AI. The new VC goldrush.

Since 2012, things have been changing in AI country. Founders and investors are equally stampeding to take part in the AI excitement. Yet, in spite of this unbridled optimism, the industry still has a long way to go as the hype precedes real business cases and proofpoints. Let’s explore what’s happening in a bit more detail.

Deep Learning resuscitates the AI industry

While the AI industry witnessed two harsh winters, a new renaissance is in full swing. As of 2007, the craze-du-jour is called Deep Learning, and “My startup uses deep learning for…” is a commonly heard phrase at many investor pitches. While deep learning is the latest excitement in AI, it is in essence a rebranding of neural networks (or “ANN”), but this time multiple ANNs are layered. And while deep learning definitely adds value in extracting and learning higher level features of data, it is not a novel scientific breakthrough that fuels it popularity.

Why the AI industry gained a lot of momentum is not that there have been new radical advancements in the field of AI, rather than the computer industry has been able to provide some critical ingredients to make things finally work; an abundance of data, and the costs of information processing and storage that continue to collapse.

The pervasive computing opportunity for AI

Perhaps one of the most unique ingredients that continues to push the AI industry, is the emergence of pervasive computing principles. With the explosion of mobile and the Internet of Things (or “IOT”), the computing industry shifted from entering data into computers, to wearable devices that continuously observe us and learn.

This shift requires new techniques that are able to interpret this continuous flow of observational data and turn this into ambient intelligence. Achieving this next level of automation will provide companies a better understanding of the likings and preferences of users.

Putting AI to work on this observational data is considered the holy grail for many companies, as obtaining a deeper understanding of consumer behavior is a competitive advantage that will mean the survival or demise of a company in the near future.

AI is cowboy country

The recent AI dawn is earmarked by lofty promises and high expectations, often fueled by persuasive marketing. And as with every promising new technology, many companies that are drawn into the excitement of this craze may find themselves struggling to find a profitable business.

Right now, most AI solutions and platform are like swiss-army knives, attracting founders that try to apply AI to every computing problem they can think of.

While it is true that AI has many applications, it is not yet clear what the high-value applications will be. What we are actually experiencing is an unprecedented winner-takes-all race that continues to match up ambitious founders with investors scrambling in all directions.

Both sides of the table seem to be equally afraid to miss out on this ‘next big thing’, but seem to have no solid roadmap to profit or an industry solution.

Dazed and confused, an AI shake-out is inevitable

Through all the hype, founders and investors seemingly forget that the true benchmark of success is not to build an algorithm that is capable of learning how to play Space Invaders, but how to turn AI into real-world solutions.

You know, good old fashioned solutions, where customers pay money to solve an actual problem.

Following a crazy wedding party with a 302% funding jump and big seed and series A rounds, this honeymoon period will eventually cool down and investors will be sobering up. This state of affairs will put many start-ups under an immense load of pressure to show that their technologies are more than impressive parlour tricks or an academic novelty.

GAFA is after AI

To make the challenge even more daunting for founders, companies such as Google, Apple, Facebook, Amazon (recently coined as the GAFA companies), but also other global companies such as Microsoft, Baidu and IBM have one resource in abundance; data at scale.

Knowing that an AI is only as effective as the data it has access to, this means a seemingly unfair and uphill battle for anyone lacking access to large amounts of data. As these incumbent companies have access to zetabytes of data, even the low hanging fruits in machine learning already provide huge advantages to their business. While these companies remain secretive about the their internal progress, they invest heavily in AI, and continue to beat the drum to which the industry marches forward.

Frequently, the incumbent companies release exciting results in their research to show they are still ahead of the curve. Facebook’s DeepFace achieves a 97.25% level of accuracy in face recognition, exceeding human-like capabilities. Google built AI algorithms to learn how to play Atari games, and IBM’s Watson beat all human contestants on Jeopardy. Microsoft is raising eyebrows with impressive results on Project Adam, and even new entrants and startups, such as Vicarious, boast their results on taking on CAPTCHA’s.

Clearly, impressive results. But truth of the matter is, the pressure on these global companies is so high, some of them, such as Baidu, are even willing to falsify results so they can claim their own bold statements.

A testament of this race, is how Google released results on algorithms that incept art following one day later by Facebook publishing their results.

This means that raking in one impressive results over another is the game these companies are playing, in order to both justify crazy investments as well as boast their pole position in the AI race. And as with any high stakes race, cutting corners is permissible.

Acquihires as a business model?

An important element of winning the AI race is being able to attract, retain and bond the best of class in the AI industry. Facebook’s FAIR team is lead by Yann LeCun, while Google’s AI race is lead by  Geoff Hinton, Demis Hassabis and Ray Kurzweil. Amazon has persuaded Carlos Guestrin, while Baidu has managed to get Andrew Ng on board.

Not only are these unicorn AI visionaries important to lead internal research and development efforts, they also act as a beacon of trust to many academics and data scientists vying to work with and learn from these industry legends.

Having AI heavyweights on board is a critical element. Indeed the war on AI talent is excruciating. And with over $650m in acquisitions (DeepMind, Deep Blue Labs, JetPac, and Vision Factory) in 2014 in the AI field, Google makes a clear statement it seeks to dominate this industry.

With acquihire valuations that exceed 5x the regular acquihire value (take DeepMind, with the equivalent of $5m-$10m per technical lead), AI startups seem to attract both founders and investors that keep a lucrative acquihire to be a valid option as a short- to mid-term end game.

As the hype precedes credible business models and proofpoints, this may very well be the best bet for many fledgling AI startups so far. While prepping a company for an acquisition might be a red flag in most cases, we cannot ignore the crazy transactions happening in this space, as the war on talent continues to heat up.

Of course, a special breed of investors and founders will be required to ride the AI wave. I’m talking about the kind that technically understands how AI will change the face of technology and has access to the financial resources, grit and unshakeable vision to support their venture.

Risky bets, but no bubble

This AI promise has lead to massive early stage investments, such as Sentient with a total funding of $143,8m, Context Relevant with $44,3m, or Vicarious with $72m in the bank.

Clearly, things have changed since 2012. AI investments have jumped 302% and more than 170 start-ups made a bet on the AI bandwagon, with over $307 invested in over 40 new deals in 2014.


While my post aims to be a critical note about the current state of affairs, I remain extremely optimist for the AI industry. Surely, after a crazy party we all need to sober up a bit, and we will, but I’m not sure we can call this a bubble.

Never in the history of computer science, we have had the explosive growth of mobile, wearable and connected devices that generate massive amounts of observational data.

Never we have had social networks generating petabytes of images and text on a daily basis, nor did we ever had such massively scalable storage and processing solutions before. All elements that prevented earlier AI excitement to mature into industry solutions.

Much to the likes of previous historical AI excitement in 1950s and 1980s, the Obama administration even made a $100m bet on the advancement of hyper intelligence. Daily progress in the field makes it apparent that AI is no longer a neat trick.

A bubble? Hardly. Overly enthusiast? Definitely.

But it seems that AI may finally be reaching a first level of maturity and start producing very exciting and practical solutions over the next couple of years. What these solutions will be, we don’t know yet…