Deep learning is a particular type of artificial intelligence based on the principle that computers can teach themselves how to improve. This approach enables the technology to build complex concepts out of simpler ones. The result is a deep layer of concepts built on top of one another – hence the name ‘deep learning’.
Deep learning lets machines train themselves to perform tasks, with huge potential for improvement over time as they access new areas of data and experience.
Scale is one of the key principles of the deep learning approach. Deep learning networks can now be huge and process vast amounts of data. What’s exciting about deep learning, in particular, is that it’s a growing area in every sense of the word.
Because it’s less dependent on human-led input to improve compared to other types of AI, there’s potential for exponential growth in this area as deep learning allows machines to make progress for themselves.
Deep learning’s neural networks get better the more data they access. The bigger the data set the network has access to, the more insightful the programme becomes. Either these programs can be trained to produce particular outputs, or they can learn to produce more accurate output by themselves over time.
These programmes are well-suited to complex challenges where it’s very hard to pre-train with an algorithm to achieve a particular predefined output. In these situations, AI that can learn and adapt and become more accurate over time through exploration can usually beat other types of AI that are more programmer-led.
Applications of deep learning technology
There are many ways in which the complex deep learning network can be applied to industry. The technology is already in use by security systems used by the banking industry, where it’s used to identify handwritten cheques and identify fraudulent transaction patterns.
It has medical applications that are just starting to be revealed, such as applying deep learning to genetics to understand how particular genes are related to vulnerability to disease.
It’s being used in the automotive industry so that self-driving cars can ‘teach themselves’ how to handle various driving conditions and scenarios safely.
Now that machines can improve themselves without much human input, growth in their capabilities is likely to happen quickly. What this means is that companies employing the technology are likely to be able to innovate very quickly.
This could mean that they get more adept at spotting commercial opportunities or they’re able to improve supply chains very efficiently. Brands employing deep learning technology may be able to make vast improvements in customer targeting because their AI can spot patterns in behavior that are more likely to lead to sales.
Deep learning offers the potential for major advances across many different industries. The common factor is data: do companies have access to it and can they apply machine learning to it in a way that’s effective. If businesses can do this, there’s huge potential to innovate very quickly via machine learning.
Inequalities of access
Even industries that don’t directly employ deep learning could benefit from indirectly. Take the weather for instance. Even in today’s sophisticated world, many industries are highly dependent on the weather conditions at any particular time. From ice cream manufacturers to clothing retail, farming to aviation, more accurate short and long range weather predictions could really improve how they do business.
The advantages of better forecasting may not be universally shared. Companies that can access more accurate forecasts will be at an advantage over those that cannot. This will benefit companies working in markets with established national meteorology services, such as much of Europe.
Many emerging markets don’t have the resources to gather the data that deep learning would need to base weather forecasts on. Many countries also restrict access to weather data, which could lead to inequalities of access to this data, particularly for smaller players.
Weather is just one example of how deep learning can both improve things for everyone and also threaten the status quo. According to Gartner, 52% of big global companies have disappeared since 2000; mainly through acquisition or bankruptcy.
Disruptive technology is the key reason for this turnover, and deep learning is likely to be a highly disruptive force in the business landscape.
One of the key growth opportunities deep learning will afford brands is probably going to be the opportunities afforded by market churn; the result of this disruptive technology creating winners and losers out of established players.
Deep learning may also see new market entrants that were previously not players in the industries they penetrate using deep learning. An example might be natural language processing startups entering the customer service space.
Companies that can employ deep learning effectively are likely to have advantages over those that don’t, whether it’s because they can improve their cost structures or just spot new opportunities thanks to clever use of market and customer data. Innovation is likely to be fast and merciless in the business landscape.
But there is one limiting factor – a lack of skilled and experienced labor working in this area. Not only is there a shortage of AI talent, but there’s also a lack of understanding about the capabilities of AI among decision-makers.
Those that can access the advice and talent they need to understand how to respond to the new technology could be a considerable advantage over those that cannot or will not engage with it.
Shortage of skills and knowledge in this area is likely to lead to inequalities of access to the technology, which will compound the ‘winners and losers’ effect of deep learning technology.