2020 U.S. Equity Index Price Returns

Dow: +6.0% | Nasdaq: +42.0% | S&P: +15.3%

Catch Up Quick

• The Trump administration declassified unconfirmed intelligence that China offered to pay non-state actors in Afghanistan to attack U.S. soldiers

• Amazon ($AMZN) announced an acquisition of podcast network Wondery

• > 6 in 10 Americans are hopeful about 2021 (Axios / SurveyMonkey Poll)

• Senator Mitch McConnell refuses to budge on approving $2K stimulus checks

• 42 people in West Virginia mistakenly received a Regeneron IV coronavirus treatment instead of the vaccine shot

• The EU reached a major investment deal with China despite US concerns with respect to forced labor, setting up tension with the Biden administration

• Jay Y. Lee, Samsung Electronics’ Vice Chairman, faces a nine year sentence in wake of bribery case

• Elon Musk says SpaceX will attempt to recover its Super Heavy rocket by catching it with a launch tower

• NSO, a company known for its offering of Pegasus spyware for governments, used people’s real-time location data to pitch its contract-tracing solution

• The price of a single Bitcoin is just ~$1K away from the $30K mark

• A newer, more contagious coronavirus variant has been detected in 2 U.S. states (though experts believe the strain will be covered by current vaccines)

Thought of the Day

• Alongside potentially anticompetitive practices and superfluous market power, facial recognition is another aspect of the technology sector that has recently garnered momentous criticism

• Taking a brief step back, the specific technical elements of facial recognition typically involve highly complex algorithms embedded in structures called neural network (a model often referred to as machine learning, or deep learning if many algorithms within the network)

• However, given the prevalence of this technology in our every day lives, from autocompleting a sentence in an email to recommending videos on YouTube, it can be highly beneficial to understand it at a high level

• Machine learning (“ML”) is a subset of AI enabling a system to receive a set of data and learn from it to achieve a certain output

• In fact, many ML components are structurally inspired by the biology of the human brain (or that of a mammal)

• The human brain is wired to recognize patterns and categorize / classify information; for instance, a toddler may touch a hot stove and thus learn via a slight pain influx to stay away from it in the future

• Neural networks use nested hierarchies of related concepts, decision trees, and feedback mechanisms (other than physical pain) to do the same for computer systems

• With facial recognition, developers feed (or “train”) an AI module with thousands of pictures of various faces

• The technology essentially develops a flexible database of faces, along with copious facial data points of all kinds

• Once fully trained, the system can receive an image of a face, parse through every pixel of that image, and cross reference the facial data points (and other algorithmic tests) with every picture of a face its ever “seen”

• If the input meets a minimum threshold of similarity to a face it has seen before, the module will recognize it

• Back to the point of criticism, facial recognition is commonly used for solving crime, in which these algorithms repeatedly exude far less accuracy on people of color

  Recently, Nijeer Parks, a 33-year old Black man from New Jersey was wrongfully arrested due to law enforcement’s use of facial recognition. This is the third person known to the falsely arrested based on an incorrect facial recognition match. In all three cases, the people mistakenly identified have been Black men (NY Times)

• Essentially, these modules are only accurate if the data they are trained with is impartial / unbiased, given it uses this data to make highly educated inferences

• Many presume that the higher accuracy rate for White or Asian faces is a result of training bias induced by ML engineers and developers (who many claim are predominantly White or Asian themselves)

The Bottom Lines

• Facial recognition is one of many examples in which a technology has outpaced legal infrastructure to the point in which it has become a societal issue

• While it certainly can reduce friction for various actions such as unlocking a phone in a highly secure fashion, it has its drawbacks, as discussed

• It’s longer term prevalence is certainly up in the air; however, the underlying ML technology behind it will undoubtedly change the world as we know it over the next couple decades!

  Leave a comment