The recent trend in the tech industry is to follow the hype and slap the label "AI" on every product that hits the market. While it is a clever marketing tactic, is it beneficial? The answer is no, according to experts.

New research suggests that the current crop of "AI" products, "Generative AI (GenAI)", which are often touted as advanced systems for mimicking human-like behavior, are not only overhyped but also environmentally unfriendly. In fact, creating and running these systems can have a significant carbon footprint, with some estimates suggesting that they may even be worse than running a search engine. 

Experts are now sounding the alarm that the "AI" hype is starting to slow down and this does not bode well for these companies.

I've chosen to avoid using GenAI which relies on the cloud due to data privacy, security and its environmental impact. To be clear: these systems are not truly intelligent - they simply mimic, like a parrot, certain behaviors based on patterns learned from their training data. It's essential to know what you're getting into before using these tools, and to be aware of their limitations.   

So, how do I use GenAI? I prefer to use them in a more contained way, meaning I'm limited to what my computers can handle. While this approach certainly has its drawbacks, at least I know the environmental impact is almost negligible. That setup also doesn't need to connect to an external network, making it private and secure. Currently, I can work with free genAI models from various companies, like Meta's Llama 3.1 model. Unfortunately, most of the models come from companies that don't disclose where the data is from and how it was acquired. Anyway, I mostly use text, so my GenAI use is mostly to re-write articles (like what I did here), though sometimes, out of laziness, I ask it to generate from scratch. *wink*.

What if I need to use more powerful models that cannot be handled by my current computer? Buy a new, more powerful computer, of course! I kid. Currently, I just suck it up and don't use GenAI at all. However, I am looking forward to the Apple Intelligence approach of using on-device text-only GenAI models and designing a carbon-neutral Private Cloud Compute for tasks that devices cannot handle. I believe that this is the responsible way to use GenAI. I wish other companies will clone and improve on this private, secure and environment friendly design. Until then, as I've said, I suck it up and don't use more powerful model, and patiently wait for Apple.