Why Was He Wrong?

This article at Futurism by Joe Wilkins caught my eye. Mr. Wilkins observes:

With so many wild predictions flying around about the future AI, it’s important to occasionally take a step back and check in on what came true — and what hasn’t come to pass.

Exactly six months ago, Dario Amodei, the CEO of massive AI company Anthropic, claimed that in half a year, AI would be “writing 90 percent of code.” And that was the worst-case scenario; in just three months, he predicted, we could hit a place where “essentially all” code is written by AI.

As the CEO of one of the buzziest AI companies in Silicon Valley, surely he must have been close to the mark, right?

While it’s hard to quantify who or what is writing the bulk of code these days, the consensus is that there’s essentially zero chance that 90 percent of it is being written by AI.

Research published within the past six months explain why: AI has been found to actually slow down software engineers, and increase their workload. Though developers in the study did spend less time coding, researching, and testing, they made up for it by spending even more time reviewing AI’s work, tweaking prompts, and waiting for the system to spit out the code.

Unfortunately, Mr. Wilkins does not actually answer the question that forms the title of this post: why was Dario Amodei wrong? Will none of his predictions come to pass?

I’ve been experimenting with large language model artificial intelligence (LLM AI) models for two years now. Based on my limited experience there are several inherent problems with their use.

The first is that applications created using AI aren’t designed. They’re just implemented. When features are added or bugs identified and reported, the applications are re-implemented with run-on effects. I’m seeing this in the software updates on my smartphone and tablet. They are being “improved” rapidly to the point of becoming unusable.

While that’s okay for simple tools only used occasionally, it can be disastrous for mission-critical applications or those that have financial aspects.

The second is that human beings just aren’t very good at explaining what they want and/or need. They leave things out. They include extraneous things. They may not recognize the run-on effects of a decision. It takes considerable skill and experience to explain things properly and completely and there are fewer people who can do those things than can grind out code.

The third is that there has been what might called “title inflation”. Seniors aren’t seniors any more. Forty years ago a senior developer had eight years or more of experience. Now five years is considered senior. It isn’t but that’s what businesses are saying these days. Also, as a past colleague of mine once observed, senior here in the U. S. and senior in another country are two different things.

Another is that you can’t rely on AI to test your applications for you. LLM AI models don’t understand anything. They just do what they’ve been trained to do (best case). The implication of that is that only human beings can determine whether something is suitable to task which in turn suggests that human beings need to do the testing. Good testing is harder than coding.

Sadly, none of this makes any difference. All that matters is that a senior developer plus 3 junior developers costs more than a junior prompt writer and a subscription to several AI models Note: there are no senior AI prompt writers because they haven’t been around long enough and such creatures may never exist because of the rapid pace at which they evolve. That’s all that will show up in the quarterly report and whatever President Trump says short term thinking is here to stay. Goosing stock value at the expense of long term risks to the enterprise is too easy a decision to make for modern managers.