SXSW 2018 Day 5 Session 4: The Power of Ideas to Transform the World is Accelerating

Session page, including audio: https://schedule.sxsw.com/2018/events/PP99241

Video of session: https://youtu.be/CiLmyA-gAZk

Ray Kurzweil, Google

Jessica Coen, Mashable

Ray: A computer will pass the valid Turing test by 2029; we will achieve the technological singularity by 2045 (https://en.wikipedia.org/wiki/Technological_singularity).

The pace of progress is increasing.  Ironically, people feel things are getting worst because information about poverty, hunger and other problems is more available and immediate; and we have an evolutionary disposition to pay attention to and focus on bad news.  However, every aspect of human life is getting better.

We will see improvements to longevity, and we are in the most peaceful time in world history.

Jessica: Elon Musk was quoted as saying that AI is more dangerous than Nuclear weapons.  What do you think?

Ray: There are risks to AI, such as privacy and bias, but I’m an optimist.  I don’t see AI as an external entity; we will have brain extenders and body extenders, so this should keep it safe.

As an analogy, 40 years ago people saw the promise and danger of biotech, and formed guidelines that govern the domain quite well.  This is a good paradigm for AI, and last year, for the first time, there was a similar type of conference for AI to discuss how to prepare.

Jessica: Do we have enough oversight on the domain?

Ray: We have a lot of regulations.

Jessica: Self-improving tech – how do we prevent it from running away from us and diverging from human interests?

Ray: Any time AI impacts the world there’s a lot of observation of it and monitoring of it.  There is a concern of its use in the military; the best strategy here is to apply our own value sets to it.  AI is an implementation of human values.

Jessica: There is a large lack of diversity in the engineers building the AI; how do you prevent that from impacting the AI that is being created?

Ray: That’s a complicated issue, because there is inherent bias in the data software is learning from and trying to correct it risks inserting the corrector’s values.

The world has a pessimistic view.  People feel things are getting worst in all parameters.  However, in 2020 we will be able to print cloths for pennies per pound.  We’re moving towards the ability to print out housing and support vertical architecture.  The 50% deflation rate of IT will apply to other resources as well.

We will be creating a synthetic neocortex in the cloud – by 2030 we will have a good model of a synthetic neocortex; we’ll be able to connect these with our own.  But even before that, we already have brain enhancers – we are already augmenting our creativity and capabilities.

AIs are not yet at the highest level of meaningful interaction.  Just a few weeks ago computers passed a reading comprehension test at average adult level.  They still can’t human level complex comprehension, but they’ll gain that ability by 2029.

Jessica: Should conscious AIs be granted human rights?

Ray: once they reach consciousness (2029), yes, they should.

Question: What’s the next major industry to die?

Ray: They don’t die, they just transform.  Agriculture will transform even more with vertical agriculture.  Transportation will be transformed not just by autonomous vehicles, but also by virtual and augmented reality, which will prevent the need to physically travel in the first place.

Question: Will we reach immortality by 2050?

Ray: 10 years from now we’ll have a flood of medical enhancements.  The third bridge is medical nano-robots, enabling non-biological blood cells and T cells, inner-body monitoring devices and organ augmentation (by around 2030s).

Question: What near term (2-3 years) innovation are you most excited about?

Ray: In that timeline, 3-D printing at the submicron level.  Vertical agriculture is about 5-6 years away.  VR/AR had a premature start, and will take another 3-4 years to mature.

Question: What are futurists going to do after the singularity?

Ray: It’s hard to see beyond that horizon.

Question: What do you think of the theory that we live in a “Matrix”-like simulation?

Ray: If it was created for someone else, we should do our best to keep them interested!  There’s another theory of thought that says the universe itself is a computer.  There is some evidence to this: there are about 50 physical constants that exist, and could only exist in a very specific and narrow range for the universe to be able to work.

SXSW 2018 Day 5 Session 3: The Final Human-Computer Interface, Your Brain

Session page, including audio: https://schedule.sxsw.com/2018/events/PP98376

Ramses Alcaide, Neurable

How do we interface with our technology?

Voice: Voice interfaces are limited based on locale of use; noisy environments get in their way.

Motion tracking: Great for things that are next to you, but harder to use for things that are out of reach.  If you are pointing at a faraway object, it can be harder to be precise.

Eye tracking: The problem with this interface is that it’s always on; it’s harder to “start” and “stop” using it.  Also, it can be hard to understand what it is the person is wanting to select, and grouping items together can also be difficult.

A brain interface is a great equalizer, and it is not limited by human disabilities; the brain interface is a great democratizer.  Other interfaces do not have this advantage.

One more advantage comes in the ability to interact with augmented reality.  Interaction devices make the interaction with augmented reality more cumbersome and reduce the immersion.  If you have to use your phone, or hand-held devices (as in VR) to control the augmented reality you see, you degrade the experience.  With a brain interface that is not an issue.

In addition, sometimes the hands can be active – for example, a surgeon performing an operation, with augmented reality, cannot use hands to interact with augmented reality, and there may be noise in the operating room interfering with a voice command.  So a brain interface is optimal in such a case.

Different types of brain interfaces:

Proactive – use the brain interface to issue commands such as typing on a keyboard or controlling a pointer.

Very slow, and the equipment required is cumbersome.

Reactive – using our innate information to interact with a software

Some examples:

The brain wave that is generated when we hear a sentence that is nonsensical

Brain waves that indicate urgency or intensity (a deer steps in front of our car)

Brain wave that trigger when something you expect to happen, happens.

The problem with these types of brainwaves is that they are slow and require gel caps, so they are restricted to normal day-to-day usage.  Neurable is creating a breakthrough in this domain, allowing simpler caps to be used and interpreting signals in real-time and with high accuracy:

SXSW 2018 Day 5 Session 2: The Humanoid Future of Journalism

Session page, including audio: https://schedule.sxsw.com/2018/events/PP80710

Robin Govik, COO MittMedia

Hanna Tuulonen, Göteborgin Sivukonntori

Hanna: Adaptation to automated (computer generated) journalism takes time.

Robin: We should not see automated journalism as scary – it’s not about computer journalism vs human journalism – it’s a collaborative environment.

What is automated journalism best at?  It’s not about creating a gripping survival story or covering a corruption scandal.  There’s room for automated journalism in weather reports, sports scores, and other similar types of topics.

Because of the revenue pressure on newspapers, the number of human journalists is going down, and we need a robot to pick up the slack.  AI can also help in investigative journalism, in analysis, searching and tracking of data which previously used to be handled manually by people.  Now the robot can manage the “dirty work” of sifting through tons of data and the human can focus putting all of the massive data in context as a coherent story.

AI can also personalize articles to the readers based on their interests, making the articles more compelling and improving reader loyalty.

Hanna: Finland and Sweden are pioneers in this domain.  Swedish people are less worried about AI displacing humans and leading to job loss because of the strong social safety net.

Robin: There are two kinds of news automation methods:

The first kind is software that generates and publishes stories directly on the site; mostly stories on sports.  Right now our newspaper publishes about 3000 AI-written articles a month.  The AI was initially used in one sport but expanded to other sports as well, due to the demand it generated.  This way one can cover a lot more teams and leagues than was possible previously, which appeals to local interest.

The second type is AI that creates short paragraphs that are sent to editors, who decide how to publish them.  These are frequently in areas like traffic reporting, weather, public transportation and so on.

Can this type of journalism generate money?  The key is relevant content.  We conducted data analysis to see what content converts users to paying users, and we can see what type of articles have high interest.

An AI can generate a richer experience than just a text story – it can get images from Google street view or from public domain image libraries, aggregate them into the story and create a richer experience.  For example, real estate reports generated by AI have been very successful.

Hanna: You can use AI to increase both quality and quantity:

• In quality, you can analyze data much more accurately, with fewer mistakes, when you use AI analytics.  This leads to better quality news reporting.
• In quantity, an AI can generate tens to hundreds of articles per second.  This leaves Journalists free to write stories they didn’t have time to previously on more interesting and complex topics.

Finnish journalists were asked how they felt about working with news robots, and answered that they loved it as it took away repetitive and boring jobs.  But this was their opinion _after_ working with the robots; before they had negative attitudes to them.  It only took about two weeks of working with the robots for them to change their view.

Robin: Two thirds of readers can’t tell that the story was written by a robot.  This allows making much wider and smaller sets of news reportable.

Hanna: Articles written partially or fully by machines score higher credibility with readers.  However, they frequently score lower on readability; you still need humans to write engaging stories.

A study that was done in south Korea showed that people felt journalists are corrupt, but algorithms are fair.

Robin: The humanoid future of journalism includes:

1. Personalization – this is much easier with robots, as it’s easy for them to generate multiple versions of a story, based on the preferences of the person reading it.  However, one thing to look out for is creating opinion bubbles, where people are only receiving news that reinforces their existing personal beliefs and points of view (as happens in social media).
2. Robot journalism + citizen journalism – an example of this is using robot journalism to report on a sports event, while using a service which sends messages to participants and possibly spectators to get quotes and pictures from them.  The combined result adds humanity to the story, but it does present a problem of curation of the provided input.
3. Dynamism – the ability to change formats of stories to support personalization.  For example, on the BBC an article can be rewritten every time it us loaded, which helps keep it up to date and personalized.
4. A new role for journalists – some journalists will focus on working with robots, finding data to feed them and working jointly on stories.  Others will continue the traditional and independent role they had (but now will be less encumbered with menial tasks).
5. Ethics and writing - Human journalists still excel in their ability to write and in soft domains such as ethics and emotions.  Robot journalism will help bring back quality of writing into focus.

So summing up:

The pros of robo-journalis

• Better journalism
• More content
• Increased quality

The cons:

• Bad data drives bad stories
• Potentially dull and spiritless writing
• May pivot news coverage to topics where there is plenty of data available.  Areas missing data sources lead to topics being under reported.

SXSW 2018 Day 5 Session 1: Where We’re Going We Don’t Need Roads: Fly eVTOLs

Session page, including audio: https://schedule.sxsw.com/2018/events/PP80710

Antonio Campello, Embraer

Mark Moore, Uber

Kym McNicholas, The Innovators Network

A year ago Uber and Embraer declared a partnership to create an eVTOLs (electric vertical take-off and landing vehicles) service, with a target date of 2020 for a test launch and a commercial launch in 2023.

Antonio: We feel that small, available aircraft will be an important way to reduce traffic jams.

Mark: We’re creating a new transportation solution with an optimal user experience.  The first tests will be conducted in the Dallas Fort Worth Airport (Texas) in 2020.  As of now, there are two companies committing to create eVTOLs, with Embraer being one of them.

These will be vehicles driven by commercially rated helicopter pilots; autonomous vehicles will be further down the road.

Antonio: Having more competitors in this domain is a good thing for the industry as a whole.

Kym: How do you deal with the FAA, that is very slow to accept change?  For example, currently it mandates two pilots to fly a commercial aircraft.

Mark: We need to work together with everyone to make collaborative progress.

Kym: What is the vehicle like?

Mark: The longest trips it will be able to make is 60 miles; we expect the average trip to be about 23 miles.  You will be able to charge its battery in about 5 minutes.  It will carry four passengers (and a pilot).  It will be very quiet – around 20 decibels quieter than a traditional helicopter.  It will have distributed electric propulsion, designed to support multiple failures.  Different companies design it in different ways.

Antonio: We will need to deal with air traffic management and interconnection among vehicles.  We need to ensure a full ecosystem of transportation (car to eVTOL to other systems).

Kym: What does the seamless experience look like?

Mark: It’ll just be another option in the Uber app.  You select Uber Air, a car will take you to the local skyport, from where you fly to a city center, where another Uber car can wait for you on the other side.  It will help make living in the suburb and commuting into the city cheaper than living in the city.

Antonio: We are designing and building the solution for affordability – everyone should be able to afford it.

Mark: Uber has no interest in this project if it’s just an elite service – we’re committed to making it an affordable one.  The air vehicles don’t waste time in traffic, so they are more productive, so they would be less expensive.  Plus, they have ride sharing, which reduces price.

Kym: What about controlling airspace?  Multiple companies are involved, who will control the airspace?

Antonio: Depends on the certification authority, the software, the integration and other factors.  Traffic control is a key element to help support 3D movement.

Mark: Uber’s main contribution to the ecosystem is its fleet and network management capabilities.  We would use an Unmanned Traffic Management system (UTM) built for NASA, which enables many thousands of vehicles to fly at low altitudes.  It works like a packet management network system.

Antonio: Current aircrafts are highly computerized and communicate with the ground.  The technology is bringing huge benefits, and in this domain it will help as well.  It’s an incremental step over what we have currently.

Kym: Will the software be shared?

Mark: No.  It will be enabled for our partners.  Other networks will be built.

Kym: Where does the FAA come in?

Mark: We’ll be collaborating with the Dallas Fort Worth Airport.

Question: Will this not amount to privatizing ownership of airspace?

Mark: It’s not privatizing ownership of airspace, but it is private ownership of networks or corridors in airspace.

SXSW 2018 Day 4, Session 4: GDPR: What Does It Mean For US Business In The EU?

Session page (no audio available): https://schedule.sxsw.com/2018/events/PP99189

Tim Bell, DRP Group

Why does GDPR matter in the US?  It matters because it is directly enforceable against US companies (and companies from other countries).

GDPR was brought about because of frustration in the EU with how US companies treat customer data.
A brief timeline of important relevant recent privacy related events:

2011: Action was brought against Facebook in Ireland because of Facebook’s facial recognition software.  Facebook disabled the feature in Europe.
2013: Following the Snoden revelations, the US-EU safe harbor for data transfers collapsed
2016: 2016: Uber lost data for 57 million drivers and passengers, and only admitted to it a year later.
2016: WhatsApp lost a case in Holland for not appointing a local data representative, as regulation requires
2017: When Facebook bought WhatsApp, they promised they would not share data across platforms.  When it was discovered that they were doing this after all, French authorities intervened and stopped it.
2018: Belgium court rules against Facebook tracking people who are not even Facebook members.

GDPR, which was agreed upon in 2016, and enforceable starting May 25, 2018, the stakes are raised with the arrival of GDPR:

• GDPR affects the data of any people in the EU, regardless of whether the person is an EU member or the company operates outside of the EU.
• The maximum punishment is extreme – the larger of $25 million or 4% of global turnover

Some GDPR terminology:

• Personal data – data which can identify an individual, including IP addresses, work contact details, biometric data, most cookies
• Data subject – the person who could be identified by the personal data
• Data controller – the organization which determines how the personal data is processed
• Data processor – an organization which processes personal data on behalf of the data controller
• Data Processing – any operation performed on personal data, including collecting and storing

What are the GDPR obligations?

Privacy by design – this is more a state of mind than law.  It requires organizations to have data protection in their DNA.

Lawful basis for processing – typically assumed to be consent, but must be active (a pre-ticked checkbox is not good enough).  However, there are some other justifications for processing data, such as performance of a contract with an individual, complying with legal obligations or performing a task in the public interest.

Data protection officer – a position which is mandated in some cases, depending on company operation in Europe

Data protection representative – mandatory for companies that are not established in the EU

Processing agreement – A data controller that appoints a data processor has to have rules in the contract to describe how data will be handled.

International transfer of data – when transferring data inside or outside of the EU, the data must be transferred with the right protection.  Some countries (such as Israel, Argentine, and others) have equivalency laws, so they are considered “in the EU” for the purpose of data transfer.  For transfer between the EU and the US, “Privacy Shield” replaced Safe Harbor as the data transfer agreement between the two.

Privacy notice – when collecting data from users, the data controller has to tell the users what it intends to do with the data, up front and free of charge.  This needs to be concise, transparent, intelligible and accessible.

Subject access requests – individuals can request to get their data and what a company holds on them.  They also have the right to be forgotten.  A request must be answered in a month, and the company cannot charge for it (but can refuse excessive requests).

Data breach notifications – a company must notify the relevant authorities in each impacted EU nation of any data breach that occurs.  The notification must be done within 72 hours of becoming aware of it.  If there is a high risk to the data subject, the company needs to let them know immediately.  The data processor must tell the data controller without undue delay.

Data processing records – must keep records of all data processing activities for inspection

Data processing impact assessments – must undertake assessments on how processing will impact customers

Estimations are that no more than 15% of companies will be ready when GDPR kicks in.  So what preparations need to be taken towards GDPR?

1. Appoint a DPO if needed.  If it is not required, appoint someone to manage data regulations.
2. Know your data – audit both physical and digital data.
3. Cleanse your data – make sure its relevant, have consent for it.
4. Appoint data protection representatives (if you don’t have an office in the EU)
5. Ensure security of data (physical and digital)
6. Update procedures to have privacy by design
7. Ensure staff training – people are the weakest link
8. Consider basis of processing – was the data collected with consent? Is the consent adequate?
9. Prepare for data events – data requests, data breach events, etc.

Q: How do you deal with right to forget for third party data processors?
A: Request only applies to you, but you need to disclose the third parties so customers can go to them and make requests from them as well.
Q: Will we see a shift in companies data gathering practices?
A: Very likely.
Q: Does it impact EU citizens living outside of the EU?
A: No – it only applies to people living in the EU.
Q: What’s the impact of Brexit on GDPR?
A: When it will happen, there will be a period of transition and confusion, but even after leaving theres a UK law that’s similar.
Q: Do you need consent for each third party you work with?
A: No – you need to list all the third parties and then have a single consent box.
Q: How about companies like Lyft or Taskrabbit, who have millions of subcontractors – how does that work?
A: What you should do is give subcontractors as little info as possible, and don’t give them continued persistent access to it.  Keep data at the controller level.
Q: What about keeping lists of subcontractors?
A: Have to be very careful about how you store it – same rules apply to it.
Q: What about chat messages?  If a user gives personal data to another user, do you have an obligation to treat it like personal data?
A: You can’t control how users use data they get from other users.  You should set up rules to inform users how data is going to be used.
Q: What about right to be forgotten in this context?
A: It applies as well; each user has the right to withdraw its data.
Q: What about the right to be forgotten with Google search results?
A: You can request Google stop linking to them today, but it’s not clear how GDPR will further impact that.