The end of software as technology

Technology is an old word for describing what is new. 

The steam engine was called a technology. The telephone was called a technology. The washing machine was called a technology. Physical things.

Military-funded research and the relative peace that followed the Second World War accelerated ever more complicated machines. This gave rise to systems engineering and the importance of code — best exemplified by its role in the Space Race. The era of code was born.

For a generation of us, the word “technology” has been synonymous with software. When we said “tech companies” we imagined cloud computing, ecommerce and social media; when we said “tech startups” we meant SaaS and mobile apps. When we said “tech workers,” we mostly meant software developers. That’s changing.

Starting with $4.6 trillion in pandemic response initiated under President Trump and continued with another two trillion dollars between a trio of spending bills signed by President Biden, the American government has forestalled recession and stoked inflation with spending. (By one measure, the US national debt has essentially doubled since 2015.)

Besides the size of the federal spending over the last few years, what stands out is what we’re spending on: building stuff. 

The $50B CHIPS Act is subsidizing semiconductor plants in Arizona and Ohio. More than $62 million in Build Back Better funds are advancing the robotics and autonomy cluster in Western PA around Pittsburgh. Coalitions from Baltimore and Philadelphia are both finalists for $70M in EDA Tech Hubs funding for bids in biomanufacturing and precision medicine. 

The outputs are intended to be semiconductor chips and robots and gene therapies. Physical things, not digital things.

That’s not to say software won’t be involved; of course it will be. It’s just that software is just kind of baked in now — a bit like washing machines and the telephone. 

AI rules are being written right now

Artificial intelligence is a kind of software, but just about everyone is betting on it marking a new stage. Moreover, the primary limits and opportunities for AI lie with semiconductor chips, energy storage and compute power — decidedly physical things. 

So, how big is the AI moment now? Despite what anyone on stage at TED might tell you, there are still roadblocks that could slow the progress of this generation of AI. There are limitations of data sources, copyright, regulation. But I still agree something has already changed. 

In recent weeks, Technically hosted three dinners of our most recent RealLIST Startup founders: one in DC, one in Philadelphia and one in Pittsburgh. Between 30 entrepreneurs whose companies were three years or younger, most told us AI tools are part of why they plan to raise less outside financing and hire less than they would otherwise.

That view from the trenches matters because it’s harder to see at the enterprise level. I spoke last month to a lead AI engineer at a Fortune 50 retail company who was just given permission to use a software copilot. Legal is very concerned about who owns code suggested by a bot. 

That’s understandable because the AI rules are being written right now:

• In January, a computer scientist appealed the U.S. Copyright Office’s decision to deny copyright to an AI-generated image (depicted above) because, they told him, only humans can hold copyright. To which he challenged: Then why can corporations hold copyright? 
• Then in February, the US Patent and Trademark Office released new guidance that an AI could not be named an inventor on a US patent — but that humans can receive patents for work done with AI. 
• Just last month, agreements of varying importance were announced by the UN, the EU and the US and UK-jointly that commit to establishing regulation for gen AI. Follow that alphabet soup.

Say what you will, but so far, we have moved faster on the threat that AI will take our jobs than we did on social media rotting our brains. And good thing, because this change is coming fast.

Technical.ly was given a preview of the internal results of what to date is likely one of the largest deployments of a software copilot within a single company. Across a sample of hundreds of developers, the large enterprise saw “a significant increase in coding speed” and “a significant increase in code quality.”

Does this mean we won’t need software developers any more? 

Not necessarily. The Jevons Paradox shows that as energy gets more efficient we often use more of it — thereby increasing demand on the resource. For example, improved fuel efficiency standards for cars results in more driving. Does far more efficient coding mean we’ll need fewer software developers over time? Or will lower software costs increase demand for coding so much that new jobs will emerge? I’ll see you at the bar.

One thing we’re sure of is that for now, the machines need to be made by someone. That’s why we’re building stuff again, and once again shifting what we mean by “technology.”