Wall Street Says the Biggest AI Breakthrough Hasn’t Happened Yet

Beyond the Chatbot: Why the Real Economic Shift Begins When AI Stops Assisting and Starts Executing

Photo by Joshua Mayo On Unsplash

Although email was widely accepted and the World Wide Web was beginning to take shape by 1994, the Internet remained largely an invisible force in the U.S. economy. For the typical portfolio manager and CEO, it was a fringe innovative idea with little practical application to their day today business activities. The underlying technology existed, however, it would be many years before this nascent medium became a self sustaining source of world wide commerce.

However, thing turn out differently than what the typical CEO and Portfolio manager predicted.

Today there is growing evidence of a similar disconnect building in the financial districts of Lower Manhattan. Even though companies in the semiconductor sector are valued at stratospheric levels and generative AI models are creating buzz everywhere, many analysts at firms such as Morgan Stanley agree that the significant AI breakthrough is yet to come. Today, we find ourselves in the same place as we did back in 1994 during the dial up era of the Internet. Yes, machines are impressively fast in drafting memos, summarizing meetings, but we haven’t yet seen them act independently as economic actors.

There is a very important difference between this current state of affairs and what is possible in the future. Currently, AI is simply a tool a powerful and fast assistant. Humans still continue to supervise each step of the process. AI is doing a great job at enhancing the abilities of humans, but has not replaced human functions. The next large step, which analysts are attempting to price into the market right now is moving from AI as an assistant to AI as a worker. At that time systems will have the capability to generate economic value autonomously and at the skill level of a highly trained expert. Therefore, there will be an almost instantaneous shift in the basic math of how economies around the globe operate.

The Expertise Bottleneck

Each significant technological advancement has addressed some type of human constraint. The steam engine increased muscle power, enabling people to go beyond their physical capabilities. Electricity enabled the efficient transmission of electrical power to distant locations. Computers improved calculations. And, the internet made access to information faster and easier than ever before. While these changes led to enormous amounts of wealth creation for both individuals and society at large, they all shared a single constraint, they required human expertise to utilize the output.

AI is the first technology to address a human constraint, it intends to enhance expertise itself. Present day Large Language Models are capable of mimicking human expertise. They are able to pass medical examinations, create viable legal briefs and write operational computer code. However, present day AI Models remain dependent upon human oversight. Attorneys must review and approve legal briefs; developers must test and debug code; doctors must verify diagnoses prior to signing off.

As long as humans remain the limiting factor, increases in productivity due to AI will be small around 10% to 20%. This will result in measurable improvements in business efficiency, sufficient to warrant purchasing a new piece of software, but insufficient to transform an entire industry’s profit and loss statement. The “real breakthrough” will occur when a given application becomes reliable enough that human oversight is no longer required to complete tasks. This represents the threshold of autonomy. Once a system can manage a marketing campaign from initiation through completion and negotiate complex multi party contracts without human handlers, the marginal cost of acquiring expertise approaches zero.

THE EFFICIENCY TRAP VS. THE PRODUCTIVITY MIRACLE

Investors view the present day era as the manifestation of the Solow Paradox. You can see the computer age everywhere but in the productivity numbers. Investors see AI in every quarterly report and in every new release of their software. Yet, the overall rate of national productivity has been historically normal. That’s because we’re still in the “tools” stage.

At present, many companies buy AI tools in order to create a competitive advantage. When a company purchases AI tools today, it is typically engaged in a competitive defense strategy. The company is buying AI tools to help it compete against other companies who are also using AI to write emails faster. As such, there exists a perpetual cycle of efficiency where each competitor is working slightly faster than before, however, the fundamental headcount and expense base of the firm is unchanged. The true economic disruption occurs when the technology enables a complete disconnection between labor costs and output.

Look at professional services firms like law, accounting and consulting. All of these types of businesses rely on a pricing mechanism known as the billable hour. The price charged by these firms directly correlates with the amount of time spent by a human expert. Therefore, if an artificial intelligence system can accurately perform 99.9% of the work that would have otherwise required a junior associate, the business model of the firm changes from managing people to managing compute. In other words, a firm that may have required 500 employees to produce X dollars in revenue could potentially now only need 50 employees. The money formerly utilized to fund payroll and lease office space will now be directed towards funding technological infrastructure and dividend payments. This is the type of future environment that Wall Street is attempting to forecast and represents a significantly larger pool of potential profit than merely providing additional GPUs.

INFRASTRUCTURE PREREQUISITE

To appreciate why the largest breakthroughs are likely yet to come, one needs to look at the enormous capital expenditures presently being invested by hyperscalers like Microsoft, Alphabet and Meta in developing large scale data center and energy grid infrastructures. The billions of dollars being invested in developing largescale data centers and energy infrastructures is not solely intended to provide power to support today’s chatbot applications. Instead, this infrastructure is being developed as part of the development of the “expertise utility” of tomorrow.

As is evidenced by the development of railroad networks during the 19th Century, until a sufficient number of tracks were laid down across the country to allow for the emergence of national retail chains and standard shipping practices, investments made in railroad systems were often viewed as speculative bubbles. At present, we are laying down the rails for AI systems. Although significant progress has been achieved regarding pure intelligence capabilities, the next major breakthroughs will depend on achieving both greater reliability and scalability at lower cost.

Wall Street analysts are waiting for several specific technical achievements: consistency, context and agency. Consistency refers to whether or not an AI model produces consistent results over repeated runs. This eliminates the “hallucination” problems common to today’s AI systems. Context refers to whether or not an AI model can process vast amounts of private data without becoming confused and losing sight of its objective. Agency refers to the ability of an AI model to utilize other tools to gather information from external sources like the Internet and make decisions then adjust course when encountering roadblocks.

Once all three elements converge, an AI system transitions from being a passive responder to an active participant. This means not so far into the future we could see AI system becomes capable of functioning as a “worker” that can be assigned objectives and subsequently figure out how to accomplish them through allocating resources and executing plans.

SECTOR SPECIFIC IMPACTS

If the transition to autonomous workers represents the ultimate breakthrough, then the investment landscape will separate into two categories: owners of expertise and users of expertise.

Already in software engineering we are experiencing some early warning signs. While junior programmers are utilizing AI tools to automate writing small pieces of code, senior architects are starting to use them to administer entire repositories. Ultimately, it is not about developing code faster, but rather it is about enabling a single individual to maintain a software ecosystem that previously required a team of twenty. With regard to software companies, this represents a huge opportunity for expanded margins. From a labor perspective, this represents a very difficult reduction in middle tier employment opportunities.

Conclusion

When the Internet was invented it didn’t change the world Immediately. It changed the world when it was economically useful on a large scale.

It could be that artificial intelligence is also coming close to such a moment. Current systems are great helpers, but the actual major advance will be when these systems can regularly carry out valuable tasks with almost no human intervention.

Once such a frontier is reached, the AI revolution characterized by the last few years may eventually only be seen as the first step. The major breakthrough might not be any of the ones that we had so far, it could be the one that Wall Street is already trying to price in.

I also wrote:

[embed]The Most Important Policy Document in America Is Being Written in Secret A quiet coalition was handed the keys to the future of American industry.medium.com

[embed]America Financial Reset Is Here How the world’s most powerful institutions quietly moved from mocking crypto to using it as the blueprint for a new…medium.com

[embed]The New Economy Belongs To The Ruthless Why the modern world is filtered for machine compatibility, and what happens when we finally succeed in erasing our own…medium.com

A Message from InsiderFinance

Thanks for being a part of our community! Before you go:

• 👏 Clap for the story and follow the author 👉
• 📰 View more content in the InsiderFinance Wire
• 📚 Take our FREE Masterclass
• 📈 Discover Powerful Trading Tools