Artificial Intelligence Doesn’t Appear Ready to Take Over the World Yet

Source: Forbes. AI Spending To Exceed A Quarter Trillion Next Year. Beth Kindig. Nov 14, 2024.

This rapid buildup of positive sentiment has been exciting for us at Beaumont Capital Management because we’ve been working in the machine learning space for over a decade. In the past, fewer people were versed in the language of AI and machine learning. It may have seemed uncertain or scary to outsource portfolio management to a computer system (albeit one overseen and implemented by human investors). Now that the average person has had much more experience interacting with AI and machine learning in their everyday life, some of the trepidation has gone away and financial advisors we talk to seem to have a more sophisticated understanding of the strengths and weaknesses of Artificial Intelligence as they incorporate those tools into their practices.[3]

However, as we near the end of 2024, researchers, businesses, and investors have begun to throw some cold water on the overheated sentiment. Some of the biggest tech companies that raced into the business are reporting low growth expectations and cautioning that AI remains a net loss for the time being.[4] While AI capital expenditures are projected to remain high,[5] some major companies have recently been punished by investors for excess AI spending and so we could see a downshift in AI-related CapEx in the coming quarters.[6]

The truth is, as “magical” as their output may at times appear, as currently deployed, these AI products are useful tools but not the new paradigm it may have initially seemed they would represent. (Perhaps they do represent a new paradigm for fraud, but that’s another matter). [7] In my view, this is not the automobile or the television or the iPhone or even the computer operating system yet.

In fact, some of the major new AI players themselves suggest their models have stopped improving.[8] In order to continue to learn, a machine learning or Artificial Intelligence system needs quality, fresh data. The challenge these AI models are confronting is a lack of both. They have largely taken in all publicly available data they can get their hands on. To get fresh data, they will likely need to pay to license it. One work around they have tested is to use AI generated data in their models, however, this has been shown to result in “model collapse” when used on a wide scale. Even more challenging, the amount of information available on the internet that is generated by AI has increased so that some ever-increasing proportion of the fresh data they can obtain without licensing is itself generated by AI. Companies will require intense human oversight of their model training results in order to correct for the potential collapse caused by training AI on AI-generated data. [9]

Many AI proponents insist that they can achieve continued learning with more computing resources. I am skeptical. As someone who builds machine learning models for a living and whose livelihood (and the life savings of our customers) depends on accurate predictions, I can say with experience that there are times when one confronts the limits of a model’s architecture. By a model’s architecture, I mean the underlying design or algorithm that drives the machine learning process. When I reach the limits of a certain approach, efforts to improve output such as more data, more variables, different sampling design, or transformations of the data all fail to increase the predictive performance of a model by any meaningful amount. That’s when I either have to set aside a hypothesis I was testing as failed or try a new modeling approach, for example, instead of using a logistic regression maybe I switch to a decision tree or a neural network.

I personally believe we may be near the limits of what may be accomplished with the large language models like ChatGPT (and others). Closer study of the outputs of these models, and knowledge of how they actually function under the hood leads me to believe that Open AI and others have not built a potential generalized intelligence, but rather just very good machine guessers. It’s amazing how well some of these systems can replicate or mimic sentient language skills. However, my wager is that a more generalized machine intelligence will require a different sort of architecture all together.

For that reason, I’m not particularly bullish or bearish on anything related to AI at the moment as an investment opportunity. It’s possible the “picks and shovel” companies may be in line for a correction if CapEx among the big players slows due to investor pressure or if the AI products rolled out continue to fail to contribute meaningfully to their bottom lines.

To avoid coming off as too pessimistic, I do believe these tools have improved the search experience and they have shown to be useful at automating low-level tasks while at the same time requiring close oversight to avoid “hallucinations” and non-sequiturs. And so, after this initial hype cycle dies down, I personally predict we’ll see less wild-eyed discussion of what’s going to happen to the office worker when AI takes over the world.

For more insights like this, visit and subscribe to blog.investbcm.com.

By Andrew Rice, partner and portfolio manager

For more news, information, and strategy, visit the ETF Strategist Channel.

[1] Tech Giants Are Set to Spend $200 Billion This Year Chasing AI

[2] Tech Giants Are Set to Spend $200 Billion This Year Chasing AI

[3] Morningstar. Behavioral Research Insights. “What Do Financial Advisors Think about Generative AI in Financial Advisors’ Workflow?”. September 2024. AI for Financial Advisors | Morningstar

[4] https://www.bloomberg.com/news/articles/2024-10-30/meta-sales-narrowly-beat-on-ai-boosting-advertising-revenue

[5] AI will force tech investors to become macro aware | Reuters

[6] https://www.bloomberg.com/news/articles/2024-10-30/microsoft-cloud-fuels-stronger-than-expected-revenue-growth

[7] FBI Warns of Increasing Threat of Cyber Criminals Utilizing Artificial Intelligence — FBI

[8] OpenAI, Google and Anthropic Are Struggling to Build More Advanced AI - Bloomberg

[9] https://www.nytimes.com/interactive/2024/08/26/upshot/ai-synthetic-data.html (footnotes 6 and 7 both cover all of the content in this paragraph)

Disclosures:

Copyright © 2024 Beaumont Capital Management LLC. All rights reserved. All materials appearing in this commentary are protected by copyright as a collective work or compilation under U.S. copyright laws and are the property of Beaumont Capital Management. You may not copy, reproduce, publish, use, create derivative works, transmit, sell or in any way exploit any content, in whole or in part, in this commentary without express permission from Beaumont Capital Management.

Certain information contained herein constitutes “forward-looking statements,” which can be identified by the use of forward-looking terminology such as “may,” “will,” “should,” “expect,” “anticipate,” “project,” “estimate,” “intend,” “continue,” or “believe,” or the negatives thereof or other variations thereon or comparable terminology. Due to various risks and uncertainties, actual events, results or actual performance may differ materially from those reflected or contemplated in such forward-looking statements. Nothing contained herein may be relied upon as a guarantee, promise, assurance or a representation as to the future.

This material is provided for informational purposes only and does not in any sense constitute a solicitation or offer for the purchase or sale of a specific security or other investment options, nor does it constitute investment advice for any person. The material may contain forward or backward-looking statements regarding intent, beliefs regarding current or past expectations. The views expressed are also subject to change based on market and other conditions. The information presented in this report is based on data obtained from third party sources. Although it is believed to be accurate, no representation or warranty is made as to its accuracy or completeness.

The charts and infographics contained in this blog are typically based on data obtained from third parties and are believed to be accurate. The commentary included is the opinion of the author and subject to change at any time. Any reference to specific securities or investments are for illustrative purposes only and are not intended as investment advice nor are they a recommendation to take any action. Individual securities mentioned may be held in client accounts. Past performance is no guarantee of future results.

As with all investments, there are associated inherent risks including loss of principal. Stock markets, especially foreign markets, are volatile and can decline significantly in response to adverse issuer, political, regulatory, market, or economic developments. Sector and factor investments concentrate in a particular industry or investment attribute, and the investments’ performance could depend heavily on the performance of that industry or attribute and be more volatile than the performance of less concentrated investment options and the market as a whole. Securities of companies with smaller market capitalizations tend to be more volatile and less liquid than larger company stocks. Foreign markets, particularly emerging markets, can be more volatile than U.S. markets due to increased political, regulatory, social or economic uncertainties. Fixed Income investments have exposure to credit, interest rate, market, and inflation risk. Diversification does not ensure a profit or guarantee against a loss.

The Decathlon strategies utilize artificial intelligence (AI) in the decision-making process, introducing inherent risks. The AI’s lack of predictability, reliance on historical data, and sensitivity to market volatility may impact investment outcomes. Technology-related risks and the dynamic nature of market conditions further contribute to potential uncertainties. Ongoing monitoring and adjustments to the AI model are essential. Investors should recognize the limitations of AI, seek professional advice, and carefully assess their risk tolerance and financial situation before making investment decisions.

---

A message from Advisor Perspectives and VettaFi: To learn more about this and other topics, check out some of our webcasts.

More Risk Management Topics >