Highlights of the February 28 edition of the HBK Risk Advisory Services webinar series hosted by William J. Heaven, CPA/CITP. CISA, CSCP, Senior Director, HBK Risk Advisory Services.
About 55 percent of people are using AI at home or at work.
Definition of AI: The simulation of human intelligence processes by machines, especially computer systems. These processes include learning (the acquisition of information and rules for using that information), reasoning (using rules to reach approximate or definite conclusions), and self-correction. AI encompasses a broad range of techniques, algorithms, and methodologies aimed at enabling machines to perform tasks that typically require human-like intelligence.
- Key aspects of AI from the definition are “learning,” “reasoning,” and “self-correction.”
- In essence it’s training machines to think like humans.
- Need to understand where the AI tool is getting its information, as it can be biased or inaccurate.
- Biggest takeaway is to test it and check up on it and not trust it as a know-all, be-all.
Four categories of AI (including generative and non-generative):
- Functionality: Everyone, aware of it or not, has been exposed to Narrow AI, like smart speaker systems that answer questions or perform tasks like turning on a radio.
- The systems are designed and trained for a specific task or set of tasks.
- Learning (non-generative AI): A great deal of promise and potentially the biggest area of benefit.
- Analytics, for example, looking at healthcare records for a diagnosis, or where to invest money
- Supervised learning: training machines through labeled data
- Unsupervised learning: reviewing how machines conclude information from data
- Reinforced learning: allowing machines to learn by interacting with the environment independently
- Techniques and Applications (generative AI): Includes content creation, like ChatGPT, the most common use of AI to date, and:
- Machine learning: machines learning from data and improve their performance over time without being explicitly programmed
- Deep learning: reviewing how machines conclude information from data
- Natural language processing: a machine learning technology providing computers the ability to interpret, manipulate, and comprehend human language
- Reinforced learning: machines learning by interacting with the environment independently
- Industry or Doman: Industries using AI the most include healthcare, finance, retail, manufacturing, and automotive industries.
- Will AI cost jobs, replace workers? Likely more job shifting than lost jobs.
- Sectors like retail and fast foods allow customers to use kiosks to order, but still need people to create and serve the orders.
- Chatbots are doing customer service in retail environments, such as Amazon recommending products based on what you buy.
Current Trends
- AI is experiencing explosive growth.
- Most companies are not mitigating AI risk: biggest risk is inaccuracy.
- Most companies expect significant business interruption due to AI.
- Companies are not prepared for widespread AI use.
- Use in cybersecurity: to detect and respond to threats in real time; analyze patterns and predict potential security breaches.
- Use in addressing language barriers: improving language understanding, including translation
- Use in healthcare: to improve medical image analysis, to diagnose, to develop treatments
Who is using generative AI?
- Marketing: to craft first drafts, personalize marketing (knowing what the customer is buying and making recommendations), summarize text documents
- Product and service development: to identify customer trends, draft technical documents, create new product designs
- Service organizations: using chatbots for forecasting service trends or anomalies, creating first drafts of documents (the most common use)
Benefits of AI
- Streamline tasks and eliminate busy work
- Collect documents
- Analyze data
- Enhance customer service via chatbots
- Threat detection
Common uses of AI in business
- Generating template policies/procedures tailored to the organization
- Configuring action to be taken if a specific event occurs
- Researching events
Risks that come with using AI
- Inaccuracy
- Ownership of data: have to understand where data is coming from and who owns the data; careful not to disclose proprietary information
- Source of data: can be biased
- Cybersecurity
- For regulated industries, concerns about violations related to sharing information, like healthcare information
AI initiatives with negative outcomes
- Netflix wanted to build a recommendation system for customers, but was using customer data, which resulted in a class-action lawsuit alleging violation of fair trade laws.
- Google Photos misidentified people as gorillas.
- Microsoft’s chatbot Tay for Twitter had to be shut down after posting offensive and inappropriate tweets learned from negative interactions.
- Equifax breach in 2017 was due to AI not properly applying an available vulnerability patch; one of the biggest data breaches in history.
- WannaCry ransomware attack in 2017: Bots were misinterpreting WannaCry and blocking traffic.
- In 2018, Amazon used AI to screen applicants and gender bias eliminated female candidates.
- Uber’s self-driving car struck and killed a pedestrian it failed to recognize.
- Facebook chatbots developed their own language which couldn’t be understood by humans.
- February 2024: ChatGPT responds in gibberish, making nonsense replies to questions for several hours.
Final note: AI is in its infancy, so use it cautiously. Remember that inaccuracy is the biggest risk.
Speak to one of our professionals about your organizational needs
"*" indicates required fields