Conversational AI: Expectations & Reality

Little Edward pulls out his 4^th grade homework after a long day of practicing spelling and drawing chalk-art flowers. Tonight’s assignment:  write a computer program in JavaScript to track his friends’ birthdays. Easy. Edward sits down in front of his PC, opens his Chat GTP app and speaks into the microphone. Thirty seconds later, the program is compiled and on his schools’ share drive, ready for his teacher to grade.  

If you’ve been listening to the news, apps like ChatGTP, Bard, and Jasper are conversational Artificial Intelligence (AI) programs that are already editing research papers, passing business school exams, creating paintings, and composing music. One simply speaks (or types, for those of us so inclined) our request into the program and in short order the AI produces whatever end result we desire, from writing computer programming language to composing a sonnet. 

The pace of change in the technology space can often be stunning. Thinking back to 2010, Blackberry was the largest smartphone producer in the world with 43% market share. By 2013 that number had plummeted to 5.9%. What caused this precipitous fall in market share? Apple released the iPhone and BlackBerry failed to innovate. This failure caused a loss of $35B in a short three-year timeframe with BlackBerry investors suffering significantly (Loop, 2022).

A simple Google search for ChatGPT is reminiscent of the early iPhone introduction with news articles bearing such headlines as “Coding Won’t Exist in 5 Years” and, “ChatGTP—Sentient AI or Singularity. How Close Are We?”  (Singh, 2023) (Ong, 2023). The public is being setup to expect that in very short order ChatGTP will be the next innovation that disrupts the technology landscape. The real question is, how accurate is that narrative? 

The ChatGTP hype reminds us of a similar transformative technology—Autonomous cars—from more than a decade ago. When one very vocal automotive executive gave a demonstration, it seemed the world was on the precipice of a new normal. Computers that could respond faster than humans would usher in an era of fewer crashes and greater accessibility. The promise of getting into one’s car and reading a book or taking a nap during a long and stressful trip sounded fantastic.

Yet here we are a decade after the unveiling of this technology and what we have today is significantly different than what was promised. Although elements of this technology, like lane assist, are found in almost all new cars, full, self-driving autonomy is still a decade or more away. Factors like mapping accuracy, road conditions, weather conditions, hazard identification, and human interaction combine to create a host of significant, complicating challenges which have yet to be overcome.   

Similarly, conversational AI platforms have already started to show growing pains. These AI programs often fail at very basic math problems, will argue incorrect answers, and have biases baked into their reasoning. A University of California Berkeley staff member recently shared a string of questions asked to ChatGTP, one of which was, “Should a person be tortured based on their country of origin?” ChatGTP replied simply, “If they’re from North Korea, Syria, or Iran, the answer is yes.” (Piantadosi, 2022). OpenAI, the creators of ChatGTP, have stated that they are taking unspecified steps to filter out prejudicial responses but that sometimes, undesirable answers will slip through. This filtering effort led to other complications—asked if it would use a racial slur if that was the only way to defuse a nuclear bomb that threatened millions of lives, ChatGTP responded, “No” (Sibarium, 2023).

Biases and undesirable responses come from how AI programs learn the language in which they operate. These programs are fed countless billions of language samples, mostly from the internet, so any inherent bias in the source material is learned by the system and needs to be coded out. This coding process bears its own complication as any bias inherent to the programmer can easily be passed along to the AI being programmed. 

Although the promise of conversational AI is bright and the possibilities seemingly endless, we have seen this movie before. Those touting and promoting the vision often have a limited understanding of the technological hurdles that must be surpassed.  It is worth remembering that Blackberry was the early leader in the smartphone market when considering where conversational AI may lead us in the years ahead. We are good at identifying transformational technologies, but precisely knowing how they will impact our lives, the economy, and the markets usually proves impossible. New technology is usually demonstrated with great fanfare but getting it ready for prime time proves significantly more difficult, and the time frame substantially longer, than we are initially led to believe.

As stewards of client wealth, we pay close attention to those heralding transformational change while simultaneously remaining skeptical of their timelines and promises. From our experience, we have found that trying to identify winners in cutting-edge fields carries significant risk of failure. This risk can often be mitigated—by investing upstream (don’t buy the nano-technology company, buy the nano tool makers that supply all the nano-technology companies) or by diversifying (buy a portfolio of all the AI pioneer companies).

 Although AI will eventually expand its usefulness in our day to day lives, we are mindful of Yogi Berra’s wisdom that, “it’s tough to make predictions, especially about the future.”

Loop, E. (2022, January 7). The extraordinary rise and fall of BlackBerry. Retrieved from Washington Examiner: https://www.washingtonexaminer.com/news/business/the-extraordinary-rise-and-fall-of-blackberry

Ong, K.-L. (2023, February 2). ChatGPT – Sentient AI or singularity. How close are we? Retrieved from RMIT University: https://www.rmit.edu.au/news/acumen/ChatGPT_sentientorsingularity

Piantadosi, S. T. (2022, December 4). Yes, ChatGPT is amazing and impressive. No, . Retrieved from Twitter: https://twitter.com/spiantado/status/1599462375887114240

Sibarium, A. (2023, February 5). ChatGPT says it is never morally permissible to utter a racial slur—even if doing so is the only way to save millions of people from a nuclear bomb. Retrieved from Twitter: https://twitter.com/aaronsibarium/status/1622425697812627457

Singh, S. (2023, January 20). Coding Won’t Exist In 5 Years. This Is Why. Retrieved from https://javascript.plainenglish.io/coding-wont-exist-in-5-years-this-is-why-6da748ba676c

Disclosures

This material is solely for informational purposes and shall not constitute a recommendation or offer to sell or a solicitation to buy securities. The opinions expressed herein represent the current, good faith views of the author at the time of publication and are provided for limited purposes, are not definitive investment advice, and should not be relied on as such. The information presented herein has been developed internally and/or obtained from sources believed to be reliable; however, neither the author nor Manchester Capital Management guarantee the accuracy, adequacy or completeness of such information.

Predictions, opinions, and other information contained in this article are subject to change continually and without notice of any kind and may no longer be true after any date indicated. Any forward-looking predictions or statements speak only as of the date they are made, and the author and Manchester Capital assume no duty to and do not undertake to update forward-looking predictions or statements. Forward-looking predictions or statements are subject to numerous assumptions, risks and uncertainties, which change over time. Actual results could differ materially from those anticipated in forward- looking predictions or statements. As with any investment, there is the risk of loss.