AI’s 3 Key Algorithms?
AIs algorithms can be broadly classified as supervised, unsupervised, and reinforcement learning.
• Supervised Machine Learning: works on predictions and searches for patterns on a given set of samples. Supervised Machine Learning Algorithms attempt to render relationships and dependencies between target prediction output and input features. In this, we start from input variables (X) and an output variable (Y) and try to map functions from the input to the output so that they establish a relationship which can be used for prediction. Common models: linear regression, random forest, support vector machines, nearest neighbour, etc. Supervised learning is commonly used in classification problems, such as digit recognition, speech recognition, diagnostics, identity fraud detection and regression problems such as weather forecasting, estimating life expectancy, population growth prediction etc.
• Unsupervised Machine Learning: arranges data into clusters and describes its structure in order to make complex data look simple and organised. Unsupervised learning can be applied when there is no labelled data available for training. Examples are k-means for clustering problems used in recommender systems, customer segmentation and targeting marketing and dimensionality reduction problems for big data visualisation, feature elicitation, structure discovery etc.
• Reinforcement Machine Learning: forwards an action according to the data point and later assess the decision. It utilises observations collected from interactions and takes actions so as to minimise the risk and maximise the benefits. It learns in an iterative fashion. Examples are Q-Learning, Deep Adversarial Networks, and Temporal Difference. Can be applied in Game AI, skill acquisition, learning tasks, robot navigation and real-time decision taking.
What will the future bring?
So what will happen in the near future? Well, as most of us know, machine learning is currently one of the hottest topics in Computer Science. The reason stems from the seemingly unlimited use cases in which machine learning can play a role. Where are we going? What will be the next big thing? Where will we be in ten years’ time? The truth is, we are most probably going to be surprised again by the next hot topic. However, some things are not all that unlikely to play a role very soon.
1. Quantum Computing: Machine learning tasks involve problems such as manipulating and classifying large numbers of vectors in high-dimensional spaces. The classical algorithms we currently use for solving such problems take time. Quantum computers will likely be very good at manipulating high-dimensional vectors in large tensor product spaces. Therefore, it is likely that both the development of supervised and unsupervised quantum machine learning algorithms will hugely increase the number of vectors and their dimensions exponentially more quickly than classical algorithms. This will likely result in a massive increase in the speed at which machine learning algorithms will run.
2. Better Unsupervised Algorithms: As of now unsupervised learning is not yet being applied very successfully in industrial software. Therefore, we expect a lot of improvements, such as discovering hidden patterns in data, or a means towards an end, often called feature learning, to happen before we can apply it to solve “real problems”. It is likely that advances in building smarter, unsupervised learning algorithms will lead to faster and more accurate outcomes.
3. Collaborative Learning: This is about utilising different computational entities so that they collaborate in order to produce better learning results than they would have achieved on their own. An example of this would be utilizing the nodes of an IoT sensor network, or what is called edge analytics. With the growth of IoT, it is likely that large numbers of separate entities will be utilised to learn collaboratively in many ways.
4. Deeper Personalisation: Personalisation isn’t the worst thing per se, but it can get very annoying. We have all experienced recommendations that seem to bear no actual relation to anything that we may actually be interested in. In the future, users will likely receive more precise recommendations and adverts will become both more effective and less inaccurate. The user experience will vastly improve for all.
5. Cognitive Services: This technology includes APIs and services through which developers can create more discoverable and intelligent applications. Machine learning APIs will allow developers to introduce intelligent features such as emotion detection; speech, facial, and vision recognition; and language and speech understanding into their applications. The future of this field will be the introduction of deeply personalised computing experiences for all.
These are things I think can and should happen in the machine learning’s near future, but it is equally likely that some new unknown disruptive technology none of us would have predicted gets introduced. In any case, the future is now, and we have 3 facts that already are changing the whole IT landscape.
They are that:
• Every company can now be a data company
• Every company can now access algorithmic intelligence
• Every app can now be an intelligent app
Yes folks, it’s really easy. And this is again one of the reasons why everyone is talking about Data Science right now.
Who needs ML and why is GeoDirectory important?
While speaking about all that artificial intelligence and machine learning hype, we often forget to address the most important question: Who needs it and why? I would like to address two major users specifically, and only list some of the others:
• Government agencies, such as public safety and utilities have a particular need for machine learning since they have multiple sources of data that can be mined for insights. Analysing sensor data, for example, identifies ways to increase efficiency and save money. Machine learning can also help detect fraud and minimize identity theft.
• Geospatial Data Organisations and Businesses: Imagine being able to train your GIS to perceive and understand the world, and give you insights based on your data. Today, geospatial experts are using machine learning for analysing big datasets (what do these 2 million points actually mean?) and predictive analytics (e.g. forecasting risk).
Financial Services, Health Care, Marketing and Sales, Transportation, and many more.
Most AI and ML algorithms require location information, which is hence becoming more and more important, therefore GeoDirectory data could be a key variable. Here are some examples:
• Market research, market segmentation
• Risk estimation (e.g. mortgages and loans)
• Spatial patterns of sentiment towards institutions/products
• Land use planning
• Housing policy and provision
• Logistics (routing)
• Understanding change processes in variety of fields (e.g. health, transport, housing, environment)
The view on AI from Industry Leaders….
And for the end, some quotes. I won’t comment on them, let them just sink in and tell me your own opinion ;)
“AI is likely to be either the best or worst thing to happen to humanity.” Stephen Hawking
“Worth reading Superintelligence by Boston. We need to be super careful with AI. Potentially more dangerous than nukes.” Elon Musk
“I am in the camp that is concerned about artificial intelligence. First the machines will do a lot of jobs for us and not be super intelligent. That should be positive if we manage it well. A few decades after that though the intelligence is strong enough to be a concern. I agree with Elon Musk and some others on this and don’t understand why some people are not concerned.” Bill Gates
OK, I can’t resist telling you what I think of AI, so here it is: I can’t help but think about what this means to us. Will machines take over the world? Are we in danger of being wiped out by killer robots? Please, of course not. At least not in the next several hundred years, I don’t think. But, there is real danger of people being controlled and manipulated by misinformation and social media. And this is already happening. Just think of recent political developments and how they started. Now, I don’t want to start a political discussion here, but think about fake news which can be produced and tailored for the target groups they are addressing. It is fairly easy to manipulate people into voting against their interests that way and this is where we need to take up responsibility and fight against misuse of artificial intelligence technology that we originally designated to the greater good.
Bojan Božić has worked on Semantic Web, Machine Learning and Data Science technologies for over 7 years. He was a software engineer in Speech Recognition at Philips Speech Recognition Systems, a Scientist at the Australian Institute of Technology and a Lecturer at the University of Applied Sciences Burgenland and University of Vienna. Linkedin.com/in/Bojan-boz
ic, Website: homepage.univie.ac.at/Bojan.bozic