AI is worth the water, but it's not the product
It feels like right now, on one side, there is a lot of doom and gloom about the AI space overall. One of the often repeated claims is that 20-50 GPT queries use the equivalent of 500ml of water.
For now we'll ignore the problems in the study, such as the fact that this water use isn't contextualized in comparison other industries, the fact that the paper is based on a study that is awaiting peer review and the dubious link that is implied between data center water use and global water scarcity. As well as the fact that, purely in the tech scene it's by far not the worst offender.
On the other side of the argument there are the evangelists that believe AI will bring us to a promised age, nothing will be the same, some "Brave New World" type vision of a life of leisure brought on by the all powerful AGI.
Honestly, I think neither are true. Or aleast that some elements are true of both. We have to admit that in some cases, AI has allowed us to do things reliably we never thought possible. Some of the recent advances you could argue have brought us into a new AI golden age. But, it is an expensive, and heavy tool.
The AI Teenage Years
Spotty, awkward and "It's just a phase!". We've moved past the baby photos where companies would tack "AI" onto anything and everything in order to boost valuations and look more modern and now we're in a different phase.
Rory Sutherland has a great expression to capture this, in terms of a Cambrian Explosion. A new kind of era has begun where we've discovered the electric motor where only internal combustion existed before, but we haven't really figured out yet what works.
This lack of "best practices" is really exciting! There's a lot of experimentation going on right now and we're figuring out this process of finding what does work. What comes after this is that we settle into familiar patterns, everything is focus-grouped and we have less choice. Consider how the number of unique car colours has collapsed since the 60's, everything is one of 3-4 colours now because that apparently works the best. That's not to say that all of this experimentation always succeeds, there are definitely some... odd choices.
"That's great and all, i'm glad it's exciting... what does this have to do with product stuff though?"
It ain't a product.
Early adopters are basically technology hipsters, they love novelty and the esoteric. But, it's a critical mistake to think that this means that everyone's the same. Almost all of your customers don't care how the sausage is made.
Your customers however do see trends that are happening, that companies are adding "AI" tools if they fit or don't because they feel that they don't want to be left behind. The problem is that these awkward integrations are a tool looking for a job, than a job looking for a tool. They're awkward, full of warts and don't help the customer.
It's not like I don't understand the company perpsective, it always goes along the lines of:
- You hear about a revolutionary new technology
- You want to be ahead of the curve, you've got exicted engineers, you want to be cool technology too!
- You've now spent $$$ on this thing, how do we sell it?
The thing is that AI isn't a product, or a feature, it's just one tool out of a huge bag of tools that could help you deliver products never thought possible. But, we need to focus on those product and features first.
AI/ML is a huge family of different tools, methods and approaches (the perceptron first came around in the 1960's!) but it's really only hit the mainstream since ChatGPT. The first instinct is to reach for those tools because they're novel but that's a mistake. There are a lot of statistical methods and heuristic models that can give you significant benefits while being orders of magnitude faster and cheaper. But the problem is that they're not really sexy...
AI can often be the worst way to solve your problem. It can be fat, slow, expensive, non-deterministic and sometimes hilariously opaque. Are those trade offs you can afford to make? But sometimes, the choice between a bad solution and no solution is a bad solution.
How does this all relate back to products? I love blueberry muffins. I'm not going to pay extra for an AI blueberry muffin, why would I? What does it give me. If you tell me that the AI puts more blueberries in my muffin, or thanks to AI the muffin is 20% larger, or perhaps that using AI you can now anticipate when i'm craving a muffin and deliver one to my door. That's different.
So where do we go from here?
- Talk to smart people about what your problems are, someone who can guide you through your options.
- Think about what you want to do! What do you want to achieve?
- Don't think about AI, explore other options first.
- Don't close yourself off to potentially better solutions by being blinded by the shiny new.
Focus on the problem, but enjoy the wild ride. AI isn't our salvation, but yeah - I think it's worth the water.