2020 - An AI Autumn....
"If we want to scale to more complex behaviour, we need to do better with less data, and we need to generalise more."(*1)
During the 2015-2019 period, it was all about "AI is going to change the world", "AI is awesome, look at the amazing things it can do", "jump on the AI band-wagon".....etc...etc...etc...
I worked on many a project during this time period and I often found myself sitting back (the ability to do that was a rare occasion, I can tell you) and asking, "Where is the AI in this solution?".
Everyone else, the project stakeholders, executives, Data Scientists (I still have beef with that job title, but I'll leave that for another day) were all strutting around dropping the "AI" word and I just couldn't see it. Yes, we had a few isolated "smart" bits, but they were just algorithms triggered by good old fashioned code on certain events that made life a bit easier / quicker for the end user - no rocket science here, I'm afraid. If you knew nothing, you'd be amazed. If however, you knew how things worked, you'd be more "meh", "is that all?"...
One of the mantra's that we've heard over and over again is, "to do AI, you need DATA and a LOT of it".... and that has always been my concern, to be able to do something intelligent you have to have millions of if not billions of bit of structured and unstructured data that you then sort and filter, slice and dice and then hope that some math genius with a PhD can make that data make sense and help your IT system / application. Sometimes it works, but it's a narrow case usually, if the data becomes contaminated, the results go haywire.
So that brings me to the quote above (*1) - that is where I believe we need to get smarter - do more with less. It's not about 25TB Data Lakes that we can analyse and try and gain some intelligence from, it's about pre-filtering that data and diluting it (homeopathic style, we keep the essence but lose the quantity), then I believe we'll start to make proper headway into the world of AI.
So, are we going into an AI Winter as this article suggests? No, I don't believe so. The "new shiny toy" phase has passed and now we are past the noise, we can start to just focus on how and where we put those "smart" bits into existing IT solutions.... AI will develop and evolve from that, but it's going to take at least another 10 years to get to where the sales people were selling it over the last 10 years.
During the 2015-2019 period, it was all about "AI is going to change the world", "AI is awesome, look at the amazing things it can do", "jump on the AI band-wagon".....etc...etc...etc...
I worked on many a project during this time period and I often found myself sitting back (the ability to do that was a rare occasion, I can tell you) and asking, "Where is the AI in this solution?".
Everyone else, the project stakeholders, executives, Data Scientists (I still have beef with that job title, but I'll leave that for another day) were all strutting around dropping the "AI" word and I just couldn't see it. Yes, we had a few isolated "smart" bits, but they were just algorithms triggered by good old fashioned code on certain events that made life a bit easier / quicker for the end user - no rocket science here, I'm afraid. If you knew nothing, you'd be amazed. If however, you knew how things worked, you'd be more "meh", "is that all?"...
One of the mantra's that we've heard over and over again is, "to do AI, you need DATA and a LOT of it".... and that has always been my concern, to be able to do something intelligent you have to have millions of if not billions of bit of structured and unstructured data that you then sort and filter, slice and dice and then hope that some math genius with a PhD can make that data make sense and help your IT system / application. Sometimes it works, but it's a narrow case usually, if the data becomes contaminated, the results go haywire.
So that brings me to the quote above (*1) - that is where I believe we need to get smarter - do more with less. It's not about 25TB Data Lakes that we can analyse and try and gain some intelligence from, it's about pre-filtering that data and diluting it (homeopathic style, we keep the essence but lose the quantity), then I believe we'll start to make proper headway into the world of AI.
So, are we going into an AI Winter as this article suggests? No, I don't believe so. The "new shiny toy" phase has passed and now we are past the noise, we can start to just focus on how and where we put those "smart" bits into existing IT solutions.... AI will develop and evolve from that, but it's going to take at least another 10 years to get to where the sales people were selling it over the last 10 years.
Comments
Post a Comment