I received an interesting call today that really got me thinking…
It was another potential client, wanting to create an AI to automate part of their business process. They had seen the new service we’re offering which is proving to be very popular. My first thought was that we couldn’t possibly take on another project. Our Deep Learning specialists have 8 projects on the go; we don’t have the resource. But I hate turning down business, so I called our head of R&D to check exactly when we might have some spare capacity. His response: “Let’s take it on, the Deep Learning platform does all the work anyway and my guys are looking for the next project.” The error in my initial thought was in the mindset of traditional software engineering. A prolonged development life-cycle, with a lot of coding, a lot more testing, code reviews, bug fixing, and re-coding. Not to mention the arduous task of manually engineering the complex computer vision pipeline: what features should we use, how shall we model the variance, what machine learning algorithm would work best?
The truth is, we no longer do any coding. We don’t conceptualise the pipeline and we don’t design special new features for that awkward computer vision problem. As for testing; well, we didn’t code anything that needs testing. Granted that we test the accuracy of the final system, but this is an automated performance measurement and we don’t need to review and correct the code that we didn’t write in the first place. Deep Learning is doing the work for us. We supply the training data to our proprietary Deep Learning platform and let it train. The coding of the platform was completed last year, it’s been tested to death, as have the Neural Networks produced – they just work. We’ve now got the whole process down to five quick and simple steps:
Free Consultation: Is an AI solution appropriate?
Feasibility study: Assess the data in detail.
Data preparation: Run automated scripts to format the training data as required.
Training: Let the Deep Learning platform do its thing.
Completion: Package the Neural Network, run a performance assessment and deliver.
So, are our software engineers going to be out of the job? Far from it. Deep Learning is proving to be so successful across such a broad range of markets (Finance, Medical, Defence and Security, Manufacturing, Food Production, Law, the list goes on) and because we are now able to tackle problems that would have been impossible in the pre-AI era, the market is huge. The first two steps described above is where all the human effort is now focused, and this does still require expert knowledge in Deep Learning. Thankfully at Aurora-AI we have these experts to hand and they can now turn a project around in weeks, what would have taken months pre-AI. So we’ve taken on that 9th project and we’re now looking for number 10.
If you’re attending PTE 2019 in London ExCel (26th – 28th March), we would love to speak to you. You’ll find us on stand #1090. Please use our contact page to make an appointment.