From Bubbles to Breakthroughs: GenAI as a Revolutionary Language Amplifier
The AI market shows signs of frothiness reminiscent of past technological bubbles. Goldman Sachs’ (GS) recent report, “Gen AI: Too Much Spend, Too Little Benefit?” has sparked debate about whether the massive investments in AI will pay off. This post, the first in a multi-part series, examines the current AI market dynamics, historical parallels, and the transformative potential of Generative AI.
Evidence of Frothiness:
Nvidia, the company at the heart of the GenAI revolution, provides a striking example of market volatility:
- Despite reporting blockbuster earnings with July quarter revenue exceeding $30 billion (up 122% year-on-year), Nvidia’s shares fell 3.8% in premarket trading. (Link)
- The company saw its share price fall 9% during the day and then 1.5% in overnight trading, with CNBC citing economic concerns and a DoJ antitrust investigation.(Link)
Are these market reactions suggesting that Goldman is right? Are AI-related stocks overvalued?
From Tulips to Tech: Are AI Investments Following a Familiar Bubble Pattern?
While it’s challenging to predict with certainty, historical patterns suggest a correction in the AI market may be on the horizon. Goldman’s cautionary stance on GenAI seems well-founded, as there are compelling reasons to believe the technology may be overhyped and corrections are likely:
1. Investor disillusionment: If AI fails to deliver expected returns quickly, investors may lose faith and withdraw capital, leading to a market correction.
2. Normalization: As AI technologies mature and speculation subsides, valuations may align more closely with fundamental business metrics. This often occurs as revolutionary technology becomes commonplace, integrated into everyday life and thus less exciting from an investment standpoint.
These factors, reminiscent of past tech bubbles, suggest that while AI holds immense potential, the current market enthusiasm may be outpacing realistic near-term expectations.
Why is it inevitable? Because we have seen this movie before.
Investment bubbles are a recurring phenomenon in capitalist economies. Examples include Tulipomania (1634), the South Sea Bubble (1720), Railway Mania (1845), the Radio Bubble (1927), and the Dot-Com Bubble (1995–2000). These episodes demonstrate how euphoria around novel technologies can drive rapid asset price inflation, followed by an inevitable crash. The challenge is that no one knows when to exit at the right time.
During the Dot-Com crash, companies like Pets.com became emblematic of speculative excess, going from IPO to liquidation in less than a year and leaving many individual investors with substantial losses. Similarly, in previous bubbles like the South Sea Bubble, even brilliant minds like Isaac Newton were not immune to significant losses. Newton and many small investors suffered substantial financial setbacks when the bubble burst, and prices collapsed. This pattern underscores the risks of speculative investing, particularly for those without the resources to weather severe market downturns.
As we witness the surge in AI-related investments, it’s worth considering whether we’re seeing history repeat itself and how society and the economy at large might prepare for and adapt to potential disruptions.
The Risk of Underestimating GenAI:
While signs of a bubble exist, we must also be cautious about underestimating GenAI’s potential. The Kodak case serves as a cautionary tale. In 1975, Kodak engineer Steve Sasson invented the digital camera. Even at that time, its transformative potential was visible:
- Instant image capture and review: Users could immediately see if everyone had their eyes open in a photo, eliminating the need to wait for film development.
- Cost efficiency: Digital photography eliminated the need for film purchases, developing costs, and physical photo albums.
- Improved accessibility: Storing and accessing digital photos proved easier than managing their analog counterparts.
These benefits signalled a paradigm shift was coming — even if not immediately realizable.
The full impact of digital photography took over three decades to unfold. It took the convergence of enabling technologies like high-capacity storage, broadband internet, and social media for digital photography to become mainstream, ultimately leading to Kodak’s bankruptcy in 2012. This underscores the importance of recognizing long-term transformative potential, even when immediate applications seem limited.
Differing Perspectives on GenAI’s Impact:
Goldman Sachs’ report presents perspectives both for and against GenAI’s potential economic impact. However, it’s essential to recognize that these optimistic or skeptical viewpoints are primarily based on broad economic estimates and generalized projections. MIT Professor Daron Acemoglu’s conservative estimate of GenAI’s impact on productivity and GDP and GS’s Jim Covello’s skepticism about AI’s problem-solving capabilities contrast with GS analyst Eric Sheridan’s more optimistic comparison to previous tech cycles.
Yet, this macro-level analysis risks overlooking the transformative potential already evident at the task level. Much like how the revolutionary nature of digital photography was initially underestimated, focusing solely on broad economic indicators may cause us to miss the true impact of GenAI on specific industries and processes.
The Hidden Impact of GenAI in the Workplace:
While economists debate the broad economic impact of GenAI, its effects are already being felt at the individual task level in workplaces worldwide. A recent BBC report highlights a striking statistic: in a February 2023 study by professional social network Fishbowl, 68% of 5,067 respondents who used AI at work said they don’t disclose this usage to their bosses.
The BBC article tells the story of Matt, a worker in a fast-paced environment, who found that tasks such as researching scientific topics, gathering sources, and producing thorough presentations for clients “suddenly became a breeze” with the use of ChatGPT. The catch? Matt and his colleague had to keep their use of the AI tool a closely guarded secret, accessing it covertly, mostly on work-from-home days.
“We had a significant competitive advantage against our colleagues — our output was so much faster and they couldn’t comprehend how. Our manager was very impressed and spoke about our performance with senior management,” Matt reported.
This anecdote illustrates the immediate, task-level impact of GenAI that may not be captured in macro-level economic analyses. It suggests that the technology is already transforming work processes and boosting individual productivity, even if these gains aren’t yet reflected in broader economic indicators.
To truly understand GenAI’s potential, we must look beyond broad economic projections and examine its impact at the task level. This brings us to the concept of GenAI as a language amplifier.
GenAI as a Language Amplifier:
To truly understand GenAI’s potential, we must look beyond broad economic projections and examine its impact at the task level. GenAI functions as a “language amplifier”.
GenAI amplifies a condensed set of words into detailed, structured outputs.
For example, this 44-word prompt:
“You are a Canadian financial auditor with 20 years of experience wrapping up year-end financial audits, specializing in small- and medium-sized businesses. The client operates carwashes. Write me a memo about the audit, regulatory, and compliance risks associated with the cash balance.”
Gets transformed by Claude.ai into a 416-word memo (See Appendix I).
The delta between the input and the output is the amplification potential, which can be measured in the following dimensions:
- Length
- Added context and detail
- Improved structure
- Appropriate tone
- Linguistic refinement
Striking the Right Balance: AI Amplification vs. Human Verification
As discussed in a previous post, we need to carefully consider how to manage the risk of abdication — the danger of over-relying on AI without proper human oversight. Unmonitored use of GenAI can result in hallucinations, where the AI generates plausible-sounding but entirely fictitious information. This risk was starkly illustrated in a recent courtroom incident involving lawyer Steven A. Schwartz.
Schwartz submitted a legal brief containing fictional legal precedents generated by ChatGPT, leading to potential sanctions. This case demonstrates how uncritical reliance on AI can backfire dramatically. By abdicating his professional responsibility to verify sources, Schwartz allowed AI hallucinations to compromise his work, underlining the critical importance of maintaining human judgment in AI-assisted tasks.
However, the problem wasn’t inherent to the use of GenAI but rather the lack of appropriate supervision. A quick verification of the cited cases would have prevented the issue entirely. This highlights two key points:
- The amount of supervision required can significantly impact the efficiency gains from AI.
- Proper prompting and context-setting can reduce the need for extensive oversight.
To maximize GenAI’s potential safely and efficiently, users should:
- Provide clear, detailed prompts
- Give the AI relevant context, such as attaching reference documents
- Use example outputs to guide the AI’s responses
- Implement efficient verification processes tailored to the task
By striking the right balance between AI amplification and human oversight, professionals can harness GenAI’s power while maintaining accuracy and reliability. The goal is to minimize necessary supervision without compromising quality, thereby maximizing the technology's productivity gains.
Conclusion:
While the AI market shows signs of a bubble, history teaches us to be cautious about underestimating transformative technologies. GenAI’s potential as a language amplifier suggests it could have far-reaching impacts across various industries and tasks.
In our next post, we’ll explore a practical application of GenAI for documenting working papers in a financial audit, offering a concrete example of its potential to revolutionize professional work.
Appendix I: Amplify Me an Audit Memo
The output below was generated by Claude.ai in response to the following prompt:
“You are a Canadian financial auditor with 20 years of experience wrapping up year-end financial audits, specializing in small- and medium-sized businesses. The client operates carwashes. Write me a memo about the audit, regulatory, and compliance risks associated with the cash balance.”
“Memorandum
To: Audit File From: Senior Auditor Date: September 4, 2024 Subject: Cash Balance Audit, Regulatory, and Compliance Risks for Carwash Operations
Overview
This memo outlines the key audit, regulatory, and compliance risks associated with the cash balance of our client operating carwashes. Given the cash-intensive nature of the carwash industry, special attention must be paid to these areas.
Audit Risks
- Completeness of Revenue
- Risk: Underreporting of cash sales
- Concern: Manual cash handling increases the risk of skimming or theft
- Accuracy of Cash Count
- Risk: Misstatement of cash on hand
- Concern: Frequent cash transactions may lead to errors in daily reconciliations
- Cut-off Issues
- Risk: Improper recording of cash receipts at period-end
- Concern: Timing differences between cash received and services rendered
- Bank Reconciliations
- Risk: Unreconciled differences between book and bank balances
- Concern: Delays in depositing cash may complicate reconciliation process
Regulatory Risks
- Anti-Money Laundering (AML) Compliance
- Risk: Failure to detect and report suspicious transactions
- Concern: Cash-intensive businesses are more susceptible to money laundering attempts
- Tax Reporting
- Risk: Inaccurate reporting of cash income to tax authorities
- Concern: CRA scrutiny of cash-based businesses
Compliance Risks
- Internal Control Weaknesses
- Risk: Inadequate segregation of duties in cash handling
- Concern: Small business environment may limit staffing for proper controls
- Cash Management Policies
- Risk: Non-adherence to cash handling and deposit procedures
- Concern: Inconsistent application of policies across multiple locations
- Record Keeping
- Risk: Insufficient documentation of cash transactions
- Concern: Difficulty in verifying the completeness and accuracy of cash records
Recommended Audit Procedures
- Perform surprise cash counts at various locations
- Test daily cash reconciliations and bank deposits
- Analyze cash sales trends and investigate significant fluctuations
- Review security footage of cash handling areas on a sample basis
- Conduct interviews with cash handlers and management
- Test the client’s AML procedures and transaction reporting
- Perform detailed testing of period-end cut-off for cash transactions
- Review and test the effectiveness of internal controls over cash
Conclusion
The cash-intensive nature of the carwash industry presents significant audit, regulatory, and compliance risks. Our audit approach should be tailored to address these specific risks, with an emphasis on verifying the completeness of cash revenue and the effectiveness of internal controls.”
