Exploring Opportunities in Artificial Intelligence

Just a few years ago the common perception was that we were in an AI winter.

Although there were lots of narrow AI applications running in the background of our daily lives, there wasn’t much enthusiasm.

But quietly in the background, a revolution was building thanks to progress across a few key areas. These areas would soon converge to produce breakthrough after breakthrough and put us on the verge of what many believe to be the most important event in human history.


Key Advance 1) More Data

Andrew Ng of Baidu explains that a massive amount of data is needed. If you put 10x the data in many of these algorithms they work, put 1/10th in and they don’t.

Previously it was very difficult to get hold of the massive amounts of data required to feed the AI systems, but thanks to the internet researchers have tonnes of data to train their neural nets.


Key Advance 2) More Computing Power

If you only have the computational power to build a small neural network it doesn’t work. But computer power has continued to increase and prices have dropped.

Moore’s Law may be stalling, but the Law of Accelerating Returns is not.


The Law of Accelerating Returns. What Steve Jurvetson calls “the most important graph ever”. 

GPUs are much better for training neural networks than CPUs and have provided the computer power needed for these algorithms to function.

Infrastructure has also improved. Today it’s possible for anyone to rent a massive amount of GPU power on cloud computing platforms (AWS, Microsoft Azure, Google Cloud, IBM Cloud).


Key Advance 3) Better Algorithms

Neural networks have been known about for decades, but most researchers had given up on them

Geoffrey Hinton of Google is one of the few who stuck with it. Despite his peers calling it a dead end, he believed that it was the right approach. It turned out he was right.

Hinton learned how to stack neural networks dozens of layers deep (deep learning) which enabled vastly more calculations and now he is considered “the godfather of neural networks”.

With these 3 breakthroughs in place neural networks finally began to work. And they worked better than almost anyone expected.


The Tipping Point: ImageNet 2012


The ImageNet project was created in 2009 to judge how well computers can see.

In 2011 computers had a 26% error rate when trying to label images. Humans only had a 5% error rate.

But in 2012, Hinton’s team made a breakthrough and reduced the error rate to 16% using deep learning.

This made everyone sit up and take notice. Massive research began in deep learning, and just a few years later in 2015 computers actually beat humans with an error rate of just 4%.

Today, just 5 years after Hinton’s breakthrough, the error rate for AI is 3%.

The portion of evolution in which animals developed eyes was a big development. Now computers have eyes.

– Jeff Dean, Google Brain

Governments, the academic world, big corporations, and startups became obsessed with neural networks and AI in general.

Tech companies began spending billions to make progress as fast as possible and a massive recruiting war began for anybody with relevant skills.

As everybody raced to develop these deep neural networks breakthrough after breakthrough occurred and it was clear that something very special was happening.

  • Self-Driving cars became a reality
  • Computers became as good as or better than humans at many types of medical diagnosis
  • Voice recognition accuracy skyrocketed
  • High accuracy language translation (for text and audio) became accessible
  • Autonomous drones were built
  • A Computer beat humans at Go, a milestone thought to be decades away.

Today the pace of funding and development of AI is faster than ever and we are still in the very early stages of exploring what these neural nets and AI systems can help us do.


The Next Decade (2017 – 2020s)

The Dream Is Finally Arriving. This Is What It Was All Leading Up To.

– Bill Gates




Better optimized logistics, supply chain management, back office processes, and communication will speed up many of our daily processes.

Productivity gains and reduced friction will occur across all industries.

These behind the scenes advances might be incremental and seemingly boring, but when combined and compounded will free up so much wealth and resources.


A lot of experts are going to be surprised at how inefficient their best efforts have been.

Google Turns on DeepMind AI, Cuts Cooling Energy Bill by 40%

We used a similar system to AlphaGo, but instead of playing Go we applied it to the cooling systems in the data centers to try and increase their energy efficiency.

We managed to save 40% of the energy that was used by the cooling systems.

The whole data center now has 15% less power usage. That’s worth tens of millions of dollars per year.

What we’re thinking now is, why don’t we optimize something like the energy grid at national scale? There’s no need to just think about data centers, there must be huge inefficiencies even at grid scale.

– Demis Hassibis, Google DeepMind


This alone probably pays for the Deepmind acquisition. Shows how far below Pareto optimal limits even Google was.

Balaji S. Srinivasan


I don’t think most grasp the significance of this. Oil companies have similar systems they pay billions trying to optimize.

– iandanforth


It’s all about optimization. It can be used in supply logistics, shipping logistics and dynamic pricing in addition to keeping an industrial area at the right temperature. We’ll be seeing AI being applied to a lot more areas.

– Dave Schubmehl


Honestly, I’m skeptical a generalized AI will go fully conscious in my lifetime. But these specialized AI? These things are going to start changing our lives over the next ten years in unimaginable ways. The energy savings alone is incredible.

– tendimensions





Long term Industrial commodity price decline

AI is very effective at figuring out where resources might be and where to drill. It can also optimize business processes, design better equipment, and suggest better techniques.

This will drive the oil price down even further, making it cheaper for all of us to power our businesses and lives.

Travel will become cheaper. Airline ticket prices will fall.

As more efficient production methods are discovered productivity will rise, pushing costs down and enabling prices to fall.

This will occur across a range of commodities from food to energy.





The event horizon of a coming economic singularity where all prices drop down an asymptote toward zero as technology advances exponentially. – James C. Townsend

Optimization and cheaper commodity prices will produce tremendous deflationary pressures.

Productivity gains should be higher than the rate of money printing. This means that prices will fall.

As things get easier to produce they become less scarce. Lower costs for producers means they can cut prices to gain more customers and sell to a larger market.

We’ve already seen this with electronics such as smartphones.

The components of the phones keep reducing in price, and supply chains and operations get more efficient, allowing the companies to expand their market to billions of people instead of a much smaller market of wealthy people.

Hundreds of years ago, light was very expensive. Clean water was very expensive. As those things became easier to produce and deliver to people (electricity and modern plumbing systems) the price dropped down an asymptote towards zero.

Nowadays those things are so cheap and easy to obtain that most people take them completely for granted.

The internet revolution has already delivered this in many ways. Entertainment is practically free now with youtube videos, file sharing, and the ability to access any song, movie, book or picture instantly.

Every industry will have costs lowered by AI bringing more efficient systems and production methods.

The path we are on is to use our knowledge and tools to drive costs down and make systems and processes more efficient.

This will result in things that were previously scarce becoming abundant. Deflation is the reward we get for improving our methods.





Fortunes will be made acquiring analogue manufacturers and digitizing them. Comparable to Russian privatization.

– Pierre Rochard

The global manufacturing sector generates about $12 trillion in annual revenue.

Industrial robotics equipped with better than human image recognition will use data from sensors to optimize how they function.

Imagine the cost savings with even just small incremental productivity gains.





This is already well underway.

Transport costs will continue to be driven down as ride sharing companies offer self-driving rides.

Car ownership rates will plummet. Vehicles will be massively more utilized (at the moment they’re idle 95% of the time). City streets will open up to be used in different ways.

It will save millions of lives and free up an uncountable number of hours.

I never thought I’d see autonomous automobiles driving on the freeways.

It wasn’t many years ago [they] put out a request to see who could build a car that could go across the Mojave Desert to Las Vegas from a place in Southern California, and several engineering teams across the country set out to do this. Nobody got more than about 300 yards before there was a problem.

Two years later, they made the full 25-mile trip across this desert track, which I thought was a huge achievement, and from that it was just a blink before they were driving on the freeways.”

– Gordon Moore





Language, even more than religion, is humanity’s central point of division.

Finally technology is overcoming this.

– Jeffrey Tucker

I used to believe that it would take a couple of decades to get a babel fish type of device. I now believe we will have languages solved within 10 years at most.

Google Translate is amazing. The recent conversion to an AI system resulted in huge improvements. I’m in awe at how accurate it can instantly translate foreign text.

Google also have a new feature to use your phone’s camera to translate text in the world around you (like on signs or on a menu).

Skype translator converts instant messages with a very low error rate. This has been available for a while and now audio translation is being rolled out.

I recently tried French – English which was flawed but good enough if you talk clearly. It’s obviously going to improve exponentially with more training.

Breaking these language barriers will reduce friction and speed up processes all over the world. It will allow much greater integration of the 7 billion minds on the planet.




Personal assistants can finally understand you and start to do useful things.

That’s a watershed moment.

– David Hanson, Jan 2017


Speech Recognition for under $1 in every product. That’s coming within 2 years.

Imagine what we see today with Amazon Alexa, that can be in every Roomba, every little consumer product, and it’s cheaper than putting buttons on a product. That’s obvious.

– Steve Jurvetson, 2016

Personal AI assistants similar to what is portrayed in the movie Her will be increasingly prominent on our desktops and smartphones (and devices beyond that like virtual and augmented reality glasses and lenses).

Siri – Apple

Viv, Bixby – Samsung

Google Now – Google

Amazon Echo – Amazon

Cortana – Microsoft

M – Facebook

Personal assistants, along with household robots, will be one of the first things that triggers a realization among the general public that a major transition is happening.




I think AI’s effect on healthcare will be far more pervasive and far quicker than anyone anticipates. 

Even today, AI/Machine Learning is being used in oncology to identify optimal treatment patterns.

– Stephen Gold, IBM Watson

Medicine doctor pushing on first aid sign with modern computer interface


AI is already diagnosing people better than humans and saving lives.

We think that its no longer necessary for humans to spend time reviewing text reports to determine if cancer is present or not.

We have come to the point in time that technology can handle this. A human’s time is better spent helping other humans by providing them with better clinical care.

Everything — physician practices, health care systems, health information exchanges, insurers, as well as public health departments — are awash in oceans of data. How can we hope to make sense of this deluge of data? Humans can’t do it — but computers can.”

This is a major infrastructure advance — we have the technology, we have the data, we have the software from which we saw accurate, rapid review of vast amounts of data without human oversight or supervision.

– Shaun Grannis



Deep learning is helping us to make sense of the human genome and know what to do with it. It’s just too complicated for us to make fast progress without AI.


Life Extension

Ben Goertzel of Hanson Robotics and OpenCog says it’s inevitable that we will learn enough through AI systems to be able to indefinitely extend our lifespan.




The greatest achievement of our technology may well be the creation of tools that allow us to go beyond engineering – that allow us to create more than we can understand.

– Danny Hillis, 1998

An area I find particularly interesting is evolutionary/iterative algorithms. Deep Learning is part of a larger family of these type of algorithms (where the AI system does something over and over again in an evolutionary manner and somehow get the best result).

These designs often end up looking completely alien and counter-intuitive.

Example 1 – Satellite Antennas


NASA has no idea why their satellite antennas are best shaped like this.


Example 2 = Jet Engines


A jet engine is far too complicated for an unaided human to design so General Electric use evolutionary algorithms.


Example 3 – Structural Nodes


3 structural nodes to hold cables.

On the left is a human design, the right is a machine learning design.

The AI design results in a 40% overall weight reduction of the total structure.


Example 4 – Producing Entangled Photons


Melvin, an algorithm designed at the University of Vienna, works by taking the building blocks of a quantum experiment (lasers and mirrors) and the quantum state desired as an outcome and running through different setups at random.

If the random setup results in the desired outcome, Melvin will simplify it. It can also learn from experience, remembering which configurations result in which outcomes, so it can use those and build on them as needed.

So far, the team says, it has devised experiments that humans were unlikely to have conceived. Some that work in ways that are difficult to understand. They look very different from human-devised experiments.

“I still find it quite difficult to understand intuitively what exactly is going on,” said team member Mario Krenn



AlphaGo is a similar example of how AI makes decisions that are completely different to what a human would come up with.

This was best displayed in the legendary game 37 when AlphaGo made a move that humans had unanimously dismissed as stupid for 3,000 year, yet turned out to be brilliant.

After humanity spent thousands of years improving our tactics, computers tell us that humans are completely wrong. I would go as far as to say not a single human has touched the edge of the truth of Go.

– Ke Jie, Go World Champion

Colonizing mars, mining asteroids, and building a space economy in the decades ahead will need much better engineering than we’ve so far come up with so far, but these algorithms will help us design it.

At some point AI may take us over the threshold of knowledge required to build some of the other most important paradigm shifting technologies speculated about such as large-scale quantum computers or Drexlerian Molecular Assemblers.





Deep neural nets will process the big data that we don’t know what to do with. This will provide insights in fields such as physics, biology, and chemistry.

Research breakthroughs at CERN’s particle collider or with the James Webb Space telescope project would give us a better ontological understanding of the nature of reality.





World GDP in Trillions $

We are already in the Singularity.

At this scale we can see it began sometime around the Scientific Revolution in the 1600s and really took off in the industrial revolution of the 1800s.

What those eras brought will pale in comparison to what comes next.





These companies trade on the stock market so are available for the public to become owners of them.

I’ve listed them in order of market cap so it’s easier to get some perspective over who are the most valued companies and who has more room to grow relative to their competitors.


1) Apple

Silicon Valley, USA

Market Cap = $711 Billion

  • Personal assistant – Siri
  • Self-driving cars
  • Massive proprietary data to feed into AI training sets and improve systems


2) Google

Silicon Valley, USA

Market Cap = $582 Billion

  • DeepMind
  • AlphaGo
  • RankBrain – AI Search
  • Personal assistant – Google Now
  • Optimization Services (DeepMind in energy)
  • Google Cloud Computing (Google Cloud Platform)
  • Hardware –TPUs new deep learning chips (Tensor Processing Chips)
  • DeepMind Wavenet –The best text to speech system in the world (50% better than traditional systems)
  • Owns D-Wave “quantum computers” for quantum machine learning
  • The Google Quantum AI Lab (Quantum Machine Learning – Hartmut Neven, John Martinis)
  • Massive proprietary data to feed into AI training sets and improve systems
  • Self-Driving Cars – Waymo
  • Robotics – Boston Dynamics
  • Google Brain Project
  • Google Translate
  • Google Inbox Smart Reply
  • Google Home (Competitor to Amazon Alexa)
  • Nest (Home automation)
  • Wearables – Smart contact lens (long-term project)
  • Languages: “In few years time we will put it on a chip that fits into someone’s ear and have an English-decoding chip that’s just like a real Babel fish.” – Geoffrey Hinton
  • Key Talent – Geoffrey Hinton, Jeff Dean, Ray Kurzweil, John Martinis, Demis Hassibis, Mustafa Suleyman (DeepMind)


3) Microsoft

Seattle, USA

Market Cap = $499 Billion

  • Languages – Skype Translator
  • AI Personal Assistant – Cortana
  • Search – Bing
  • Speech Recognition
  • Cloud computing provider (Microsoft Azure)
  • Researching deep learning with FPGAs. Possibly better result than GPUs
  • Researching quantum computers
  • Massive proprietary data to feed into AI training sets and improve systems
  • Key Talent – Jian Sun, Li Deng, Matthias Troyer.


4) Amazon

Seattle, USA

Market Cap = $404 Billion

  • Cloud computing provider (AWS)
  • Amazon AI Services
  • Open sourced their Deep Scalable Sparse Tensor Network Engine (DSSTNE)
  • Personal Assistant – Amazon Alexa
  • Autonomous Drones
  • Investing significantly in robotics to improve their processes, reduce costs and increase profits.
  • Massive amounts of difficult to obtain proprietary data.


5) Facebook

Silicon Valley, USA

Market Cap = $387 Billion

  • Big data – Training AI on it.
  • AI Personal Assistant – M
  • Oculus – Integrating AI into this. (Long term)
  • Acquired voice-recognition AI startup Wit.ai
  • Key Talent – Yann LeCun


6) General Electric

Fairfield, USA

Market Cap = $270 Billion

  • Using AI to improve engineering


7) Samsung

Seoul, South Korea

Market Cap = $259 Billion

  • Acquired much hyped AI assistant Viv Labs (Siri cofounders new company)
  • The next Samsung phone will come with a digital assistant called Bixby.
  • Wearables – Smart contact lens (long term)
  • Part of a $3 billion public/private research centre with, LG, telecom giant KT, SK Telecom, Hyundai Motor, and internet portal Naver.
  • Pumping Millions into Artificial Intelligence startups
  • Deep learning to detect cancer – Samsung Medison


8) Alibaba

Hangzhou, China

Market Cap = $253 Billion

  • Cloud Computing AI Platform


9) Intel

Silicon Valley, USA

Market Cap = $175 Billion

  • Nervana Systems is a deep learning chip set company recently bought by Intel for around $400 mill which accelerates algorithms.
  • Intel FPGAs Break Record for Deep Learning Facial Recognition


10) Oracle

Silicon Valley, USA

Market Cap = $172 Billion

  • Lots of data
  • AI Cloud Apps


11) IBM

New York, USA

Market Cap = $171 Billion

  • Cloud computing provider (IBM Cloud)
  • IBM Watson – a tool for doctors, business people, and scientists.
  • IBM Watson making inroads into the $3.8 trillion healthcare industry
  • IBM Watson -Diagnosis greatly in demand
  • FPGA – Neuromorphic Chips


12) SAP

Walldorf, Germany

Market Cap = $106 Billion

  • Access to lots of data
  • Cloud computing platform
  • Sells AI services
  • SAP believes the next technology adoption phase for businesses will be around how they can use intelligent applications to assist them with their operations.


13) Qualcomm

San Diego, USA

Market Cap = $84 Billion

  • Competes with Nvidia in the GPU market
  • Neuromorphic Chips
  • The Internet of Things


14) Softbank

Tokyo, Japan

Market Cap = $82 Billion

$32 billion acquisition of ARM (makes GPUs)

  • Robotics
  • Pepper, customer service robot
  • “Nao,” a customer service robot that answers basic questions and is designed to speak 19 languages
  • Masayoshi Son (Chairman) – “ARM will play a key role in bringing about advanced artificial intelligence. I have unfinished business with the Singularity.”


15) Daimler AG

Stuttgart, Germany

Market Cap = $72 Billion

  • Mercedes-Benz, one of the leaders in self-driving cars
  • Mercedes and Uber plan network of self-driving cars


16) Baidu

Beijing, China

Market Cap = $65 Billion

  • Baidu AI Lab
  • AI Cloud Services
  • 2nd largest search engine in the world.
  • 4th most visited website in the world (behind Google, Youtube, and Facebook)
  • Baidu’s AI Supercomputer Beats Google at Image Recognition
  • Key Talent – Andrew Ng


17) Nvidia

Silicon Valley, USA

Market Cap = $56 Billion

  • Nvidia have 70% of the GPU market and will try to defend that.
  • They have a huge head start over their competitors by investing in a $2 billion R&D program before anyone else.


18) Salesforce

Silicon Valley, USA

Market Cap = $56 Billion

  • Salesforce Einstein is artificial intelligence (AI) built into the core of the Salesforce Platform
  • Bought deep learning startup Metamind


19) General Motors

Detroit, USA

Market Cap = $54 Billion

  • Self-Driving Cars


20) ABB Group

Market Cap = $50 Billion

Zurich, Switzerland

  • Industrial robots and robot software.


21) Ford

Detroit, USA

Market Cap = $50 Billion

  • Putting $1 billion into an AI startup, Detroit’s biggest investment yet in self-driving car tech


22) Hon Hai Precision Industry Co Ltd

Taipei, Taiwan

Market Cap = $50 Billion

  • Apple’s Supplier
  • Industrial Robotics
  • Already had a fully automated factory that can run 24 hours a day with the lights off


23) Keyence Corp

Osaka, Japan

Market Cap = $48 Billion

  • Develops and manufactures automation sensors, vision systems, barcode readers, laser markers, measuring instruments


24) Tesla Inc

Silicon Valley, USA

Market Cap = $42 Billion

  • Self-Driving cars
  • Plans be able to add your self-driving car to the Tesla shared fleet just by tapping a button on the Tesla phone app and have it generate income for you while you’re at work or on vacation.


25) Fanuc Corp


Market Cap = $41 Billion

  • One of the largest makers of industrial robots in the world


26) Hyundai Motor Co

Seoul, South Korea

Market Cap = $32 Billion

  • Part of a $3 billion public/private AI research centre.
  • Self-Driving Cars.


27) Mitsubishi Electric Corp

Tokyo, Japan

Market Cap = $32 Billion

  • Self-Driving Cars


28) HP

Silicon Valley, USA

Market Cap = $28 Billion

  • Venture Fund that invests in AI


29) Nidec Corp

Kyoto, Japan

Market Cap = $27 Billion

  • Manufactures electric motors


30) Intuitive Surgical

Silicon Valley, USA

Market Cap = $27 Billion

  • The Da Vinci, a flagship robot performs surgical operations


31) Illumnia Inc

San Diego, USA

Market Cap = $24 Billion

  • Applying Deep learning to healthcare


32) Naver Corp

Seoul, South Korea

Market Cap = $22 Billion

  • Korea’s biggest search engine
  • Lots of data


33) Delphi Automotive

Gillingham, England

Market Cap = $20 Billion

  • Built 1st driverless car to travel across the USA


34) SMC Corp

Tokyo, Japan

Market Cap = $19 Billion

  • Industrial robotics


35) SK Telecom

Seoul, South Korea

Market Cap = $16 Billion

  • Part of a $3 billion public/private AI research centre.


36) Tencent

Shenzhen, China

Market Cap = $15 Billion

  • Tencent AI Lab


37) Twitter

Silicon Valley, USA

Market Cap = $14 Billion

  • Lots of data


38) AMD

Silicon Valley, USA

Market Cap = $12 Billion

  • Competes with Nvidia in the GPU industry.
  • Investing heavily in AI markets


39) Mobileye

Jerusalem, Israel

Market Cap = $10 Billion

  • Self-Driving Cars


40) LG Electronics

Seoul, South Korea

Market Cap = $9 Billion

  • Part of a $3 billion public/private research centre with, Samsung, telecom giant KT, SK Telecom, Hyundai Motor, and internet portal Naver.


41) Omron Corp

Kyoto, Japan

Market Cap = $9 Billion

  • Japanese electronics company.
  • Bought Adept technologies which focuses on industrial automation and robotics in 2015


42) Toshiba Corp

Tokyo, Japan

Market Cap = $9 Billion

  • Robotics
  • Toshiba receives bid as high as $3.6 billion for chip business stake
  • Toshiba gets on the starting blocks for its latest NAND fab


43) Trimble Navigation

Silicon Valley, USA

Market Cap = $8 Billion

  • Makes Global Positioning System receivers, laser rangefinders, unmanned aerial vehicles, inertial navigation systems and a variety of software processing tools


44) KT Corp

Seoul, South Korea

Market Cap = $7 Billion

  • Part of a $3 billion public/private research centre with, Samsung, LG, SK Telecom, Hyundai Motor, and internet portal Naver.


45) NEC

Tokyo, Japan

Market Cap = $6 Billion

  • Internet of Things


46) Nuance Communications

Boston, USA

Market Cap = $5 Billion

  • Provides speech and imaging applications


47) Yaskawa Electric Corp

Kitakyushu, Japan

Market Cap = $5 Billion

  • Industrial robots


48) Cyberdyne Inc

Tsukuba, Japan

Market Cap = $3 Billion

  • Robotics company
  • Introducing ‘service’ robots with artificial intelligence


49) iRobot

Boston, USA

Market Cap = $1.7 Billion

  • The Roomba, the floor cleaning robot from my previous company, iRobot, is perhaps the robot with the most volition and intention of any robots out there in the world. Most others are working in completely repetitive environments, or have a human operator providing the second by second volition for what they should do next – Rodney A. Brooks, 2014


50) Pacific Industrial Co

Nagoya, Japan

Market Cap = $0.7 Billion

  • Robotics


51) Nippon Cermaic Co

Tottori, Japan

Market Cap = $0.5 Billion

  • Robotics
  • Sells various types of sensors


52) Kawada Technologies Inc

Tokyo, Japan

Market Cap = $0.4 Billion

  • Industrial Robotics
  • Its robot is designed to work alongside humans, and can be taught new tasks without the need for programming expertise.




At the moment the public cannot buy stock in these companies.

Some might IPO soon or be bought up by one of the large public companies.


1) Uber

  • Acquired self-driving truck company Otto.
  • Uber AI Research Labs
  • AI Labs at Uber, which is using AI in everything from self-driving cars to dynamic ride scheduling
  • Transitioning to self-driving ride shares
  • Massive amounts of proprietary data.


2) Human Longevity Inc

  • Creating the largest database of human genotypic phenotypic and microbiology data ever assembled and using machine learning to analyze it


3) Hanson Robotics

  • Create life-like robots.
  • These robots are infused with OpenCog’s opensource artificial intelligence to “think”.


4) Rethink Robotics

  • Sells “Baxter” Industrial robot.


5) Kernel

  • Building advanced neural interfaces to treat disease and extend cognition.


6) Vicarious

  • Working on artificial intelligence; replicating the human visual cortex and creating machines with human-level intelligence in vision, language and motor control.
  • Funded by Mark Zuckerberg, Elon Musk, Peter Thiel, and Jeff Bezos


7) Berg Pharmaceutical

  • Uses machine learning, to learn the various health associations and correlations. This led to the development of Berg’s first drug, BPM 31510, which is in clinical trials.




It’s probably smart to be an owner of Artificial Intelligence.

The large public companies are spending billions.

Private companies and startups getting involved.

Lots of theoretical research is taking place in academia (especially in Canada).

Governments and militaries are inevitably involved. Some of these projects are public knowledge, for example the Chinese government have pledged to invest $15 billion by 2018 and South Korea has a $3 billion project. Most likely there are classified projects already well underway.

There are several opensource AI projects such as OpenCog and OpenAI. Maybe some mindblowing unforeseeable opportunity to invest in AI will emerge from a project similar to these the same way Bitcoin came out of nowhere and changed everything.

Some random kid in the middle China could come up with something amazing. History shows that no matter how impressive a big company’s projects look, it’s nothing compared to the guys working out of their garage.

At this stage though, the easiest way to become an owner is to buy shares on the stock market of the companies most prominently involved and likely to succeed.



I laugh when people say tech is a bubble. The establishment is the bubble.

Who’s around in 2025 – Google or the EU?

Balaji S. Srinivasan

Of the public companies heavily involved, the one that stands out head and shoulders above the others is Google.

Google snaps up every machine-learning or robotics company it likes the look of. They have supposedly the greatest Artificial Intelligence Lab in the world. They even managed to convince the neural network godfather Geoffrey Hinton to join them.

Google never pays a dividend or does stock buybacks because they aggressively reinvest cash into long-term projects such as DeepMind who they bought for $400mill (a bargain in hindsight). Founder Demis Hassibis describes it as “the Manhattan Project of AI” with hundreds of the best minds working on it.

The scope of their focus is breathtaking and they are clearly the front-runners to make major breakthroughs or even develop General AI.

Kevin Kelly tells the story below of when he realized Google have knowingly been building an AI from day one:

Around 2002 I attended a small party for Google—before its IPO, when it only focused on search. I struck up a conversation with Larry Page, Google’s brilliant co-founder, who became the company’s CEO in 2011. “Larry, I still don’t get it. There are so many search companies. Web search, for free? Where does that get you?”

My unimaginative blindness is solid evidence that predicting is hard, especially about the future, but in my defense this was before Google had ramped up its ad-auction scheme to generate real income, long before YouTube or any other major acquisitions.

I was not the only avid user of its search site who thought it would not last long. But Page’s reply has always stuck with me: “Oh, we’re really making an AI.”

I’ve thought a lot about that conversation over the past few years as Google has bought 14 AI and robotics companies.

At first glance, you might think that Google is beefing up its AI portfolio to improve its search capabilities, since search contributes 80 percent of its revenue. But I think that’s backward. Rather than use AI to make its search better, Google is using search to make its AI better. Every time you type a query, click on a search-generated link, or create a link on the web, you are training the Google AI.

When you type “Easter Bunny” into the image search bar and then click on the most Easter Bunny-looking image, you are teaching the AI what an Easter bunny looks like. Each of the 12.1 billion queries that Google’s 1.2 billion searchers conduct each day tutor the deep-learning AI over and over again.

With another 10 years of steady improvements to its AI algorithms, plus a thousand-fold more data and 100 times more computing resources, Google will have an unrivaled AI. My prediction: By 2024, Google’s main product will not be search but AI




Nvidia are also a very exciting company with massive demand for their products.

The shares are worth picking up because if they do hold on to their position as the best chips to use for AI over the next few years their sales and opportunities to expand into other interesting areas are going to be staggering.

If We Were a Hedge Fund We’d Put All Our Money Into Nvidia

– Marc Andreessen


For somebody who wants general exposure to AI, but doesn’t want to pick a winner, investing across a basket of shares by building a portfolio of the best stocks is a good idea.

The Global X Robotics & Artificial Intelligence Thematic ETF is an actively managed AI fund with a management fee of 0.68%. It trades under the stock symbol “BOTZ”.

The 10 biggest positions in this portfolio are:

  1. Mitsubishi Electic Corp
  2. Fanuc Corp
  3. ABB
  4. Keyence Corp
  5. SMC Corp
  6. Intuitive Surgical
  7. Yaskawa Electric Corp
  8. Omron Corp
  9. Trimble Navigation Ltd
  10. Mobileye.


Personally, I think the 10 stocks below would make a better AI fund.

  1. Google (GOOGL:NASDAQ) – $846 per share – $582 billion marketcap
  2. Nvidia (NVDA:NASDAQ) – $107 – $56 bill
  3. Tesla (TSLA:NASDAQ) – $272 – $42 bill
  4. IBM (IBM:US) – $181 – $171 bill
  5. Microsoft (MSFT:US) – $65 – $499 bill
  6. Samsung (005930:KS) – 1.9 million KRW – $259 bill
  7. Amazon (AMZN:NASDAQ) – $845 – $404 bill
  8. AMD (AMD:US) – $13 – $12 bill
  9. Facebook (FB:NASDAQ) $134 – $387 bill
  10. Baidu (BIDU:NASDAQ) – $185 – $65 bill




Short WTI Crude Oil (Currently $53 per barrel)

Short the Bloomberg Commodity Index (Currently at 88)

Short the First Trust NASDAQ Global Auto Index Fund (Currently at 36)




The hard part was getting the neural nets to work. This took decades but now we’re there we will see years of easy gains as different industries and people roll them out into every product and service they can think of.

Is there a risk of another AI winter and we’ll be disappointed again?

It’s unlikely at this stage. Computer Science just overtook Economics as the most popular class at Harvard and more and more people are going into machine learning, so better talent is going to be coming through over the next few years.

Companies are spending billions of dollars, and there’s so much low hanging fruit just from applying the current deep learning techniques we already have.


Key Area 1) Data

The amount of data is growing at an exponential rate, doubling every two years.

While a lot of data remains proprietary (which gives companies like Google their edge) there are more and more public data sets becoming available for anyone to use and train their neural nets on.

With more sensors, the internet of things, and more companies willing to launch services which actually lose money just to get data for training AI, this area is going to keep growing exponentially.


Key Area 2) Computing Power and Infrastructure

  • We’ll soon see the power of computing increase way more than any other period. There will be trillions of products with tiny neural networks inside. – Jen-Hsun Huang, CEO of NVIDIA

Hardware is something we have a very good track record on. It’s extremely unlikely that this will be a limiting factor in the decades ahead.

Ray Kurzweil says there’s general agreement that we’re close to the hardware requirements of strong AI. He believes that we’ll be at the human brain of 10^14 calculations per second for $1,000 in the early 2020s.

Hardware costs are coming down as competition increases and chip companies are racing to compete with Nvida in the GPU area.

Every company is thinking about how to run AI faster to enable larger neural nets. Similar to the bitcoin hardware race, there’s a range of hardware, from GPUs to FPGAs to ASICS.

Deep learning ASICs are a very interesting area with some huge engineering efforts going on in the space of neuromorphic chips, Google’s TPUs, and quantum computers.



Microsoft has done research to show you can do better than GPUs with these.


Tensor Processing Units (TPUs)

Google are building their own machine learning ASICS known as Tensor Processing Units (TPUs).


Quantum Deep Learning

I would predict that in 10 years there’s nothing but quantum machine learning–you don’t do the conventional way anymore

– Google’s Hartmut Neven


We actually think quantum machine learning may provide the most creative problem-solving process under the known laws of physics

– Google

The Google Quantum Artificial Intelligence Lab was set up to investigate how quantum computing might help with machine learning

Hartmut Neven leads the team and is known for developing the first image recognition system based on quantum algorithms using a D-Wave quantum computer.

Google have bought every model D-Wave has ever produced (at around $15mill each) and has signed a contract to buy all of their machines for the next 5 years.

Lockheed Martin also bought a D-Wave quantum computer to help train a deep neural network.

Google just keeps saying breathless things about D-Wave. I don’t know why Google’s competitors don’t wake up and say “God, what if Google’s right?… What if this is actually gonna knock the socks off everything else.. Maybe we’d want to buy one.

Strangely it’s only Google who’s actually buying them, at least to date in this category. Microsoft or Apple haven’t bought one, but we’ll see.

– Steve Jurvetson, 2016

Google’s Quantum AI Lab also has John Martinis who is regarded as one of the leading experts.

Martinis is part of Google’s effort to build a quantum computer and his qubits are widely regarded as way higher quality than D-Waves.

Building a quantum computer is a massively ambitious goal, although progress appears to be going much better than most expected. Martinis is confident that Google can demonstrate “quantum supremacy” within the next 1-3 years.

They are definitely the world leaders now, there is no doubt about it. It’s Google’s to lose. If Google’s not the group that does it, then something has gone wrong.

– Simon Devitt, 2016


Other Efforts in Quantum Deep Learning

IBM and Microsoft are also investing massive resources to develop their own quantum computers that can be applied to AI.

In 2014 a team of Chinese physicists demonstrated ‘quantum artificial intelligence” by training a quantum computer to recognize handwritten characters.

There are still massive challenges, but even if quantum machine learning turns out to to be too difficult to scale and fully implement over the next couple of decades it’s inevitable we’re going to get exponentially better hardware for machine learning which will allow much deeper neural nets.


Key Area 3) Better Platforms and Algorithms

This is an unprecedented time in openness. All of the big companies are openly publishing their work, although they keep their data-sets proprietary and build better systems than their competitors by exploiting their data edge.

It’s easier than ever to build machine learning systems, especially with opensource platforms like OpenAI’s Universe and Google’s Tensorflow. Start ups can use these to create an AI product that solves a business problem.

As more money and talent comes into the space, better ways of building AI will inevitably emerge.

Most exciting is AI systems working on better ways to build AI systems.

This is already happening.

1) AI Designing deep neural nets

2) Machine learning system writes machine learning software.



Building Stronger More General AI

The big moment is when AIs become general in their abilities and have an almost human-like level ability to generally figure things out, to switch from context to context and remember skills applied in different areas.

General AI is hard, and we’re not there yet.

Most of the magic produced by AI today comes from a surprisingly simple technique called supervised learning.

Most researchers believe the next big breakthrough will come from unsupervised learning.

Unsupervised learning involves learning from unlabeled data. We’ll probably need a lot more progress in unsupervised learning to get to General AI.

In the brain, synapses adjust themselves but we don’t have a clear picture for what the algorithm of the cortex is. We know the ultimate answer is unsupervised learning, but we don’t have the answer yet.

– Yann LeCun, Facebook AI Research


In 2013, hundreds of experts were asked when they thought AGI may arrive. The median prediction was 2040.

In “Future Progress in Artificial Intelligence: A Poll Among Experts” by Bostrom and Vincent C. Müller, the authors come to this conclusion: AI systems will probably (over 50%) reach overall human ability by 2040-50, and very likely (with 90% probability) by 2075.

As AI matches the range of tasks a human can do and switch between them, improvements will take it far beyond a current humans capabilities in similar situations.

How long would it take for a general AI to go from unenhanced human level abilities to superintelligence?

Nick Bostrom has a great slide showing how long it took AlphaGo to progress from a beginner human level of Go to crushing one of the world’s best players.

Image result for lee sedol vs alphago

Lee Sedol v AlphaGO

October 2015: “Based on its level seen in the match (against Fan), I think I will win the game by a near landslide”

February 2016: “I have heard that Google DeepMind’s AI is surprisingly strong and getting stronger, but I am confident that I can win at least this time”

March 9th 2016: “I was very surprised because I didn’t think I would lose”

March 10th 2016: “I’m quite speechless… I am in shock. I can admit that… the third game is not going to be easy for me”

Mar 12th 2016: “I kind of felt powerless.”



A general AI doesn’t have to be self-aware, or conscious. There’s a big debate over what consciousness even means. An AI could have general abilities even if the lights aren’t on.

It is a huge leap to go from something as unalive as a microwave or iphone to something which has the ability to be self-aware and have subjective experiences.

In a report to the Pentagon, JASON claim that current neural network architecture is completely unrelated to consciousness and not even close to the right path to replicating human like self-awareness.

Andrew Ng of Baidu agrees and says “there is no clear path to how AI can become sentient. Part of me hopes there will be a technological breakthrough that enables AI to become sentient, but I just don’t see it happening. That breakthrough might happen in decades, hundreds, or thousands of year from now. I really don’t know.”

We know creating consciousness is possible because we have it, so there must be a way, but how to do it we still don’t know.

David Deutsche feels that current models are a dead end to creating consciousness but “it is plausible that just a single idea stands between us and the breakthrough. But it will have to be one of the best ideas ever.”




Smartphones and the internet gives us so much opportunity for more knowledge, more intelligence.

I’m really looking forward to AI giving us more of that in 10 years or so when everyone is wearing augmented reality glasses with deep learning built into it. Then beyond that somehow integrating AI into my brain processes.

– David Chalmers

A common theme is us against the AI but I think this is about us and the AI.

We are not facing an invasion of intelligent machines from Alpha Centauri. These are our tools, an extension of us. Our devices will get smaller and more embedded into us through wearables, VR/AR glasses, VR/AR contact lenses.

Once nonbiological intelligence gets a foothold in the human brain (this has already started with computerized neural implants),  the machine intelligence in our brains will grow exponentially

– Ray Kurzweil

Obviously there are massive dangers involved, but that’s where this is leading. It will be a very incremental process. Bit by bit the technology will become more and more a physical part of us rather than an extension.

This seems ridiculously science fiction, but breakthroughs are becoming common in neuroprosthetics and implantable electronics.

Can we inject electronic circuits into the brain, then connect & monitor it?

Yes,we can, & that’s where we are today


Charles Lieber’s team at Harvard is making good progress by implanting electronics in mice. Lieber commented by saying “In science, I’ve been disappointed at times, and this is a case where we’ve been more than pleasantly surprised,”

Elon Musk has spoken at length how a “neural lace” is essential. Neuroprosthetic startup Kernel is employing top neuroscientists to build an implantable chip in the hope that eventually it will be able to enhance intelligence.





The road ahead might seem outrageous, or too dangerous and difficult from where we sit today.

But in all likelihood, when the time passes and we get to that stage, many will take the accomplishments for granted. The goalposts always move on what seems impressive.

An often repeated quote is “AI will be the last invention we ever make”. But what seems to us today like the hardest challenge possible might seem relatively trivial compared to what we will face after this.

Who knows what kind of journeys, adventures, and problems we’ll be enabled to overcome when there is no distinction between us and our most advanced technology.


My Writing Profile Picture

Follow me on Twitter @leebanfield1

Bitcoin: 1Jwh6nZiASJf4d3hNytjxqiimWBmEJvJ4S

Bitmessage: BM-2cXjeAykLT7gbjzNHZFnCxdawvyryyb4Nf




  1. […] – Exploring Opportunities in Artificial Intelligence, Lee Banfield […]

  2. […] – Exploring Opportunities in Artificial Intelligence – Lee Banfield, February 19th 2017 […]

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: