By Chris Adams, Head of Data & Tech Services at Madgex.com
Editor’s Note: Don’t miss TAtech’s Leadership Summit on AI & Machine Learning in Talent Acquisition on April 6-7 in Chicago, IL. It begins with a special half-day program – Artificial Intelligence 101 for non-technologists – followed by a one-day Executive Forum covering the best practices for building, buying and using smart machines and systems in recruitment.
Artificial Intelligence is the latest hot topic in technology, and so, subsequently, it’s become the latest thing in ALL industries. Whether you are #rechtech, #fintech, or #regtech, somebody somewhere is purporting the use of AI to transform your business. As with any latest tech topic, much hype is made, with AI being heralded as being as revolutionary as the discovery of electricity and the invention of the light bulb, and often claimed as the new gold-rush of the 21st century. For many commentators, AI will be the 4th industrial revolution, changing our workplace beyond anything that is recognizable today. Having as much, if not more impact on industry as the introduction of steam, electricity and modern computing have had.
There is some reason to believe that this may be the case – a large number of world leaders are on board. The British Prime Minister declared at Davos in January 2018 that the UK will be a leader in AI and China declared in 2017 that it will be the Global leader by 2030. Russia’s President Vladimir Putin ominously stating that whoever leads in AI will rule the world. The political elite clearly think that this technology is a game changer. With newspapers reporting that up to five new AI companies are being created each week, it definitely feels like a Goldrush environment and everyone is eager to get on-board.
So, within the #rectech space, should you be rushing to get on board? How do you separate the hype from the reality, and know when and how to use AI? The first thing to work out is what AI is and what it isn’t.
AI Isn’t New
The first thing to remember is that AI isn’t new. The science and understanding of AI have been around since the late 1950’s. The field of Artificial Intelligence was established in 1956, but it took until the 21st Century for the technology to reach a maturity where it escapes the research labs and enters the workplace.
One of the biggest drivers of this has been the tremendous power of computers today. What may have taken over 3 weeks to compute in 1970 today can take only minutes. The democratization of AI and cloud computing in recent years has seen AI take center stage, bringing the power and the tools out of the lab direct into the hands of engineers and data analysts.
What is AI?
AI as a term has been abused by the tech media to explain anything automated. For example, programmatic advertising which determines how long an advert should display in certain slots on your page based on the number of clicks, views or applications may seem intelligent, but there is no intelligence there. The system is automating rules – there may be thousands of rules being implemented but there is no understanding, self-learning or self-rule creation. In my view this is not AI, and anyone who tries selling you this as AI is misleading. So when someone talks about AI today, what do they mean?
Intelligence comes from the ability to make predictions – being able to look over data, identify patterns, understand the drivers of those patterns and then use that information to make a prediction through which similar data presented back can then be categorized. For example, a Doctor looks at MRI scans and over years has been trained to identify the patterns in the images which identify the presence of cancer. When the Doctor looks at a scan, they are making a prediction based on their experience and knowledge of those patterns. That prediction can be wrong, but with practice the Doctor is less wrong over time. In the realm of recruitment you may have someone in your organization who categories job postings before they go live. When they see the job title, text, skills and employer they make a prediction as to the industry, function and sector of that role. Hopefully they are correct, and over years of practice they become more right than wrong.
Predictions like these is where artificial intelligence starts coming into play with the use of Machine Learning. Machine Learning is a set of technology and math that can process a set of data and spot/learn patterns then create models from which a computer can predict the likelihood that a new set of data matches the previously learned patterns. Going back to our examples above, Data Scientists have fed the digital MRI scans through Machine Learning techniques and have created a model that can predict when the scan shows signs of cancer. The point where this artificial intelligence becomes useful is when the AI model has a better hit rate than the human counterpart. The key link though is that AI is driven from Machine Learning technologies. It is the machine learning that provides the ability of a computer to predict and provide that Artificial Intelligence.
If a sales person says that their system is powered by AI, what they mean is that they are using Machine Learning to identify patterns, and then are doing something with those patterns to predict going forwards.
What Makes Good AI?
The basis of all Machine Learning is data. In order to build models, a machine needs data, but it also needs good data, and a large amount of it in order to make good models.
Machine Learning can be thrown at any set of data, but that does not mean you will get good results or predictions back. Feeding a ML algorithm a full set of Harry Potter novels, can be fun, but doesn’t mean you get anything useful out.
If you’ve already gone through a Big Data strategy, then you’re likely to already have a good set of centralized data. You can probably start to apply some Machine Learning to that data today. Be careful though, with small data sets, or incomplete or un-representational data you may introduce bias, or embed existing biases that you’re hoping Machine Learning will remove. AI is not a silver bullet, and to expect it to be can lead to costly mistakes.
Using AI in the #rectech Space
So, should you be rushing head-strong into the AI stream, and starting to employ data scientists? The answer is, it really depends. It’s an exciting area and definitely one that has big potential benefits from providing new services, to streamlining business processes, to solutions to your problems through computer intelligence. All successful AI projects in any space though have a clear problem that they are trying to solve. Without that clear business problem, all you’re doing is spouting out bad Harry Potter novels.
So, if you have a clear business problem to solve and a good set of data, then you’re probably in a good position to start using Machine Learning to help solve that problem.
If you’re not, then starting now to identify a strategy for AI and ML is essential. Identify those problems you’re trying to solve, and work out what data sets are needed, then put in the processes to capture that data. If you can’t capture that data yourself, can you identify partners who can provide you access to that data, especially a wider set of data that will increase the quality generated AI?
As more and more businesses jump on the bandwagon, you’ll be presented with tech opportunities that are ‘powered’ by AI. Some of these may be great, but in the gold-rush many may be fools-gold. Before rushing in, make sure you understand:
1. How are they using AI in the process?
2. Where are they getting their data from, how much data they have access to?
3. How are they dealing with bias of that data?
If they can’t explain the AI process to you, then the risk of the AI being biased and un-suitable for your problem may be high.
Clients using Madgex job boards benefit from being part of one of the world’s largest recruitment data sets with over 23,000,000 jobs and over 2,000,000,000 job alert email sends. With Madgex’s new Data Charged initiative – our Data Science Team – are focusing on solving our clients’ problems through using Machine Learning and AI.