SXSW 2018 Day 3 Session 1: 2018 Emerging Tech Trends report

SXSW 2018 Day 3 Session 1: 2018 Emerging Tech Trends report

Session page (including audio): https://schedule.sxsw.com/2018/events/PP97100

Amy Webb: Quantitative futurist, founder, Future Today institute

11th annual edition of the report, 225 emerging tech trends and some weak signals across 20 different industries for 2019.  One of the biggest mistakes people make is to only pay attention to the trends in their own industry; you need to simultaneously pay attention to the future of many industries.  They are connected.

The signals of 2017:

• Outside the US, China and Russia are setting up to extend a single leader’s rule, while in the west there is instability – In Europe Brexit is sowing instability and in the US people are being shuffled in and out of the administration almost weekly.
• ICO is the new IPO, and everyone is engaging in it.  Uber had an ICO, Kodak had an ICO.
• AI is everywhere – in toothbrushes, in toilets
• Robots being built for everyday activities – Robot baristas, robot burger flippers, robot sock sorters.

There is wiz-bang technology all around.  How do we distinguish between the trendy and real trends?  Tech trends share four characteristics:

1. Tech trends are driven by basic human needs.
2. Tech trends are timely, but they persist over long periods of time.  On the report, may have been there for a decade or more.
3. Tech trends evolve as they emerge – they are not static.  They change as they interact with consumers, government, environment, etc.
4. Tech trends have dependencies and multiple points of convergence.  They materialize as a series of un-connectable dots which begin as weak signals on the fringe and move to the mainstream.

The emerging tech trends that are worth paying attention to usually cannot be encapsulated in a buzzword; they usually take a sentence or two to describe.

When confronted with an early tech trend, it’s easy to dismiss it or miss it.  That’s why many organizations with smart people do miss them.  Senior leadership frequently goes through the cycle of doom:

Paying attention of trends is the way to escape the cycle of doom.

How are the trends identified? Through a data driven process:

• Identify interesting technologies and look at them through 10 different lenses – 10 modern sources of change:

• Identify weak signals at the fringe
• Use pattern recognition to identify emerging trends
• Use trends to build possible, plausible and probable scenarios for the future, to calculate risk and opportunity.  What are the second, third and fourth order implications?
• Apply to strategy, to make better strategic decisions.

Important to track all the landscape of a technology and trend and make connections among all relevant topics.  For example, for cryptocurrency:

If you can track the trends and all its interdependencies, you can change the cycle of doom to the cycle of opportunity:

Never heard of it -> Do dedicated research -> calculate timing -> Have first mover advantage -> inform our strategy -> motivate and engage our workforce -> Brand positioning strategy -> sustainable growth, better planning for future.

Key findings from reports – 3 key findings clustering 14 trends in 3 trend clusters, showing connections and plausible scenarios for the future:

Key finding 1: 2018 is the beginning of the end of smartphones

The trend clusters that support this key finding are:

1. Digital assistants
2. Machine comprehension and natural user interfaces
3. Voiceprints
4. Faceprints
5. Generative Algorithms
6. Augmented + Mixed reality

New technology is bubbling up from the horizon and smartphones have not leveraging it.  Smartphones have peaked, and in the last quarter we’re starting to see the sales drop off.  Part of the reason is we’re not replacing our phones as much as we used to, but why?  Because we’re only seeing incremental benefits.  If in the past we merged all of our devices – camera, mini-disk, computer, etc. into one device, the direction has reversed and now we’re decentralizing the computation – smart watches, smart wristbands, smart earbuds, and pretty soon smart glasses.  Smart glasses combined with smart earbuds and a watch is where we’re headed.

Trend 1: Digital assistants will become ubiquitous.  We are surrounded by digital assistants.  Within the next 10 years we’ll move to conversational interfaces or non-visual UI.  By 2021 50% of the people living in industrialized nations, over half their interactions with machines will be done with voices.  This connects to another trend:

Trend 2: Machine comprehension.  For example, today, when you type a question into Google, you mostly get a list of links.  Machine comprehension will allow the interface to provide the actual answer (some of this is already showing up in Google) or at least pull up where in the article the answer will be.  This is laying the foundation for our future interaction with our machines.  This is going into all of our machines going forward.

Trend 3: Voiceprints. The ability of a computer to identify your unique voice is key to enable ubiquitous voice commands.  We all sound different, using different inflection, sounds, and intonation.  You can use all of this to create a voiceprint, like a fingerprint.  There’s a lot of research around this right now.  It can be used to recognize you, your age, your health, your emotional state, and even where you are: what type of room you are in; what the walls are made of; how many people are in the room with you and even a good approximation of where you are in the room itself at the time you are speaking.  Also, MIT is researching machine’s ability to replicate the voice print, and they are able to do it so well that humans cannot tell the difference.

Combining the four trends above gets us to a future where you can access your accounts using your voice command to your assistant, making passwords unnecessary and deprecating other forms of biometrics such as fingerprints and retina scans over the next decade.

Trend 4: Faceprints.  Every person has different structures of face – bone and flesh below the skin.  Even if we shut off all lights in the room, you could still use thermal prints to identify a face uniquely:

With a 3D scan you could make a faceprint of someone, which a computer could use to identify a person from multiple angles.

Megvil, a Chinese company, is combining faceprint technology with body posture and gait analysis in a lot of applications: Alibaba enables smile to pay, in airports they deployed face recognition to catch criminals, and they have cameras that catch and publicly shame Jaywalkers by identifying the people jaywalking and publishing their personal information on big display boards and over social networks.

Trend 5: Generative Algorithms.  Generative algorithms enable taking basic data and generating extrapolated data from it in a way that makes that data seem natural.  instead of having to take a 3D scan of a person, which is complicated and expensive if you want to apply it to the whole population, you can take a regular photo of them and generate an accurate 3D scan.  Or, alternatively, if you have video of someone, you can take the motion characteristics and apply them to a 3D modeling of someone else.  Combined with voiceprint generation technology, you could now use one video to generate fake video of someone else that looks and sounds as though the other person was filmed, even though they were not.  This technology has advanced to an extent that it is commonly available and can be accessed online.

Trend 6: Mixed reality.  VR came quicker, but AR has longer stay.  Its projected revenue growth:

10 years forward: where does this take us by 2028?

Framing	Outcome	Likelihood
Optimistic	Companies work together to make the operating systems interoperable. Consumers have total control over their data.  There’s transparency and we can block ourselves to preserve our privacy.  We’re all wearing combinations of smart glasses, watches and earbuds.  We are information millionaires.  We have better information about each other while we are interacting – could bring about the end of social isolation (looking to each other again, instead of at our devices).  We have better security for all our devices.	0%
Pragmatic	The developmental transition away from smartphones leads to different portable OS’s.  We don’t have interoperability among devices.  Consumers grow frustrated.  The ecosystem develops on a decelerated track.	50%
Catastrophic	We have a new digital divide.  The wealthy get the best devices and OS’s.  Everyone else is subjected to a barrage of unwanted marketing.  We realize too late that the era of privacy is over.  Leads to heavy-handed regulation.  We are all miserable.	50%

Key finding: Era of Artificial Intelligence.

The trend clusters that support this key finding are:

1. Reinforcement machine learning
2. Generative adversarial networks
3. Machine reading and comprehension
4. Multitask learning

We’re all using AI every single day; it doesn’t feel like AI because a lot of what we use as part of our daily lives, so we don’t notice it.  This type of AI is Artificial Narrow Intelligence (ANI), where a system can perform a single narrow task as good or better than a human.  For example: ABS, spam filters, automatic playlists, etc.

AI is not a tech trend by itself, it represents the next era of computing; there are 27 different trends in the AI domain.

Trend 1: Reinforcement learning.  AI learns through reinforcement, like animals (and humans) do.  You can train a robot through positive and negative feedback.  Recently a Neural network developed a sub system to do a work for it for efficiency – sort of a child AI.  This AI was trained to recognized various visual images.  This system learned via reinforced learning, and achieved very high accuracy (82%), outperforming systems that were created by humans.

Trend 2: Generative adversarial networks. Software developers have been exploring ways to find and take advantages of weaknesses in AI thinking patterns, in particular in recognition algorithms, by using adversarial learning.  Adversarial learning takes the form of trying to “hack” the learning algorithm by finding and taking advantage of small flaws in it, so that it fails on what should otherwise be a success.  In the case of image recognition, taking a picture the algorithm correctly recognizes, and changing a few pixles in it so it fails to recognize it.  For example, the following image:

The two images on the left are the original “clean” images, which the algorithm successfully recognizes as washer and dryer.  The two images on the right had a few pixles altered in them, such that the algorithm no longer recognized them as a washer and dryer, but rather as a safe or a loudspeaker.  To a human, they still look the same.  Taking an adversarial approach allows finding flaws in the algorithms, and then the machines can be trained to avoid these flaws.  But further than that, an AI can train itself to avoid failures by purposefully adding small mistakes and analyzing the results, and this is generative adversarial learning.  This allows an AI to help learn without needing human guidance.  This is particular useful against malevolent adversaries, who are looking for flaws in AI to cause harm.  For example, research has shown putting a few stickers on a stop sign can make an AI recognize it as a different sign.  This could have negative consequences on self-driving cars, who rely on AI for image recognition.

Trend 3: Consolidation.  There is a lot of consolidation in the AI ecosystem.  Currently 9 companies hold the future of AI:

These companies all partner with academic institutions, but the money flows from these companies to the researchers.  Three of them – Tencent, Alibaba and Bidu – are Chinese.  China as a nation is the largest world investor in AI and Robotics, sovereign fund level of investment ($200 Billion in the next 3 years).  But unlike other investors, they are not looking for a financial return on investment, they are looking for IP.

Trend 4: Splinternets.  Different countries are trying to assert control over the internet and what information passes through it.  China and Russia already have to a large extent succeeded in this, Brazil may create its own curated version in the future, and the US may be as well.  There is also regulation that could splinter the internet – GDPR applies in Europe, Canada has laws forcing Google to scrub pirated products from search results, Germany has laws forcing social networks to delete hate speech within 24 hours of posting and so on. Down the line, something posted in one country may not be visible in a neighboring country.

15 years forward: where does this take us by 2033?

Framing	Outcome	Likelihood
Optimistic	The developmental track of AI changes to include more diversity.  Norms and standards are adopted to ensure transparency – how decisions are being made.  The Federal government devotes money to basic research.  Government creates new social systems to deal with unemployment.  Less consolidation in the commercial sector – the big 9 still make money, so everyone is happy.	0%
Pragmatic	We continue on this path without big changes.  Research becomes commercialized.  Much of the IP drains to China, which becomes a formidable business threat.  US and Europe have a hard time competing.	60%
Catastrophic	People don’t understand how decisions about them are being made.  They revolt.  Misinformation is worse than anything we’ve seen in history.  Widespread technological unemployment.  Governments aren’t prepared.  Collapse of civil society.	40%

Key finding: Biology is emerging as the most important technology platform of the 21st century. 

The trend clusters that support this key finding are:

1. Genome Editing
2. Biological DVR
3. Molecular self-assembling robots
4. The Human cell atlas

Trend 1: Genome Editing.  CRISPR/Cas9 is a genome editing technique that allows precision modification of DNA.  It is a hot area of funding and R&D.  It promises a multitude of revolutionary breakthroughs in healthcare and biology.  For example: instead of external or internal devices used to pole the blood sugar level of diabetics, a gene that activates a skin tattoo whenever blood sugar drops.

Trend 2: Biological DVR.  New research will allow us to record and store information about our cells as they age.  This will allow us to observe the aging process, but also, if we can quantify aging at a cellular level, save earlier versions of ourselves. 

Trend 3: Molecular self-assembling robots.  Self-folding, single strands of DNA that can be made inside living things.  This isn’t exactly new – goes back to first experiments in the domain in 2010.  This will be able to lead to plug-and-play nanomedicine - biological autonomous nanobots that can deliver medication precisely where it’s needed and optimized individually.

Trend 4: The Human cell atlas.  A joint, global effort of researchers to map every facet of every cell of the body.  Once that happens, it will enable better understanding of how to apply the other new capabilities in biology.

25 years forward: where does this take us by 2043?

Framing	Outcome	Likelihood
Optimistic	Our bodies are being monitored continuously by smart medical systems and by us.  No more hospital waiting rooms.  No more 20 pills a day, each one counteracting the next.  Less probability for another opioid crisis because we won’t be prescribing pain killers. Lots of new jobs.  Hybrid-trained doctors/roboticists, etc.  People live longer.	0%
Pragmatic	Pharmaceutical companies, hospitals and insurance companies can’t transition away from their current business model.  Our current health systems around the world lag behind the technology.  Lots of strain and aggravation.	70%
Catastrophic	No one plans ahead for precision medicine and people living longer.  Like we see today with phones and cars, the wealthy get the latest and newest versions of monitoring tech.  Everyone else has to trade in their personal data for use in experiments.  For them, privacy is gone. To deal with overpopulation in wealthy nations, the autonomous systems – the robots inside us – decide who lives and who dies.  Nanobot abortions based on probability of success and contributions to society.	30%

We still have the ability to choose the future we want.

• If you are a businessman, you have an opportunity to start making smarter strategic investments.  Do not be distracted by noise, such as bitcoin.
• If you are a government person, don’t wait, and don’t politicize science and technology.  Now is the time to think about the future, now when it arrives and you are feeling the pressure of its impacts.
• If you are a regular human, you have a chance to fight for the future you want to live in, but you have to start taking action on trends in the present.

Trends report: www.bitly.com/FTITechTrends2018

All forecasting tools: www.bitly.com/FTITrendsFolder2018