TLDR:
This article explores the potential profitability and transformative impact of neural network-based AI compared to traditional software. It distinguishes neural networks from algorithmic machine learning, emphasizing their strengths in tasks such as pattern recognition, noise reduction, and data mining. The piece highlights real-world applications—from market analytics and trading signals to social media sentiment analysis during elections—demonstrating how AI can automate mundane tasks and drive innovation. Despite dystopian warnings from public figures, the focus remains on the lucrative opportunities emerging from AI's ability to process and predict complex data patterns.
○
Terminator is a work of fiction, let us be very clear about that before we start!!! What is of interest here is whether there is cold hard cash to be made out of Neural Network based technology. Up until now the vast majority of software has been based on sequentially processed instructions with parallel threads. Defined inputs give defined outputs, and computers always perform the instructions they are given. Neural Networks are a different proposition. They are programmed based on test data and expected outputs (rather than low level programmatic instructions).
Neural Networks: Revolutionizing Software Innovation
Hold on a minute I hear you cry, both The Hawk (Stephen Hawking) and Elon Musk have said this type of AI could be the end of humanity. Well yes, that may well happen, and a big comet might also smash into the earth unless we send Bruce Willis up with a nuke to split it into pieces. In the race between Skynet, The Matrix, Global Warming and a general desire to blow each other up anyway, it is probably a close run thing. If you are really worried about the machines taking over and wiping the skies the best advice I can give is to take the red pill when offered. If you see the same cat walk past twice, adopt the duck and cover position (preferably under a small child's writing desk).
In the meantime back on present day earth, with the likes of Google spending $400m + on small AI research companies, could we be about to see a new sector to rival legendary historical booms? The answer is possibly yes, but not necessarily for the reasons that you will see in the popular press.
Also it is worth making a clear distinction here between Neural Networks and Algorithmic Machine learning. Algorithmic machine learning is based on making predictions using statistical analysis and heuristics. What we are specifically looking at here are Neural Networks which are "trained" by repeatedly giving them test data with re-enforcement signals whenever the network responds with the correct answer.
When the industrial revolution arrived, the luddites feared the end of life as they knew it, which was mainly true, but it was not the end of life generally. Machines did not replace humans, they did replace some of the roles humans performed and they vastly augmented other roles humans performed. Here is the secret to understanding where the big opportunity lies in AI.
The immediate opportunity with AI, like the industrial revolution, is in performing menial tasks that humans can do, on a scale that is far more cost efficient using using function specific targeted Neural Network blocks. To understand this better let's go over some of the activities that Neural Networks are very good at.
- Pattern recognition
- Background noise subtraction
- Taking diverse input streams e.g. stock prices AND social media trending trigger words
- Staying awake 24/7
- Self training with long, intricate and essentially boring datasets
- Testing themselves relentlessly with historical data to see "how good they are"
- Creating clones of their most successful configurations
With the exception of staying awake 24/7 and cloning themselves, these are many of the activities human brains currently do better than linear software.
If the above does not impress you then you probably have far more exciting opportunities on your plate, but if you read carefully then you should realise the enormous potential here. This time unlike the industrial revolution, once you have created your first factory, making your second is more akin to copy and paste than building a whole new set of production lines. To spell this out, think incredibly accelerated timelines to build lucrative new processes you have not even thought of yet.
So far this has all been fairly ethereal, like all those weight loss adverts on Facebook there comes a point where you want to know whether another 5 minutes of reading is going to get me to something useful or just a Paypal page asking for £13
The true answer to how you might make money with AI, is as varied as how you might have made money in the dotcom boom, but never the less let's look at a practical use of AI and maybe a few lightbulbs will start flickers amongst the entrepreneurial readers.
AI in Market Analytics and Data Mining
There is a saying in the markets "buy on the rumour sell on the news", and you often see this when big companies release great quarterly results only for their stock to go down slightly. Some of you are smiling at this and others asking "really are you sure??". So the reality behind this is that before those great quarterly results come out, the trading community already has a pretty good idea of what is going to happen and got all the trades in early. When the figures finally come out, it is time to move on and pick up something new. The City spends a small fortune tracking a vast variety of sources, and when you look back over the results, whilst Ipsus Mori may not have known how to call elections and referendums, quite often the FTSE and other markets have clearly called it long before the doors have opened.
Real-World Applications: From Trading to Social Media Monitoring
For some of our clients during the UK Elections we spent a lot of processor power tracking the emotional temperature towards the leaders during campaign, and also the interactions between key players in the media spotlight.
Initially our human coordinators struggled to make sense of the ever changing landscape, but then certain patterns started to emerge.
Support in social media for David Cameron would peak between 7.30am and 8.30am
Support for Ed Miliband would peak at about 11am in social media
Popularity of the candidates was also influenced by the day of the week and TV schedule
Once we had 14 days of data we started to feed that data into a neural network, and what we discovered was fascinating. The Neural Network was pretty accurately able to predict the expected sentiment ratios and volumes, and where it was not able to predict it, that was because some event was happening which was changing opinion. Very quickly we became less interested in the raw data and far more interested in the actual data with AI predicted data subtracted. This was the data that clearly showed us which way the landscape was changing and in which direction public opinion was moving.
We often track Obama in social media, simply because when significant news hits America he will be involved voluntarily or not in the conversation.
Have a look at this plot below when another black man gets shot by police. The troughs in the data are nighttime in the US. Its pretty clear to see where the spike for Actual Data does not fit in with the Neural Network prediction and then it is a fairly straight forward task to research what has cause the spike.
I hope this article has whet your appetite for what can be achieved with AI. Trading information and data is one of the most lucrative industries on the planet right now, and using AI to help mine that data is just one way focussed Neural Networks can replace boring repetitive tasks typically preformed by humans.
Finally I would like to leave you with the story of a good friend and expert in AI. He is training a Neural Network to recognise his own cats, and has the output linked to a sprinkler which keeps any unwanted visitors out of his garden. As you can start to imagine, the possibilities are endless....