Once in a while, an evolution will occur in the world of technology that has the potential to change everything else from that point onwards. Some key revolutions we can look back on are Henry Ford’s introduction of the modern moving car assembly line in 1913, the rise of television advertising throughout the 1950s, or the birth of e-commerce in the wake of the 90’s dot-com boom. The next evolution is already occurring and it’s being driven by Artificial Intelligence (AI).
The concept of AI is by no means a new one. The notion of artificial beings having intelligence can be traced as far back as classical Greece. However, modern AI started to impact our daily lives in the early years of the 2020s. AI is beginning to change many things we do every day, whether we know it or not, and we’re still at the dawn of this revolution.
Different types of AI
One area that has supercharged the impact AI has had on our daily lives is Generative AI but that’s only a small part of the puzzle. While some of these concepts might not sound familiar, the way they’ve been employed into real life will be recognisable.
Narrow AI
Narrow AI is designed to perform simple tasks as they are and nothing else. There is no learning element with Narrow AI – its purpose is around understanding queries and providing relevant information. Some examples of Narrow AI include voice assistants like Alexa and Siri. Narrow AI can make sense of your voice commands and provide an appropriate response.
Reactive Machines
Reactive Machines are designed to be task-specific to deliver consistency based on predefined rules. One example is IMB’s chess-playing Deep Blue AI. Deep Blue understands chess and will use the rules to make specific moves within predefined parameters, but only when the other player has taken their turn.
Limited Memory AI
Limited Memory AI is defined as AI that is trained to learn from data and generally improve over time based on its experience. This is a similar way of learning to the human brain’s ability where neurons connect to one another. Self-driving cars are good examples of this form of AI as they can collate data from the driving conditions, assess them, and ultimately make an informed decision about the route it should take, the safest speed etc, all while avoiding other cars or pedestrians and obeying the rules of the road.
Generative AI
Generative AI is a category of Narrow AI that is trained from large subsets of existing data to create unique outputs. It can create new content resembling a human-created output, such as images, text, or music. These models learn patterns from existing data and generate unique examples based on those patterns. ChatGPT, Jasper Chat, Gemini and Midjourney are all examples of Generative AI.
The world is changing with AI
You don’t have to look far to see where AI is being harnessed to make our lives easier and businesses more efficient. There are a variety of tools that are now publicly available and are already changing how a lot of different jobs are executed:
Industries are embracing AI in exciting ways
The tools listed above give some insight into how individuals can employ AI to help them with existing roles. But on a wider scale, AI is revolutionising entire industries, too.
Healthcare
One specific area that has shown promise in applying AI is within healthcare, in particular with patient diagnosis and treatment. Similar to how healthcare professionals build up their experience from practical learning and experience, AI algorithms can use patterns and rules to teach themselves how to diagnose and treat illnesses.
One fascinating example comes from object detection technology and AI deep learning known as Computer Vision (CV). CV brings optical sensors and cameras together with AI in order to allow technology to perceive people in the same way that humans do. CV can be used to improve image analysis for diagnosing illnesses from medical images such as X-rays and MRI scans. It can even be applied to detecting breast and skin cancer. Google have gone so far as to develop their own app to assist healthcare professionals in detecting anomalies early.
Finance
The finance industry is often on the cutting edge when employing technological innovations, and the use of AI is no different. One area where banks need to stay highly innovative is with fraud. AI is being utilised to assist in fraud detection through machine learning models. These models can be trained to recognise individual customer habits and behaviours. The more data the models can get on the habits of a particular customer, the better it can become at predicting specific behaviours and spotting anomalies.
Should there be any kind of unusual behaviour, real-time fraud detection systems can pick it up, and both the customer and the bank will be alerted. Other common fraud detection systems used by financial institutions include biometric verification. Many new smartphones have facial recognition and liveness detection that apply to banking apps to ensure that the person is the account holder. AI-powered facial recognition is crucial in stopping identity fraud via deepfakes technology.
Customer service
In the last decade, we have witnessed the rise of chatbots to assist with customer service enquiries. In their infancy, some chatbots used preloaded responses to funnel queries to appropriate customer service agents. Thanks to AI and its ability to now understand the nuances of individual queries, the capabilities of chatbots as a customer service tool for all types of businesses have become even better than they were before.
Given AI’s ability to perform repetitive tasks, things like frequently asked questions and personalised instant support can now be offered by businesses via chatbots. This is especially valuable to small businesses given the increasing customer expectations to get instant support. The advantages apply to larger businesses as well. If they are taking on customer queries at scale, they can employ a chatbot to quickly deal with simple requests, taking the burden away from their existing workforce, regardless of the type or size of business they’re engaging with. Customer service leaders agree, as 73% believe that all consumers will expect AI-assistance in their buying journey within the next five years.
Food production
Farming is an industry that exists in near constant evolution. The last few decades have been challenging for agriculture and the industry has been quick to explore different ways that AI can take on some of the human burden.
One highly specific example comes from Japan where the son of a cucumber farmer employed machine learning to identify and categorise the crop of the farm. Cucumbers that are straight, vividly coloured and thick are considered premium grade, fetching much higher prices. Before AI was employed, it was largely the job of the farmer’s wife to sort the cucumbers. This would sometimes take up to 8 hours a day during the harvest season. Automating this process by implementing a machine learning system to analyse and categorise the cucumbers is saving this single farm hundreds of hours of time each harvest.
Preserving nature
The decline of the global bee population is an area that many environmental organisations are working hard to solve. Bees are vital to the pollination of plants and are therefore a crucial element of the global eco-system and our food supply. AI is helping to surface and understand data that can help us to help secure the future survival of bees.
The World Bee Project remotely monitors the behaviour of bees, globally. Data is gathered from individual hives, tracking everything from hive temperature and the sound levels. Once collected, the data is then analysed by AI to scan for trends and patterns in the behaviour of the bees and to help determine if they are behaving in a way that suggests that their colony is in jeopardy. The World Bee Project can step in and ensure they’re able to survive.
AI and GDPR regulations from around the world
The deployment of AI systems are moving faster than legal regulations. Around the globe, various governments and authorities are working to create legal frameworks to contain the technological advances we’ve seen over the past decade. Some of the most significant changes are reflected in this timeline:
Source: https://www.mindfoundry.ai/blog/ai-regulations-around-the-world
Cyber risks posed by AI use
There is no shortage of insight that tells us how the threat landscape has been changed through AI. While we can see advantages to using AI to streamline and build efficiency, we can’t discount the growing threat. A recent report published by Lloyd’s focussed on what they described as “the profound impact that unrestricted generative AI models are set to have on the cyber landscape.”
Their report found that the initial impact is minimal, but that shouldn’t build complacency; “Significant impacts of generative AI technology on the cyber threat landscape have been minimal thus far thanks to industry safety measures, the effectiveness of AI model governance and current barriers in hardware. However, as generative AI and large language models become more accessible, they pose a growing risk. While 2023 was already a record-breaking year for ransomware attacks, the impact of generative AI on the cyber threat ecosystem is likely to increase the frequency and severity of smaller-scale cyberattacks. That threat is expected to grow over the next 12 to 24 months, levelling off as security technologies catch up with threat actors.”
Some of the developing threats that we are likely to see include the following:
Reconnaissance by threat actors using AI
A lot of manual work is done by threat actors to target companies, especially when it comes to network perimeter scanning. AI can make this easier by taking the strain out of repetitive tasks like gathering data on key corporate officers from published directories, then targeting them through personalised content and more effective phishing lures, etc.
Polymorphic Malware Generation
Polymorphic malware is notoriously hard to contain because it can alter itself while maintaining its original function. This type of malware generation was once the domain of highly-skilled threat actors, but the ubiquity of generative AI means this type of code can be created and deployed for clandestine purposes more easily.
Zero-Day Identification
The ability of AI to analyse large volumes of data and subsequently identify patterns and anomalies presents an increased risk for threat actors to exploit zero-day vulnerabilities. While zero-day vulnerabilities are less common than malware attacks, they can significantly impact a business and should have preventative measures.
Social Engineering Risks
AI is a core element of Deepfake technology. Deepfakes can produce remarkably convincing counterfeit audio and video, enabling precise impersonation of specific individuals. Deepfakes can be exploited in scams, disseminating misinformation, and circumventing biometric security protocols, posing substantial risks to both individuals and organisations. This could be used to expose vulnerabilities in multi-factor authentication, which is increasingly becoming the norm amongst organisations.
Sharing sensitive information with AI tools
While generative AI tools like ChatGPT might be helpful for an organisation’s employees to increase efficiencies, it’s important to remember that these models learn from the data you put in. Last year, OpenAI was forced to respond to reports that a tool glitch allowed some users to see the titles of other users’ conversations. This would have huge implications if a business used these tools to parse sensitive information.
What we have mentioned above is by no means an exhaustive list of the potential risks, as the threat landscape will continue to evolve while AI does.
How can businesses defend themselves?
While the global application of AI in business might seem daunting, there’s no denying that businesses can create efficiencies and find competitive advantages through using AI. The risk presented by threat actors is also worth careful ongoing consideration. One effective method is keeping an eye on some of the big profile data breaches that have occurred due to AI exploiting other businesses’ vulnerabilities.
A publicly known cyber-attack outlines the attack path used to compromise a business and can be used by internal teams to ask whether their business has the same exposure from an external attack surface point of view. If an attack path is shared online, a business can review and see if the same controls are in place to prevent that attack path from being carried out.
For a business owner to fully understand what they are using from an AI perspective, they should talk to experts about the risks when implementing tools and keep up to date with security issues. Make sure you’re having ongoing conversations with your clients, that they’re fully aware of what they’re getting into and their employees are aware of the risks.
We can see that AI continues to have a tangible impact on both our daily lives and how your client’s businesses are run. The rewards and risks of AI adoption come in equal measure. For every process that is streamlined with support from AI, there’s a development to the risk landscape and what can be exploited by threat actors, potentially putting your clients at risk.
Having cover from cyber experts is therefore crucial. Brit’s policyholders can use the outside in scanning tool and unlimited virtual CISO service with Datasafe to understand and mitigate the cyber risks to their business. Get in touch with our cyber team to find out more.
