Second Cyberattack on CDK Global
CDK Global, a data provider whose software is integral to operations at 15,000 auto dealerships, experienced a second cyber incident that severely hampered operations on Thursday across dealerships in the US and Canada. According to a message sent to dealers, the incident caused operations to slow to a near-standstill. CDK Global offers a variety of software products that auto dealers rely on for managing tasks such as recording negotiated deals, scheduling, and service communications. While not all dealers use CDK’s products and some use them selectively, the system shutdown has impacted many dealerships. Some dealers interviewed by CNN mentioned difficulties accessing previously negotiated customer agreements, complicating sales closures. These agreements are crucial as they encompass more than just the car’s price, including rebates and incentives that require customer qualification. Beyond car sales, disruptions have extended to parts and service operations as well.
Read More: CNN
Spotting AI ‘hallucinations’
AI models can generate inaccurate responses, known as ‘hallucinations,’ when uncertain about answering a query. Researchers have developed a new method to detect when generative AI is likely to produce such hallucinations—fabricating information due to a lack of knowledge—and prevent these incidents. A recent study by University of Oxford researchers introduced a statistical model capable of identifying situations where a question posed to a large language model (LLM), commonly used in generative AI chatbots, might result in an incorrect response. Hallucinations are a significant concern with generative AI models due to their advanced technology and conversational abilities, potentially presenting fabricated information as factual in response to queries. As more students utilize generative AI tools for research and assignments—activities these models are marketed to assist with—many industry experts and AI scientists advocate for increased measures to address AI hallucinations, especially concerning medical or legal inquiries.
Read More: Independent
Groundbreaking test for Parkinson’s disease
Using artificial intelligence, scientists have pinpointed a biological signature of Parkinson’s that they believe could pave the way for a straightforward blood test capable of detecting the condition at least seven years before symptoms manifest. Researchers from University College London and University Medical Centre in Goettingen, Sweden, employed machine learning—a type of artificial intelligence—to analyze blood samples from individuals with Parkinson’s. They identified eight key proteins, or “biomarkers,” consistently present in those with the disease. The team then applied their machine learning tool to study blood samples collected a decade earlier from people diagnosed with Rapid Eye Movement Disorder, a condition linked to a high risk (about 75%) of developing Parkinson’s later in life. Impressively, the AI accurately predicted which individuals would go on to develop Parkinson’s, up to seven years before their initial symptoms appeared. Dr. Michale Bartl from UMC Goettingen emphasized, “By assessing eight specific proteins in the blood, we can identify potential Parkinson’s patients several years ahead of time. This early detection could enable earlier initiation of drug therapies, potentially slowing down disease progression or even preventing it altogether.” The researchers are now focused on validating the test’s reliability and refining it for future clinical use.
Read More: Sky News