What can AI do for you?

What has AI done for you lately?

Is the AI hype exaggerated? I asked Gemini and here's what it says

That's a fantastic and timely question. The short answer is: yes, the AI hype is likely exaggerated in the short term, but the underlying technology's potential is probably understated in the long term.

and it explains further:

The "General AI" Fantasy: Many headlines imply we are on the verge of Artificial General Intelligence (AGI)—an AI capable of performing any intellectual task a human can. We are not. Current models are narrow AI; they excel at specific tasks.

It then concludes that:

However, once the noise dies down, the underlying technology will continue its slow, steady, and inevitable advance. The ultimate impact of AI on the global economy, science, and the nature of work will likely be greater than most people can imagine—it just won't happen as fast as the most excited venture capitalists hope.

I'm actually surprised by the candid and balanced answer I received. It seems to me that in the short term there aren't many high-impact AI solutions. If I missed any, please reach out to correct me.

No free lunch?

There is a saying, "the people making money during the gold rush are those that are selling shovels". As in any engineering endeavor, the more you dig into a problem, the more questions come out and details become progressively more intricated and layered. Creating agreement across teams, obtaining permissions and overcoming obstacles becomes paramount on the managerial side, while a mesh of issues accumulate on the technical side. One-click solutions are nowhere to be found.

AI affects different parts of an organization in different ways. Outside of simple automation, it has critical problems when used at scale. This complexity involves costs, security implications, data protection regulations, and, critically, the inherent variability of AI output. A single negative experience, such as a poorly written email, is enough to severely reduce confidence in the entire automation system.

Conclusion

AI does offer advantages in specific data engineering tasks, especially those that are repetitive or research-intensive, such as reading and writing documenattion which can be time consuming for new or not-well-knonw APIs. However, it seems to me that, as of now, AI remains mostly limited to basic automations that cannot be considered true business multipliers.

While AI is getting better at churning out text, the rest of the world is getting better at producing non-AI tools that achieve the same goal with fewer problems than AI, e.g. generating boilerplate code, formatting and automatically generating documentation. There are advantages to these non-AI tools. Yes they're not as seducing as the new cool AI stuff but they are typically reliable and maintaned, often part of a CI/CD lifecycle and battle tested by humans. They're not an hallucinated AI concoction.

As with everything in life there are pros and cons. You may go faster with AI but perhaps you lose reliability. I'm suggesting that the shiny lights of AI and "productivity" are hiding the benefits of solid but unassuming technology that has a lot to offer.

Cheers.