Is this AI worth anything

“is all of this (AI) actually worth anything?” from CNN Business post is important question.

Tech/AI people often sounds like this: “I tried this 8B 4-bit quantized LLM model with this multilanguage sentence embeddings, everything running in our datacenter on CPUs. Showed my demo to few people and they think it would be helpful.”

While experimenting and learning is very important this type of thinking misses the point. Whatever technology we are implementing, it always gets down to two simple questions:

1. How much money will it cost?
2. How much money will it bring (or save)?

There can be discussions about details. Cost of development vs cost of running, cost of securing the technology. In the end, the most impacftul data products solves real problems a critical amount of people have. Quantizations, neural networks, stable diffusion, emergent abilities, … are for conferences and meetups. The core competency of data science/AI departments is bringing value.

I see a lot of business people learning about technical and AI topics. And not enough technical people learning about business, management and domain topics.

I also highly recommend listening to the Experiencing Data. They deliver approachable episode about differentiators between successful and unsuccessful data science teams.