Toyota steers millions into U. Michigan AI project

Toyota has pledged $22 million to a four year artificial intelligence, robotics, and autonomous driving project led by the University of Michigan.

The investment will be spread throughout the four years and go towards all sorts of tech projects, including stair climbing wheelchairs, cameras for blind people, and autonomous cars.

“Toyota has long enjoyed an excellent working relationship with the University of Michigan, and we are excited to expand our collective efforts to address complex mobility challenges through artificial intelligence,” said Toyota Research Institute CEO Gill Pratt.

“We look forward to collaborating with U-M’s research faculty and students to develop new intelligent technologies that will help drivers travel more safely, securely and efficiently. We will also focus on expanding the benefit of mobility technology to in-home support of older persons and those with special needs,” she added.

Earlier this year, Toyota announced a $1 billion research facility for autonomous vehicles in Ann Arbor, Michigan. That is also in collaboration with the University of Michigan.

This is Toyota’s third self-driving campus

The facility is Toyota’s third in the U.S., after Palo Alto, CA and Cambridge, MA. All three are connected to the local university, in Palo Alto that’s Stanford, in Cambridge its MIT.

Toyota clearly sees America’s universities as a way to advance its self-driving technology without being too active in the industry. Others, like General Motors, have followed a similar model.

The Japanese automaker still seems hesitant to make a large play in the autonomous market, despite the investments in research centers. In 2014, it said it had no interest in autonomous cars, and since then has pushed testing to its luxury car brand Lexus and research to universities.

That might be a smart strategy to avoid public scrutiny and lower the costs. The tech centers will retain their value in a few years, while General Motors $1 billion acquisition of Cruise Automation or Tesla’s large expenditures on AutoPilot will not have any value if they don’t win.

LinkedIn sues anonymous data scrapers

LinkedIn is trying to lock down its exclusive relationship with its users.

The professional networking company filed suit against 100 unnamed individuals last week for using bots to harvest user profiles from its website. The lawsuit is a preliminary step to revealing the identities of the scrapers — LinkedIn intends to ask the court to reveal the true identities behind the scrapers’ IP addresses — and a way to maintain its exclusive hold on users’ resumes.

But LinkedIn’s lawsuit also raises questions about how to police bot use. The company, which was recently snapped up by Microsoft for $26.2 billion, has invoked the controversial Computer Fraud and Abuse Act (CFAA) in its suit against the unidentified scrapers, claiming that collecting user profiles from the site amounts to hacking.

“During periods of time since December 2015, and to this day, unknown persons and/or entities employing various automated software programs (often referred to as ‘bots’) have extracted and copied data from many LinkedIn pages,” the lawsuit claims. “To access this information on LinkedIn’s site, the Doe Defendants circumvented several technical barriers employed by LinkedIn that prevent mass automated scraping, and have knowingly and intentionally violated various access and use restrictions in LinkedIn’s User Agreement, which they agreed to abide by in registering LinkedIn member accounts. In so doing, they have violated an array of federal and state laws, including the Computer Fraud and Abuse Act.”

The CFAA allows companies like LinkedIn to bring cases against anyone who gains “unauthorized access” to a “protected computer.” The law has been criticized for essentially criminalizing Terms of Service violations, and Representative Zoe Lofgren and other lawmakers have pushed unsuccessfully for CFAA reform.

LinkedIn’s case accuses the anonymous scrapers of building a massive botnet and circumventing the restrictions LinkedIn uses to prevent profile collection by undesirable third parties.

The lawsuit details several of LinkedIn’s automated tools that prevent data harvesting. Dubbed FUSE, Quicksand and Sentinel, these tools monitor the web traffic of LinkedIn users and limit how many other profiles a user can view, and how quickly a user can view those profiles. This tracking is intended to prevent scrapers from signing up for fake LinkedIn profiles and then vacuuming up vast amounts of data. The company also uses a tool called Org Block to block IP addresses it suspects of scraping and uses Member and Guest Request Scoring to track page requests.

But paradoxically, LinkedIn doesn’t want to prohibit scraping altogether. Search engines like Google use bots to index websites and turn up relevant results — and LinkedIn wants to allow this type of scraping to occur.

“LinkedIn ‘whitelists’ a number of popular and reputable service providers, search engines, and other platforms so as to permit them to query and index the LinkedIn website, without being subject to all of LinkedIn’s security measures,” the company explains in its suit. The scrapers targeted in the lawsuit circumvented LinkedIn’s bot-blocking tools by sending their requests through one of these ‘whitelisted’ entities, a third-party cloud service provider.

A LinkedIn representative declined to comment on how the company differentiates between good and bad scraping, referring TechCrunch to the complaint, which does not discuss how the company makes that determination.

It’s also not clear what kind of behavior LinkedIn is trying to prevent, since the lawsuit doesn’t specify what the scraped data is being used for. Does LinkedIn want to squash a competitor? Or is it targeting a research project like ICWATCH, which archives the resumes of individuals in the intelligence community?

LinkedIn likely defines ‘bad’ scraping based on the scrapers’ effort to circumvent the company’s preventative measures. While it gives special access to search engines and other friendly bots, LinkedIn obviously didn’t give permission to the data harvesters it’s suing.

Similar CFAA lawsuits, like Craigslist’s against 3Taps and Facebook’s against Power Ventures, have been favorable to the plaintiffs, so LinkedIn has a good shot at shutting down its scrapers. Twitch filed a comparable CFAA lawsuit against view-bots earlier this summer, in which the live stream site alleged that using bots to inflate a channel’s view count amounts to an unauthorized access of Twitch’s ‘protected computers.’ However, Twitch’s complaint also claims a number of other violations, including trademark infringement.

Clearly, companies are interested in stamping out certain kinds of bots. But other scraping, like that done by search engines and web archiving services like the Wayback Machine, is welcomed. That dichotomy could create an anti-competitive business atmosphere, the Electronic Frontier Foundation argues.

“If you make it illegal for bots to access websites, you’ve given existing search engines a monopoly,” EFF staff attorney Nate Cardozo told TechCrunch. “Google and Bing got started by crawling the entire web. That’s essentially what LinkedIn is talking about here. To call scraping a CFAA violation is extremely anti-competitive. Using the CFAA to stifle innovation is certainly not what it was intended for.”

But LinkedIn says that fighting some bots and allowing others is essential to protecting its members. The case is scheduled to be heard in U.S. District Court in San Jose.

XPRIZE launches AI 2020 competition with IBM Watson

Close your eyes. What do you see when you think about artificial intelligence? Hal 9000defying human orders? Ava savagely murdering her maker? What about ARIIA taking over the country from deep inside the Pentagon?

With a new competition, XPRIZE and IBM Watson challenge us to think slightly more pragmatically (or at least less apocalyptically dystopian) about the implications of artificial intelligence research on the future of humanity.

To date, more than 1,000 people have registered to form teams with plans to tackle such issues as health, climate, transportation, space travel, robots, city planning, surgery, education and even civil rights. Because the competition is an open challenge, teams are expected to come up with criteria on which they will ultimately be judged in 2020.

Some teams may be backed by corporations or angels, while others may go it alone as a team. Labs, startups, a group in a corporation or even a spinoff can constitute a team.

“We think this prize specifically might drive a lot of financing activity,” said Amir Banifatemi, XPRIZE lead for the competition.

The interdisciplinary nature of the competition is what makes it so exciting. Teams may be comprised of some folks adept at traditional skills, like machine learning, and others with knowledge of cognitive science, mathematics, linguistics and even design.

The AI competition officially starts in March 2017 and offers a number of check-ins and milestone awards. The final presentation will be at TED 2020. Because the challenge will occur over multiple years, final projects will include technologies and utilize skill sets that don’t exist today.

XPRIZE was founded 20 years ago to catalyze radical breakthroughs that could benefit humanity. The nonprofit gained recognition for coordinating the Ansari XPRIZE for suborbital spaceflight, which resulted in the flight of SpaceShipOne in 2004.

The winner of the AI prize will receive $3 million courtesy of IBM, with the runner-up getting $1 million and the third-place recipient getting $500,000.

If you think you have what it takes to change the world with AI, register here.

Autopilot in Tesla Model X helps driver get safely to a hospital

A Missouri man might not have made it safely to the hospital without the aid of his Tesla Model X’s self-driving features (via Slate). Joshua Neally, the 37-year old Model X owner, was on the way home from his law firm after work when he was struck by a pulmonary embolism. Neally allowed the Autopilot system to take over for 20 miles of highway driving, which brought the vehicle and its driver to an off-ramp close to a hospital, where he was able to finish the trip.

Tesla’s Autopilot system may be under scrutiny after an accident in Florida earlier this year resulted in a driver death, but this is an example of where it definitely helped the operator of the vehicle. The obvious question that comes to mind is whether it would’ve been safer on the whole for Neally to pull over and call an ambulance instead of relying on his own car’s self-driving features, since Neally admits he didn’t “remember much of the drive” after his embolism began.

But whether or not Neally made the right call in continuing to drive himself, it proves the importance and necessity of offering safety features like Autopilot: In clutch situations like a medical emergency where we feel like our life is on the line, very few people can be relied upon to make the sane, logical and most conscientious choice about our actions.

Neally told Branson local NBC affiliate KY3 that he “just knew [he] had to get [to the ER]” once he began feeling the symptoms of the pulmonary embolism – in other words, he was essentially panicking and his instinct kicked in, convincing his brain the best course of action was to get to the hospital as quickly as possible.

Continuing to drive while under extreme physical duress might not have been the best choice, but it was the human one, and that’s what makes Autopilot’s role in getting Neally to the hospital in one piece so poignant. Plus, the fact that Neally had a choice at all is a chance outcome of the severity of the embolism: If it had been something more severe, and he wasn’t able physically to continue driving or make a decision, Autopilot would’ve engaged its failsafe mode, which decelerates the vehicle, turns on the four-way hazard lights and slowly moves the car to the roadside.

You can argue about whether or not a driver in this situation is more or less likely to continue driving rather than pull over and call an ambulance since the system is present to begin with, but you can also debate whether allowing Autopilot to drive was more or less safe than other options. What you can’t argue is that people will always make the most rational and informed decision when their life feels immediately threatened, and that’s where autonomous technologies stand to make the most impact.

The human role in a bot-dominated future

Imagine a world where bots are ubiquitous… a world where nearly every online interaction takes place with a Siri, Alexa, Cortana or some soon-to-be-named artificial being. Here, banking is a breeze, as a customer service bot can quickly extrapolate your banking preferences from your online search history. In this world, your cupboards and refrigerator are always full, because your groceries are reordered every week automatically, based on consumption data.

But in such a world, where bots provide the ultimate convenience of a futuristic lifestyle, is there still room for human help?

At the most recent F8 Conference, Facebook CEO Mark Zuckerberg made some bold claimsabout a bot’s place in the future of commerce. Using 1-800-Flowers as an example, Zuckerberg argued that by integrating bot chats into the sales process, a customer would never have to dial 1-800-Flowers or speak to a human again. In theory, using chat with bot support can expedite the buyer-seller interaction and bring the consumer closer to a sale. While Zuckerberg may be correct to say that customers much prefer chat interaction to using the phone, we can’t necessarily leap to the conclusion that their preference includes chatting with a robot.

Human interaction is, and has always been, vital to a high-quality customer experience. Even Facebook supports this claim, as they’ve partnered with several companies to help with the hand-off from bot to human during live chats. Facebook COO, Sheryl Sandberg, has publicly stated that, “…we simply don’t have the technology yet to actually imagine that a bot could replace humans in the sales process.”

So, what is the human role in a bot-dominated future?

Remember that frustration every time you call customer service and end up with an automated voice (IVR)? That’s where we are with bot chats. And guess what happens every time a bot fails? You end up talking to a real, live customer service representative — back to square one.

There’s no doubt we’ve come a long way with artificial intelligence, but for the strides we’ve made, we’re still a ways away from the creation of a bot that can successfully pass all facets of the Turing test. While a bot can handle a good portion of a conversation with a human, there will undoubtedly be times when it gets confused and fails (especially when it comes to switching from one topic/area to another). In these situations, your bot chat gets handed over to a human to complete your engagement.

Even with all the buzz today, the bot is nothing new. In the late 1990s, when AOL Instant Messenger was all the rage, I remember chatting with SmarterChild. At its core, SmarterChild was essentially an early version of a bot. You could chat with it about school, life or sports — much like you would do with your real friends. SmarterChild worked great (most of the time) and seemed quite sophisticated. Though, in fairness, a majority of the chats were conducted by twelve-year-olds.

So, the real question today is whether the bot is really going to define the future, or are we all just falling for the same hype we did when we were ‘tweens?

To answer this question, it’s important to understand the technology behind a bot. While we’ve seen tremendous strides and advances in computer technology and software development over the last 20 years, bot technology has essentially been siloed into two categories: those based on simple logic trees (SLT) and those that rely on natural language processing (NLP) or machine learning (ML).

An SLT bot relies on the traditional logic tree to gather information and redirect the user. For example, an insurance bot may ask several questions to determine which policy is ideal for you. If your answers match what the bot has anticipated, the experience will be seamless. However, if your answers stray from those predicted and stored in the bot database, you might hit a dead-end. If you’re lucky, you’ll be handed off to a human to complete the interaction. If you’re not, you’ll end up in bot purgatory. Most bot technology today relies on SLT.

An NLP and machine learning (ML) bot is meant to function more like a real conversationalist by picking up keywords and phrases from the user’s input to gather information, instead of requiring direct answers to specific questions. In theory, this category of bot sounds like the better option. Examples of this type of bot are Apple’s Siri and Amazon’s Alexa.

While Siri and Alexa are pretty good at simple functions like giving the weather or telling a joke, they still have a ways to go with complex functions and lengthy commands.

Regardless of whether you are engaging with an SLT or NLP bot, the likelihood of ending up needing to speak to a real person is incredibly high. SLT bots often lack the complexity we expect from technology today, while we are unable to fully utilize the technology necessary for NLP or ML bots at this point.

Fortunately, customers enjoy the efficiency of interacting with a real person. While the trend is moving away from long, formal conversations, customers still expect the same quality of service from chat, whether live or bot. In fact, a recent study by American Express found that 78 percent of customers have bailed on a transaction or not made an intended purchase because of a poor service experience. The same study also found that 67 percent of customers have hung up the phone out of frustration when they could not talk to a real person. In most of those cases, the customer was forced to endure a dialogue with a bot.

We will likely see a two-stage industry transition as we attempt to adopt bot technology into daily transactions. The first will be very human interaction-intensive, as real people will be required to take over every interaction that a bot fails to handle. The risk of a poor customer experience is simply too much for top brands to stomach, so staffing their engagement centers to take over when the bot fails will be a very real side effect.

It seems inevitable that at some point ML and NLP will allow for a bot to become more intelligent, to a point where the failure rate is minimal. At that time, it isn’t unreasonable to believe that a majority of interactions will take place on bot channels. Will the bot channels stand alone, or will they be integrated into the existing channel landscape? If they exist on their own, what happens to the other channels?

In a scenario where the preferred customer engagement moves away from a brand’s website and onto a bot channel like Facebook Messenger, a question of scale is presented. Even with low failure rates, the number of human hand-offs will most likely grow because of the sheer volume of engagements that will likely occur.

It’s more than likely that we’ll see exponential adoption of bot technology, but human capital for engagement is both inevitable and necessary for sales and customer support. Depending on the lifetime value of the customer and the margin of the product, the specific value of an engagement with a real person can be very significant.

As we prepare for a more automated future, it’s important to not forget about the role of the human in engagement. And, as great as R2-D2 was in “A New Hope,” we all must remember that there was a human in that bot.

Chatbot can answer all of your cannabis questions

With the rapidly spreading legalization of medical marijuana, and it becoming a more available and acceptable form of treatment for individuals, many people are left with questions that need answers.

A Canadian company has stepped up to this task by creating the world’s first AI-powered chatbot for answering questions about cannabis. Lift Cannabis Co., based in Vancouver, has developed a chatbot named, Elevator, that is designed and programmed specifically to answer questions pertaining to marijuana.

Elevator offers a more private and personalized method to help consumers learn about marijuana.  With decades of misinformation about this plant, customers are in need of help regarding the right strains of cannabis for their health issues, medical benefits, how to safely use it and where and how to purchase it.

Finding answers by using artificial intelligence

Elevator is an artificial intelligence app that is text-based.  People can ask questions and get immediate and accurate responses. Tyler Sookochoff, the CEO of Lift, which is the company behind Elevator, stated that the goal for Elevator is to make one’s cannabis learning experience easier.

“Cannabis as a legal industry is quite new, yet there’s so much false and misleading information floating around the web about its effects, benefits, and how to access it safely – and if you’re new to medical marijuana it can be really hard to cut through all the noise to get answers specifically tailored for your situation.  We developed Elevator as a response to this, and we’re harnessing the latest in AI bot technology to give real-time, credible and informative responses. We do the heavy lifting behind the scenes on your behalf, in terms of sorting through and collating only the most credible and up-to-date cannabis information. On top of this, Elevator will remember your questions and its own responses, and the more it learns about you the better it will be able to assist your exploration in the future.”

Lift is already an information leader in the Canadian medical marijuana industry.  Therefore, Elevator is a sensible extension of the company.

“We see Elevator as your personal ‘Cannabis Concierge’, and by using a simple and familiar chat interface – ironically, the same type of interface many people have been using for years to text their dealers – it has universal appeal,” says Sookochoff.

Elevator will demo at Lift’s trade show later in May in Toronto.