What are our thoughts on AI? Not that it really matters…
We didn’t use AI to create this website, as you can probably tell, because an AI would have done a better job. We don’t use AI to write our content either, that’ll be obvious too :)
Joke’s aside, and the tricky to measure (good and bad) global Sociological, Environmental, and Economic impacts on the world aside, AI is here to stay.
In years to come (which might be very soon) there’s no doubt in our mind that we will look back on the AI Revolution, just like we did and continue to do so when we talk about the Industrial Revolution.
Let's put AI into context
Putting everything back into context, to be clear, we’re not anti-AI, even if it might seem so. AI is proving to be instrumental for business.
Staying in context from now on (because AI is such a broad “thing”), as mentioned in a previous blog post, AI is great for producing impressive simple websites, prototyping, and for rapidly implementing systems that are “ready” for the market in the short term, maybe not the long term though.
Why short term and not long term?
This is obviously just our opinion, but we’ll explain our thoughts. If you ask an AI platform to build a system for you, it does a good job of doing what you ask. If you ask an AI to make adjustments because the first attempt wasn’t quite right, the AI does a very good job of responding to your demands.
This is all great right? Well, we have some concerns… Behind every website or online technology platform there will be backend and/or middleware and/or frontend code. The code that the AI’s are producing isn’t optimal, but in the short term this probably doesn’t matter too much.
In the longer term there are some things businesses should be worried about:
- The sub-optimal code just gets worse as you ask the AI to make changes.
- The AI platforms are often using proprietary technologies on top of common frameworks. It’s the proprietary technologies that over time gain a tighter grip leading to unexpected “lock in”, even if the service provider claims otherwise.
- If the service provider decides to change their technologies or just do things differently, there isn’t much you can do about it. Due to the sub-optimal code, such changes will affect the way your website/system functions, and as such impacting the service you provide to your customers.
- As the code gradually gets worse and worse over time the option to export the code and data because “the code and data is always yours if you want it” becomes somewhat useless because it’ll be really hard to do anything with it.
- If the system generated by the AI is your business, we’re confident that the underlying infrastructure being used will scale well. However, we think in the longer term (perhaps more like medium term) that there will be a tipping point where it’s easier just to throw everything away and start again. This will be more to do with sub-optimal code become harder and harder for the AI to keep appending to, and security holes appearing due to the complexity and messiness of the code.
- ... one more thing, more complex systems built using AI platforms become very difficult to test. Test environments aren’t really a thing, so changes are constantly being applied to the live system. User acceptance testing is what you rely on, which is a bit uncomfortable because you become a customer for your own service at the same time as your real customers.
What’s the alternative?
The best of both. We believe that AI platforms are excellent for prototyping, and going to market quickly. But, if the online system you build, using the AI platform, is business critical; as soon as you’ve proven the market (if not before) you should be making plans to take ownership of the code, and have one or more developers re-write it for you.