I find myself constantly shifting between two camps…
I think that the 70% problem sums up the issue really well. I highly recommend reading it. I am not an AI engineer. I am not a marketer, researcher, or scientist (unlike the person who wrote “The 70% Problem,” who looks more than qualified to contribute valuable insight), but I am a person who, like just about everyone I know, has to use generative AI in some shape or form.
So my day-to-day goes something like this:
- Wake up
- Turn on an electronic device
- Get bombarded by AI-generated content, whether it is in the form of a video, a Google text response, or something else that I see getting passed off as original content
- Feel annoyed
- Drive to work. Find out that my boss has some new requirements, so I desperately need to do/understand x, y, and z by Tuesday morning or I might lose my job
- Use some AI tools that the company is not just allowing, but encouraging us to use. Sit in stunned silence as it gives me code that solves every problem I had in five seconds
- Feel happy
- Realize that the code it generated caused a different problem
- Debate whether to use AI or “traditional coding” to get over the next hurdle
- Rinse and repeat
The tl;dr for anyone who doesn’t like reading bullets? Generative AI solves problems but causes different ones, often fulfilling a coding request by “unintentionally” introducing some completely unrelated issue. This is where I find discussions tend to diverge. Some people believe that generative AI is still getting there, and that it is only a matter of time before AI can solve almost 100% of coding problems without hallucinating. Others believe it will never quite get to the level others are advertising. It will not deliver enough value to justify the high cost of energy it consumes. It will never truly top human capability, because (and I doubt anyone would disagree on this point) generative AI does not truly understand anything. It generates words based on probabilities. It’s not some second brain.
Why I Can Sympathize With Camp #1
Sal Khan likes AI. I know that a lot of people would argue, rightly so, that those four words do not constitute a legitimate argument (nor are they surprising). But in my mind, this is a big deal.
Programming has been a lucrative field for quite some time because of ways in which it is not intuitive to the average person. Small mistakes like a missing semicolon (or async, to the JavaScripters out there) can have significant results, and reading code can be so confusing that a single word could require ten pages of documentation to understand (for example, async).
Generative AI could be like a great savior, something that allows a user to communicate in plain language and watch as dreams come to life. No more gate-keeping. No more friction. Everyone who wants to code can, without a need for a rigorous bootcamp or college education.
This is yet another point addressed by the “70% paper.” The author underscores AI’s potential but questions why he has not observed tremendous increases in the quality of software we use on a day-to-day basis.
Why I Can Sympathize With Camp #2
As a programmer, I find that AI is a flawed but exciting tool. If nothing else, it is good at idea generation. Generative AI is best, I think, when coming very close to a working solution – it can generate the missing line of code or identify the bug holding me back from making progress.
As a writer (er, blogger), I absolutely despise AI with a visceral rage that is confusing, even to myself.
Am I simply annoyed that Medium has become so AI-heavy? Articles about AI, written by AI, with comments left and then responded to with AI? Have I observed AI and human AI users alike post unironically about why we should strive to write AI-generated content for AI users? Yes, and at times I think I am losing my mind.
But some people dislike writing, and for them AI is a good tool. The exact same argument could also be used against me – how often have I relied on AI for grammar corrections, or in Google searches? If I actually wrote for a living and had AI generate ideas for me, would that be such a bad thing?
Closing Thoughts
I watched am AI-generated corgi attempt to scale a rock-climbing wall, and my optimism returned.