Festival de Datos reflections (3): Are evolving definitions of AI helpful or confusing?

I am currently attending Festival de Datos in Uruguay.  While my work is more… ‘data adjacent’, it’s a short flight from Peru (my current temporary home) so why not spend 3 days immersed in a world of data for development.  I am sharing a few reflections / questions from the sessions I’m attending here.  If you’re interested in following along, session recordings are available on YouTube.

There has, unsurprisingly, been a lot of AI discussion at Festival de Datos.

However, it seems that since AI went from niche to mainstream over the last few months, the way people use the term and the things included under the banner have shifted.

It seems like AI used to be well-understood in the Computing Science world (basically various types of Machine Learning), while the public perception was way off and much more aligned with Terminator/Skynet-type visions of sentient robots and dystopian visions of the future.

AI in late 2023 in reality is primarily focused around the newly released power of training and applying Large Language Models, but what has happened to the perception and usage of the term AI.

On the one hand, it is good that the use has broadened to include the reality of AI today (even if it does continue to include a fear of world-ending sentient robots, as evidenced in this word cloud from day 1 of the conference:

However, another interesting shift has also happened.  I’m starting to notice that some of the time, when people talk about AI, what they are actually talking about is pretty much any sort of algorithm or data analysis – even pretty basic stuff that we’ve been doing since the 1990s or before!

It’s almost certainly just marketing bandwagon-jumping as AI is so popular now, but … how helpful is this – is it better or worse than the historic disconnect between reality and usage?

Do we need to start using new/better terms to be clearer?

I think I’ll go ask ChatGPT what she thinks 😛