Artificial Intelligence is one of the biggest buzzwords of the 21st century. It’s affecting how we work, play, shop, and stay safe. These astounding pieces of technology are largely misunderstood, however. In the same way with cars we know a Smart Car isn’t the same as a Ferrari, two pieces of A.I software marketed to the same demographic may differ even more, yet some people still think all A.I is the same.
In this article we’ll go through some of the differences in Security-focussed A.I software terms you may hear thrown around in sales dialogues, to help you know if you’ve actually got the right A.I in your software to match your needs and expectations. Untangling yourself from the web of tech mambo-jumbo, we hope this article also helps you save time, money, and stress.
What is A.I in security (in general)?
In security, Artificial Intelligence is used in a number of ways. The most common is utilising its’ capacity to alert a human operator to pre-set patterns of activity in an attempt to prevent crime. Examples of this is to set a ‘virtual boundary’ where an alarm will be raised if anything crosses the A.I-set path or area of monitoring. A.I is also used in tracking vehicles of interest using Licence Plate Recognition (LPR). These LPR cameras use A.I to analyse and compare licence plates viewed with licence plates stored in numerous databases held by the police or private sector (ensured to be POPI compliant).
What is the difference between old-school and new-tech A.I in security?
Traditional:
The way traditional A.I is used (things are moving so fast there can still be older ‘traditional’ processes in an emerging market), is basic pre-set pattern recognition to relieve the drudgery and fatigue human labour would struggle with.
Traditional A.I security strategies, if not backed up by off-site security verification, are hardly better than a motion detection spotlight. Temporarily shines a light on the situation, but if nobody is there to follow up or see it, it lacks effectiveness.
New-tech:
But A.I can do so much more than the basic use most people are familiar with. Advanced A.I is no longer typically housed in servers away from the action, subject to load-shedding or communication issues, but instead housed on the ‘edge’, I.e. in the CCTV camera or device itself!
Modern A.I utilises deep learning to increase the scope and accuracy of detection over time. Along with integration with measures like LPR and IoT devices, modern approaches to A.I can greatly expand the overall clarity of your site or property.
What keywords or phrases should I know/ask for?
Edge? Deep learning? What terms should I ask for when talking to a supplier? And what’s the difference? We’re glad you asked. Here are some common terms, their counterpart, and a quick description to help you make informed decisions.
Key differences:
Machine learning vs deep learning.
Machine learning utilises pre-written/trained instructions, fixed and strictly defined, to help notify an off-site monitoring operator of an anomaly. The problem here is that the algorithms are only as effective as their training. (built-in classification library, trained on an initial data set) They may be good at what they detect, but if they don’t get trained well they will never be able to ‘learn’ over time and self-improve.
Deep Learning starts like Machine Learning with pre-trained instructions, as strictly defined and accurate, yet it can ‘learn’ based on what criteria the human operator needs and selects for. This form of A.I allows new patterns to be formulated (algorithms) after initial training, and therefore becomes more accurate over time, and can be trained to have a wider scope than Machine Learning.
Smart software:
Beware of blanket terms such as ‘smart’ everything. ‘Smart’ devices mean they have embedded A.I capabilities, either machine learning or Deep Learning. When we use the term ‘Smart’, it means we can utilise all manner of A.I powered devices. If someone is selling you on the ‘smart’ aspect of the device, don’t forget to ask which of the above it has.
Edge vs cloud-based
Edge-based A.I is Artificial Intelligence software housed in the physical device. There are numerous benefits, such as potentially faster alert times as the time between initial alert and acceptance/processing by A.I is almost instantaneous. The uploaded file sent to off-site monitoring companies can also be much smaller as the device doesn’t need to send continuous video to off-site A.I, only a notification of an alarm for the operators to process. They would then dial into the camera if needed.
Cloud-based A.I means the Artificial intelligence software isn’t housed in the device itself, but rather off-site. There are drawbacks to this type, as continuous signal to an external A.I system needs to be maintained, otherwise the system is effectively without A.I. The benefits of this form of A.I however, are that the cost of regular CCTV cameras is less than edge-based CCTV devices, and that existing security/CCTV systems without existing A.I can (often) be enhanced with A.I after market.
For most of our sites, we utilise a combination of edge and cloud-based A.I systems to achieve optimal monitoring for an optimal cost.
Additional terms:
IoT: Internet of Things. Essentially there are devices that can be integrated with off-site monitoring platforms to monitor and alert specific things, such as a freezer system malfunctioning or a generator system not kicking in.
LPR: Licence Plate Monitoring, used to track suspicious vehicles, or to alert wanted vehicles when entering a premises. Useful for early warning!
Algorithms: In the case of off-site A.I security, these are patterns of code which determine if something is within the accepted ‘rules’ or not. If something is unusual it will flag it to an off-site monitor.
Biometrics: Elements of a person’s unique biology that allow a security system to grant entrance or not. Examples include fingerprint access control, facial or retina scans, and much more.
Integration: Integration refers to a device’s ability to connect and communicate seamlessly with another device or software. Some pieces of hardware (cameras etc) may need a bridging software to connect to an external software. Our VCC platform seamlessly integrates with most brands of devices, and models old or new.
Why does this matter to me?
Costing issues 1: Are you being over-sold on under-spec tech?
Are you being sold ‘smart’ devices that aren’t as smart as you think they are? Or the opposite, are you being over-quoted with the latest and greatest devices for a back alley nobody ever uses when a simple device with entry-level Machine Learning will do?
Costing issues 2: Are you getting the best bang for your buck? Can the old technology integrate properly? Early warning etc? IoT integration?
Are you being correctly advised on the best types of A.I system for your property type? Off-grid scenarios may benefit from edge-based A.I devices, whereas commercial sites in the city with generators and backup systems may only need cloud-based A.I. Are you being quoted for devices made by reputable brands? If we haven’t heard of a brand, it’s likely poor quality as we independently work with all major security device brands.
How can I make sure I’m getting the right technology for my site?
Do a bit of background research: a bit of knowledge and a few terms dropped in conversation also ensures you are taken seriously in negotiations, and vice versa to see if the person you are dealing with knows their stuff. (If you’re reading this article, I take it this point may not apply to you, but I thank you for making it this far!)
Ask questions: it’s your site, your money, and your rules. Ask questions even if you don’t know the answer. The way a supplier answers basic questions can be an indicator in itself.
Get multiple quotes: one of the best ways of ensuring you get the best bang for your buck is to speak with two or three suppliers and get quotes with costing. Make sure it’s tax included for accuracy. You don’t need to ask a ton of suppliers, but a few will give you a good picture.
Ask about suppliers: if you would like a list of suppliers we work well with, or we have pre-vetted, feel free to ask us.
Think about reputable brands: Do a background check on the names of popular and reliable brands. If they’re not quoting you for one of these, we can advise you further on whether the brand is reliable.
Ask us for help: Our consulting services can help you with transparency and streamline connections between yourself and reputable suppliers we work with. We want to make your off-site security experience as pleasant as possible!
Conclusion:
With a bit of knowledge, and a crash-course on the difference between certain terms, we hope you will feel more comfortable speaking to installers, suppliers, or anybody else in the off-site security industry. Should you wish to discuss the use of A.I for your security system further, we welcome a call! Our talented and knowledgeable sales and technical support team are more than happy to help guide you along your A.I security journey!
Comments