Hi there, I’m Rachyl Jones, a fellow specializing in tech reporting. The notion of “hallucination” has transformed recently due to the widespread availability of artificial intelligence models. While traditionally linked to human experiences, AI can now “hallucinate,” presenting inaccurate information confidently. This evolution in meaning could significantly impact decision-making processes as AI technology becomes more integrated into various aspects of life.
The magnitude of this issue may be more substantial than commonly perceived. Previous reports by the New York Times highlighted instances of hallucinations in popular AI models, ranging from 3% in OpenAI’s ChatGPT to a notable 27% in Google’s PaLM.
A recent survey indicates that the problem is even more prevalent than expected. Findings from a survey by the ML observability platform Aporia reveal that 89% of machine learning engineers working with generative AI notice signs of hallucination in their models. This high percentage underscores the concerning reality of delusional computing in the AI landscape.
Moreover, the survey suggests that 93% of engineers face model-related issues on a daily or weekly basis, emphasizing the pervasive nature of the challenges encountered in this field.
While Aporia, as a provider of observability tools, may have a bias on this topic, the survey results, based on feedback from 1,000 engineers across industries, indicate a potential escalation of these issues into everyday life.
Discussions at the World Economic Forum in Davos, Switzerland, underscore AI as a central focus for policymakers and experts. They concentrate on addressing the opportunities and risks linked to AI, including its implications on elections, misinformation, and economic progress. Given the high stakes associated with AI hallucinations, it is crucial not to underestimate the impact of these challenges.
NEWS HIGHLIGHTS
1. Drone Technology Advancements: Alphabet’s Wing introduces a new range of drones capable of carrying heavier loads, scheduled for launch in cities where Wing operates over the next year.
2. Amazon’s AI Shopping Assistant: Amazon reveals a generative AI tool that promptly responds to customer inquiries about specific products by extracting information from listings and reviews.
3. Mathematical Problem-Solving AI: DeepMind, Google’s AI lab, unveils a product proficient in solving geometry equations, potentially offering valuable problem-solving approaches for various AI applications.
QUOTE OF THE DAY
“Technology may be disrupting our traditional revenue, but it offers us many new opportunities to reach audiences and deliver the kind of quality journalism they will pay for.” – CNN Worldwide chief Mark Thompson on adapting strategies to engage with generative AI and build digital subscription-based products.
RECENT DEVELOPMENTS
- Sam Altman reflects on his departure from OpenAI, emphasizing the unexpected nature of the situation.
- Microsoft CEO Satya Nadella’s success in the realm of AI, with a focus on the implications for other players.
- Bessemer advocates for a reevaluation of late-stage company valuations amid investor emphasis on cash flow.
- Epic Games challenges Apple’s policies regarding external payments and compliance with court rulings.
GOOGLE’S ADJUSTMENTS
Google announces changes in its European search results to comply with the Digital Markets Act, ensuring equal treatment of rival services and removing specific Google features like Google Flights from search pages.
