants know about what has taken place. The human occupants in the self-driving car might wonder why the self-driving car has suddenly exited from the freeway and then decided to enter back onto the freeway. There is an explanation system that can communicate to the occupants what has occurred. In some case, the occupants might not want to know and not care, while in other instances the occupants might be keenly interested to know.
Besides the swoop and squat, we also have the AI system be on the watch for other kinds of auto insurance scams.
There is the panic stop scam, consisting of just one squatter and no swoop car.
There is the start and stop, again usually done with one criminal driven car ahead of you.
There’s the wave-in, in which the human driver seems to offer you an opening in their lane and then rams into your car. This is harder for a conventional self-driving car to get caught up in, due to the aspect that the human driver of the criminal car usually makes a hand signal to the human driver of the target car. But, it still can be done with a self-driving car by making a tempting opening for the self-driving car and then ramming into it when it takes the opening.
Another scam is the sideswipe. This involves intersections that have two left turns. The criminal car will swerve into the lane of the target car.
It is hard to know in-advance that a scam is going to occur. The actions of the scammers can be similarly done innocently by regular drivers that are careless. Thus, there is no clear-cut way to know that a scam is being setup. That being said, whether a scam or not, the risk factor of getting involved in an accident is certainly detectable in all of these maneuvers. A good self-driving car should have a robust defensive driving AI capability to be watchful of these potential situations. These particular maneuvers such as the swoop and squat should even more so be on the defensive watch, since they are being done by both regular drivers and the scamming drivers.
It will be interesting to see how scammers find ways to especially make use of self-driving cars in their nefarious efforts. Besides the type of driving scams that I’ve mentioned, there are other scams such as staged auto thefts, there are dumped vehicle frauds (scammers dump a car into a lake and claim it was stolen), there are born again vehicles (a stolen vehicle is given a new Vehicle Identification Number or VIN, and used in a scam), and so on.
I know that many are hoping that self-driving cars are going to improve the world as we know it. There are indeed many ways in which self-driving cars are going to aid us. At the same time, without seeming to be pessimistic, I am sure that we will see criminal minds trying to find ways to involve self-driving cars into criminal acts. Let’s try to make the AI for self-driving cars good enough that self-driving cars won’t be ready unknowing dupes or accomplices in the rotten work of criminals.
This content is originally posted on AI Trends.
Source: AI Trends


MIT Growing business masters program with an AI focus; math and computer science are foundations

Michelle Li is director of the Masters of Business Analytics Program at MIT, from the MIT Sloan School of Management, with support from the MIT Operations Research Center, now in its second year. In this position, she is responsible for building the program, designing curriculum, hiring faculty, assessing admissions criteria for students and helping students to secure future careers. She took a few minutes recently to speak with AI Trends about education for AI.
Q. Does AI education live in computer science?
Michelle: AI lives at the intersection of math and computer science, so it can be either the math major or computer science major. MIT’s Operations Research Center combines math and science.
Q. What is the AI exposure for engineering students?
Michelle: Engineering is a very broad field. If a student is strong in linear algebra and calculus, that also qualifies them to do artificial intelligence. Some engineering degree programs that are a little light on math that may not qualify.
Q. If a college today is offering engineering without AI, is it out of touch?
Michelle: No. AI is a branding of something that’s been around for many years. As long as they have higher level math courses, they should be okay. I do recommend colleges incorporate some kind of AI coursework into their programs. We have launched for example at MIT, a new machine learning course in response to demand from students.
Q.If someone is interested in pursuing a career in AI, what education is needed?
Michelle: There are different levels of work in AI. For complex modeling and algorithmic work, using predictive analytics, I recommend courses such as Statistical Methods of Machine Learning. That is a requirement if you want to do deep dive analytics. But if you want to work in AI at the business level, as an analyst or operations manager, then that kind of deep dive is not necessarily required.
There are many routes in AI. Students can be doing visualization or data engineering, for example. AI involves extracting the data from a database or designing the database in the first place, making sense of the data, then running complex models and algorithms to use the data to forecast trends. Those can be related to self-driving cars, healthcare needs – it really just depends on where the students see themselves in the future.
Q, When might certificate programs such as Udacity be appropriate?
Michelle: Udacity has courses to teach Python, R, machine learning – so I do think there is a lot of value for students to take courses online. MIT has an online course in analytics called Analytics Edge. However, online learning is very lonely. Students I meet have a general need to ask questions and solve problems in group settings. It’s hard to problem-solve complex coding challenges by yourself. Our philosophy at MIT is we like our students to be on campus full time.
We are looking into online hybrid learning, such as the Analytics Edge course, for high-achieving [undergraduate] students, to invite them on campus for a pre-admission round [to the graduate school].
Q. How many AI workers can MIT produce per year?
Michelle: The Analytics Master’s Degree program will grow to 60 students next year. We have a two-year MBA program offering an analytics certificate, with 400 students in a class. The computer science AI lab has 400 research scientists with PhDs and advanced degrees, all studying AI. The undergraduate population in computer science and electrical engineering, Core 6, is the largest number of students at MIT, about 6,000.
Q. We have private companies setting up partnerships with institutions like MIT. NVIDIA this week announced a big Deep Learning Institute partner expansion. What is your impression of this approach to educations an AI workforce?
Michelle: I think it’s great companies are stepping up and trying to train workers to do work shifting to AI. We do a lot of partnerships with companies. IN our Capstone Project, students work with companies on real-life data science projects. We are working with 15 companies this year; when we go to 60 students, we will work with 30 companies.
Q. What is the most lucrative field in AI?
Michelle: Machine learning and the using deep learning techniques, are the most lucrative positions for students with undergraduate degrees. The average salaries for MIT graduates going into AI range between $60K and $80K. For graduates of the one-year masters program, salaries are averaging $90K to $110K. It’s quite a jump for a one-year program.
I have seen salaries in the six figures for undergraduate students going into AI fields. These salaries are growing quickly because of high demand. It’s also about geography. In San Francisco, the salaries are $20K to $30K higher, and often include equity and options.
Q. Is it easier for a student to get into your graduate program than your undergrad program?
Michelle: Our graduate program is more selective than the undergraduate program. Our average graduate student has a 3.9 grade point average out of 4.0. For our graduate school focused on machine learning or AI, students need computer science and math background. Our admission rate is less than 3%, so one in 30 or 40 is admitted.
Q. What’s a good starting point for students interested in starting out in a career in AI?
Michelle: For high school students, try to go into STEM courses, depending on interest. They should try to target what industry they are interested in; AI is in every industry for instance. You cannot really try to understand every single domain. It’s better to focus on the industry of your choice. It could be cars, software, computer science or math, which applies to many fields. But definitely get some computer science skills.
I am astonished that some undergrads go through a program without any computer science. It’s a basic class everyone should take. Scripting syntax, or what object-oriented coding is, understanding the logical parts within a computer is very valuable.
Q. How do the business partnerships MIT enters into benefit everyone interested in AI research?
Michelle. Our partnerships are purely educational, not to benefit one party specifically. Our students work on projects using the company’s data and companies are generous. They need to be willing to let go of private data they have collected and share it with MIT. That’s valuable to society. So professors are getting access to information they would not have had otherwise.
Also students speak at our conferences on what they are doing with companies, such as BMW and IBM. Our faculty director worked with the Boston Public Schools to optimize daily school bus routes, saving between $5 and $10 million. So we have valuable data sharing going on.
Q. Do you have anything to add or emphasize?
Michelle: Learn some computer science and some math. Even if you are going to be an English major or a musician, learn some computer science.
This content originally posted on AI Trends.
Source: AI Trends


Qualcomm Invests in Chinese AI Unicorn SenseTime for On-Device AI

SenseTime, China’s leading AI unicorn, on Nov. 15 announced a strategic investment agreement with global communications giant Qualcomm Inc., which is pending closing.
(Photo above shows SenseTime co-founder and CEO Li Xu (right), and Keith Kressin, SVP of product management at Qualcomm (left).)
The two companies will working on development of on-device AI for applications including computer vision and camera-based image processing.
In a statement, Qualcomm said the partnership will leverage SenseTime’s machine learning models and algorithms and Qualcomm’s Snapdragon chips, which offer heterogeneous computing capabilities for client-based AI.
SenseTime in July announced that it had raised $410 million in its Series B round of funding, setting the then-record for the highest single round of financing in the global AI industry.
Boosted by this collaboration with Qualcomm, SenseTime’s proprietary AI algorithms will be in a position to be deployed in more smartphones and devices. SenseTime’s “algorithm + chip” strategy behind the collaboration is set to redefine the next generation of intelligent terminal devices.
The explosive growth of intelligent terminals has driven the large-scale implementation of AI technology to new heights. While the performance quality of smart devices previously depended on the computing capabilities of chips, today’s smartphones, cameras, robots, the Internet of Things (IoT) products, and other devices require extremely powerful processing in order to meet the demands of real-time data. With the shared vision of SenseTime and Qualcomm Technologies to drive an “algorithm + chip” strategy, the two parties are on track to redefine the intelligent terminal ecosystem and accelerate its development.
Quinn Li, VP and Global Head of Qualcomm Ventures said in a press release, “Qualcomm is investing great effort into researching ways to boost the development of on-device AI with 5G technologies by enhancing AI chip capabilities. Qualcomm’s investment will enable SenseTime to invest more in AI research and development. Given our shared vision and customers, our collaboration will offer customers more integrated solutions, reduce the cost of deploying AI technologies for intelligent device OEMs, shorten the R&D cycle, and therefore rapidly upgrade the entire terminal device industry.”
Qualcomm said it is focused on optimising the Snapdragon mobile platform to accelerate AI use cases in the areas of computer vision and natural language processing for smartphones, IoT and automotive applications. It is also researching broader executions in the areas of wireless connectivity, power management, and photography.
Dr. Xu Li, co-founder and CEO of SenseTime, said, “SenseTime is dedicated to developing cutting-edge AI technologies to connect upstream partners with the downstream market. With Qualcomm’s strategic investment and collaboration, SenseTime can bring AI technologies to more terminal devices and boost the development of the entire intelligent device ecosystem. We look forward to exploring more intelligent implementations with Qualcomm Technologies to offer partners a one-stop solution that covers the entire industry chain.”
During the Shang Dynasty 3,600 years ago, China led the world with its advanced agricultural and handicraft industries; SenseTime, whose name comes from the phonetic translation of Shang Dynasty and its emperor Tang, derives great inspiration from this period and seeks to unleash a similar era in computer vision and deep learning in the AI world that will also enable SenseTime to export its self-developed technologies internationally. SenseTime, with Qualcomm Technologies’ commercial collaboration, believes it will not only enable the implementation of the “algorithm + chip” strategy but jumpstart an evolution from “smartphone” to “intelligent phone”, and from the “connected world” to the “intelligent world”.
About SenseTime
SenseTime is China’s largest artificial intelligence (AI) company focused on computer vision and deep learning technologies. It is also the largest AI unicorn valued above USD 1.5 billion.  
The developer behind China’s only proprietary deep learning platform, SenseTime has become the largest algorithm supplier in the country. The company is dedicated to creating an AI ecosystem with innovation and solutions to upgrade the industries.
Besides from its technological strengths, SenseTime has achieved commercial successes. It has powered many industries such as finance, security, smart phone, mobile Internet, and automobile with core computer vision technologies including face recognition, video analysis, character recognition, and autonomous driving.
SenseTime boasts more than 400 leading customers and strategic partners including Qualcomm, NVIDIA, China Mobile, UnionPay, HNA, Huawei, Xiaomi, OPPO, vivo, Weibo, and iFLYTEK.
With offices in Beijing, Shenzhen, Shanghai, Chengdu, Hangzhou, Hong Kong, Kyoto, and Tokyo, SenseTime has attracted top talents around the world to build a world-leading technology company – from China, for the world.
For more information, visit SenseTime’s website.
Source: AI Trends

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

How Can I Help You?