AI and the Long View

?Interesting article in NY Times on Meta’s AI glasses .? Quick recap of the idea: you look at something through the glasses and ask your glasses “What am I looking at”.?? The bulk of the article is an exploration of how accurate [or not] the tech is when identifying an object.? Cool but this type of technology already exists via Google Search where you can take a picture of something and ask Google what it is.? These Meta glasses simply cut out the need to take a picture.?

?What came to mind as I read the article, and the reason for this post, are two issues that I don’t feel have generally been well covered by the mainstream media as it relates to today’s AI craze.? A craze that was kicked off by announcements and innovations related to generative AI.

?Classic AI vs Generative AI

?The first issue is the idea that AI just burst onto the scene from nowhere. It is true that there has been an explosion of innovation, VC investment and AI whitewashing due to significant and recent advancements in generative AI.? However, the discipline of AI has been around for decades.?

?A quick scan of the literature on AI shows that neural networks first appeared as a topic of exploration in the late 1940’s? and had morphed into a discernable discipline by a decade later.? I personally remember working on expert systems when I was in college in the 80s.? Rule based expert systems were one of the earliest manifestations of attempts to create artificially intelligent systems.?

?The world saw a huge boost in AI capabilities in the 90’s as increases in computing power intersected with increased maturity in the area of deep learning.? This potent combination allowed computers to really begin to exploit pattern recognition capabilities that most closely simulate how our brains solve problems.

?People learn by bringing together past experiences and outcomes in order to predict future outcomes.? Computers have the benefit of being able to rapidly crunch through massive amounts of data to recognize patterns and then predict a near infinite number of outcomes with probabilities of success computed for every outcome.?

?Computers make true a great saying I heard a long time ago: “life is too short to make all of the mistakes yourself, so you need to be able to learn from the mistakes of others.? Improve decision making by Increasing the amount of data available to you so you can generate better predictions of possible outcomes.? Classic AI is all about this idea.? Leverage massive amounts of data to recognize patterns and predict likely outcomes.

?The Impact of LLMs

?The use of large language models (LLMs) by AI really adds another dimension to AI and is the foundation of generative AI. The buzz about generative AI is that it enables the ability to create novel solutions.? Novel content creation is a really exciting aspect of today’s ability of AI to work with LLMs.?? What’s as least as powerful, and maybe more so, is the idea that for the first time, mere mortals can begin to directly leverage the raw power of AI in their daily activities.?

?Until recently only data scientists and programmers could interact directly with the models being created to drive pattern recognition.? Using generative AI built on LLMs, non-data scientist using natural language, can now generate new, original content—everything from text, images, video, music, and computer code.?

?The many potential downsides of AI use in our day to day lives has gotten plenty of play in the press.? Everything from wild hallucinations where fiction is substituted for fact, unseen biases in AI algorithms and the question of whether AI algorithms can be trusted to act ethically in any number of situations.? Those are all great issues and should be top of mind for anyone charged with figuring out if and how AI evolution will be regulated.

?That brings me to my second issue.? AI, especially generative AI, will have a profound impact on many professions that rely on knowledge and creativity to create novel content.?

?The Unpredictability of Long-term Benefits and Risks

?It seems funny that a technology that is all about predictive capabilities will leave society facing so many questions about how this technology will change the way we work and by extension the way we live.? Two quotes from a McKinsey Study on the impact of Generative AI illustrate this point.?

?“Current generative AI and other technologies have the potential to automate work activities that absorb 60 to 70 percent of employees’ time today.”

?“…generative AI has more impact on knowledge work associated with occupations that have higher wages and educational requirements than on other types of work.”

?In the future, once generative AI reaches a significantly higher level of maturity, there will be the need for less people across many professions.? We’ve already seen the increased potential to further displace low to middle wage workers with the rise of autonomous vehicles.? Broadscale use of driverless taxis, buses and even long-haul freight delivery probably aren’t that far off.?

?But the impact to professions where people use words or imagery to earn a living will also be greatly impacted.?? Professions that include software developers, lawyers, journalist, and marketers to name a few.? The need for people that are experts in these domains will not disappear, but the number of individuals needed in each of these professions may dramatically decrease as the productivity of each individual worker dramatically increases.

?Getting ahead of the power curve

?The impact to society caused by the broadscale adoption of AI technologies is hard to predict with any kind of accuracy.? However, based on the impact of significant past technological advancements we can predict the following:

?The impacts will be both positive and negative.? AI is likely to improve productivity across a broad number of domains and professions.? However, it likely will also mean that we need way less people to do certain jobs.? This in turn may or may not be a good outcome based on two other offsetting long-term trends: (a) people are living longer while at the same time (b) the global birth rate continues to decline.?

?What is more certain is that there will be dislocations to large segments of society as the work of many individuals undergoes significant transformation.?? There will undoubtedly be winners and losers as these transformations take place.? There always are.

?In the past, societies haven’t really been able to get ahead of the power curve as it relates to these kinds of major technological and economic upheavals.? Look no further than the example of what happened to manufacturing in the US in the latter half of the last century due to a combination of the application of automation and the shifting of manufacturing to lower cost geographies.

?Let’s Not Wait

?Massive transformation of how we do work will not occur overnight.? It is also important to recognize that generative AI is simply a continuation of a long-term trend to replace labor with automation that goes back to creation of the first machines.? Generative AI is a step function, an accelerator if you will, that drives us closer to an infinity point that we chase but will never reach.?

?We shouldn’t wait until the disruptions of Generative AI become deep seated and painful to begin to lay out plans for what will surely be a dramatic change to how people think about work.? These changes are in our future; even if we don’t know the exact shape or timing of things to come.