The Emerging AI Bubble

Many readers might remember the “dot-com bubble” (also known as the tech bubble, and the Internet bubble ). It was an historic economic bubble and period of excessive speculation that occurred roughly from 1997 to 2001. It was a period of extreme growth in the usage and adaptation of the Internet by businesses and households. During that period, many Internet-based companies, commonly referred to as dot-coms, were founded and many of them failed. Sadly we’re likely to enter a new technology “bubble” related to Artificial Intelligence (AI).

Even the generally conservative, CNN Money, is now publishing articles touting “Robot revolution: 22 stocks to buy”, which is an item that should be corrected immediately. Robots do not mean AI and AI does not mean Robots. They may be combined but they are mutually exclusive technologies. The Accounting firm Deloitte, proclaiming it’s “Cognitive Advantage”, says “Emerging technologies are finally able to emulate and augment the power of the human brain.”

Goldman Sachs is advising clients that AI is defined as “any intelligence exhibited by machines or software.” That can mean machines that learn and improve their operations over time, or that make sense of huge amounts of disparate data. It’s even using phrases similar to those used during the dot-com bubble like; “who are the players going to be?”

Goldman considers Sentient Technologies to be a “player’. It claims Sentient is the company seeking to solve the world’s most complex problems using a phrase that is a clear throwback to the dot.com era, “‘massively scaled’ artificial intelligence”.

Just like during the dot-com bubble Goldman is sharing it’s advice in ways that distinguish it. Goldman “believes” Japanese hardware company NEC— the number one facial and text analysis company in the world — is a good investment. They also recommend several companies that sell AI components into cars for scenarios like helping drivers park: Nidec, MobileEye, Nippon Ceramic, and Pacific Industrial. So once again the “advisors” are more than ready with their analyses and investment reports.

66 years ago in 1948 Claude Shannon introduced the idea that information was a measurable element and defined the basic unit, which would later be called a “bit.” Almost seven decades since Shannon introduced us to bits work is underway to accelerate the evolution of human intelligence. The outputs of these endeavors can be both entertaining and informative. Generally, however, the issue is financing and getting people to pay for the work which more often than not causes focus on entertaining rather than informing. This also means AI work becomes more focused on applying it and less on “developing” or advancing it.

Within ten years of Shannon introducing the concept of a bit, John McCarthy introduced the term Artificial Intelligence (AI) in 1956. It’s been on a bumpy development ride ever since but AI technology has made significant advances in recent years with statistical AI and accelerated all the recent hype and economic bubbles likely to surround it’s financing.

AI researchers first thought they would use “logic-based” AI but that became intractable so they switched to “statistical” AI. Regardless of the technique used to produce intelligence, logic-based or statistical, the scope of AI is categorized as:

  1. Weak artificial intelligence (weak AI), also known as narrow AI, is non-sentient intelligence focused on a narrow task. Apple’s Siri is a good example.
  2. Strong or general artificial intelligence which is a machine with the ability to apply intelligence to any problem.

All currently existing systems considered “AI” are weak AI at best. It’s these weak AI systems that are most prominent and generating the energy related to an AI bubble.

According to MIT professor Max Tegmark most controversies surrounding strong artificial intelligence (that can match humans on any cognitive task) center around two questions:

  • When (if ever) will it happen, and
  • Will it be a good thing for humanity?

He then classifies the views of researchers on these two questions into five classes:

  1. Techno-skeptics — are convinced that human-level artificial general intelligence (AGI) won’t happen in the foreseeable future
  2. Digital utopians think it will happen but is virtually guaranteed to be a good thing.
  3. Beneficial-AI movement feels that concern is warranted and useful, because AI-safety research and discussion now increases the chances of a good outcome.
  4. Luddites are convinced of a bad outcome and oppose AI.
  5. Virtually Nobody Thinks strong AI will happen in the next few years

Perhaps the AI-enabled application likely to take-off and generate some of the most energy related to an AI bubble is“ Facial Recognition Systems” (FRS) Facial recognition proceeds in three steps: detection, faceprint creation, and verification or identification.

Starting with an image — generally from an unknown person, FRS decompose the image into data segments called a faceprint which is a set of characteristics that, taken together, uniquely identify one person. They then attempt to match the face print to data segments from facial images in a database of known people’s faces. Elements of a faceprint include the relative locations of facial features, like eyes, eyebrows and nose shape.

Providing a lot of energy for the AI bubble Google claims it’s FaceNet, Facial recognition application represents the most-accurate approach yet to recognizing human faces. According to Google, it’s FaceNet achieved nearly 100-percent accuracy on a popular facial-recognition dataset called “Labeled Faces in the Wild”, which includes more than 13,000 pictures of faces from across the web.

Never one to be outdone in the arena of hype, Facebook said it’s Facial Recognition App can identify individuals with 83% accuracy using a method dubbed PIPER, an acronym for pose invariant person recognition. According to the Wall Street Journal “the technology could help Facebook develop more products akin to its newly launched photo-sharing app Moments, which uses face recognition to group images based on who is in each photo.”

“Face ID is the future of how we unlock our smartphones and protect our sensitive information” said Phil Schiller, Apple’s senior vice president of worldwide marketing.

iPhone Facial Recognition

A “dot projector” beams out more than 30,000 invisible infrared dots, and the infrared camera captures an image. Apple uses the infrared image and dot pattern and pushes them through neural networks to create a mathematical model of your face, and then it checks that mathematical model against a stored image captured earlier. Once it detects a match, the phone unlocks.

iPhone X will come with an A11 bionic neural engine to process faces. “Bionic Neural Engine”! Sounds pretty “bubbly” doesn't it?

Though Google, Facebook and Apple advances in facial recognition are relatively new, computer systems like them can be found all around us today. As sophisticated as Facial Recognition Systems sound they are still examples of “Weak AI”. They are not the type of General AI used by Humans. FRS incorporate an artificial intelligence technique called deep learning, which has proven remarkably effective at so-called machine perception (also a hype phrase) tasks such as recognizing objects (by some metrics, machines are now better at this than are people), recognizing voices, and understanding the content of written text.

Facezam recently engaged in what appears to be a promotional event highly reminiscent of the dot.com bubble. Jack Kenyon, the founder of Zacozo Creative (the ad agency behind the campaign) told the Observer newspaper in an email interview that he and his coworkers were playing with the music recognition app in a London pub when they got the idea for Zacozo’s first public campaign. “We believed a hoax facial recognition app would quickly go viral because of its controversial nature, the invasion of privacy and dubious public uses,” Kenyon said. Kind of like hyping the hype or bubbling the bubble!

___________________________________________________________________

William Smith is a software engineer with forty years of experience developing applications for global businesses in the consumer products industry along with large government agencies. He holds advanced degrees in industrial engineering and international relations.