What does the AI bubble mean? It means software has no intrinsic value. Since software is worthless, creating good software is a waste of effort.
This is a conclusion of reached from observing the software industry for a half-century
Gather 'Round The Fire All You Youngs
Once upon a time, software had value.
It was all proprietary. You either bought it from a vendor or licensed it. There was no Open-Source Software Movement. Before 1983, the idea barely existed. I saw the seeds of open-source in the form of vendor-specific user groups that might distribute clever software among customers.
Since the accountants wrote checks to companies, software had to be an asset on the balance sheet.
Since there were vendors and service providers, there were contracts and contract negotiations. And three-martini lunches to discuss terms and conditions.
This changed.
The Outsourcing Years
There came a time when management consultants told CIO's to outsource. This did this by reciting catch phrases that amounted to "Focus on your core business, outsource everything else."
The management consultants could then recommend their firm as an outsourced IT provider. (This was my career for decades.)
The vision of a modern CIO's job was oversight of contracts with services companies. Minimal in-house staff. Roomy offices in largely empty buildings.
Note the shift in perspective required for this. When the accountants wrote checks to services companies, IT was an expense. That's where their interest in the subject ended.
Software? Bah. Software was an asset of some flock of external IT service providers. The accountants could wash their hands of it. All they saw was a stream of expenses, and they were happy.
The Software Patent Years
Starting somewhere after 1994, software became patentable. (Maybe around 1998 and the State Street Bank & Trust v. Signature Financial Group decision.)
This decision forced a new perspective. The accountants had to consider software as an asset (again.) If software was part of a patent, and a patent was balance-sheet intellectual property, then, software is an asset.
Creating software can (and should) be a Capital Investment. The accounts liked it when money could be spent to add assets to the balance sheet.
However. The accountants realized there was a potential problem. They were concerned about the human resources. When the people creating intellectual property assets were outsourced to a services company, the intellectual property could (in whole or in part) belong to the services company. This was bad. The asset's location on a balance sheet had become ambiguous.
The only way to really get ahead in creating software assets was to stop outsourcing software development and bring all the work back in house again. This doesn't happen too quickly. It takes some companies a full decade to finally cease outsourcing of software development and maintenance.
Note the shift in perspective: from "software is someone else's problem" to "software is our intellectual property asset."
The Advent of Free and Open Source Software
In the '00's, we debated the real meaning of "Free and Open-Source Software" (FOSS). There were two camps:
- "Free" as in "Free Beer".
- "Free" as in "Free Speech."
The free speech crowd did not embrace unconstrained commercial use of open-source software. The free speech crowd felt authors and maintainers should get compensated from people using their intellectual property to generate revenue. They formulated things like the creative-commons no-commercial-use licenses (https://creativecommons.org/licenses/by-nc/4.0/).
The free beer crowd has mostly won out. Free beer is a tangible asset on the balance sheet. Free speech? Not so much.
The accountants asked the CIO to look at the enterprise software portfolio. The accounts could see the company pays almost nothing for backbone products used many times per day. There are no fees, no checks being written, no visibility into how this software arrives in the enterprise.
A shrewd CIO can negotiate lower prices for services because the service providers aren't paying vast licensing fees for proprietary products. The price of software is approaching absolute zero.
Since there's no money being paid to use FOSS, then, what is the value?
Current State
It appears the consensus is: FOSS is worthless.
The accountants aren't paying for it. They aren't paying anyone to maintain it. It can't go on the balance sheet. It's as free as air.
This leads to: Zero cost = zero value.
By extension from FOSS, all software is worthless. As worthless as an undeveloped tract of forest. As worthless as an unimproved building lot surrounded by parks and playgrounds. As worthless as the river flowing past a factory that emits hazardous runoff.
Because it's worthless, software is off the balance sheet, it's not a thing to consider.
Want to use unproven, hallucination-generating LLM sparkling autocomplete tools to generate software?
The approved accounting short answer is "Go ahead. We don't care." The long answer is that using zero-quality tools to build a zero-value thing is within the realm of a "rational" use of capital.
Remember, the presence of FOSS proves software is never an asset. Throwing money at an LLM is simply an operating expense (and a tax deduction.)
The AI Bubble Can't Burst
The AI bubble can't burst until we can demonstrate software has an intrinsic value. Until we show that software has value, the AI bubble can only leak and sag.
This can't happen until a number of companies with a specific profile fail spectacularly. More than one company that layed off skilled staff to use LLM tools has to fail. And the (billionaire-owned) media outlets have to report -- emphatically -- on how these companies fell behind in the technology race because LLM code isn't good and doesn't keep up with people. Further, the (billionaire-owned) media outlets must also perform some rudimentary analysis of the failed company's software to show that software has an intrinsic value. (I'm not optimistic.)
It will take a lot of work to convince people that software captures business knowledge in a form that is so clear that a silicon chip can make sense of it.
I'll try to help.
An LLM takes knowledge (in the form of a prompt) and dilutes with rubbish additional text or code generated from unknown sources.
Important
Software Captures Knowledge
Software does have intrinsic value as a form of knowledge capture.
An LLM can only dilute the knowledge.
An LLM's summary of software means details are elided. LLM's hallucinate. The missing details have replaced with nonsense from a large training set that includes the Bee Movie.
It's difficult to demonstrate the value of intelligible software written by people for people. In spite of the difficulties, we need to do more of this.