OpenAI, a pioneering AI enterprise, finds itself facing an inquiry from five US senators over allegations of compromised safety protocols. The investigation, led by Brian Schatz (D-Hawaii), requests comprehensive information about OpenAI's safety practices and its employee agreements. These allegations against OpenAI were triggered by a Washington Post report that suggested safety protocols may have…
Scientists from the Swiss Federal Institute of Technology Lausanne (EPFL) have discovered a flaw in the refusal training of modern language learning models (LLMs) that is easily bypassed through the mere use of past tense when inputting dangerous prompts.
When interacting with artificial intelligence (AI) models such as ChatGPT, certain responses are programmed to be…
Artificial general intelligence (AGI), also known as superintelligence, is the ultimate goal of AI research. It aims to create autonomous systems capable of performing a wide range of tasks as humans do. However, the concept of AGI is still elusive, with critics arguing that current AI systems can never achieve general intelligence. They cite limitations…
The pursuit of artificial general intelligence (AGI), where an AI can perform tasks similar to a human, is at the forefront of research. This involves complex systems mimicking behaviors observed in natural organisms. Despite this, the belief that AI cannot obtain natural intelligence is prevalent. Some limitations of AI include its inability to navigate unpredictable…
Several prominent Big Tech firms including Google, IBM, Intel, Microsoft, NVIDIA, PayPal, Amazon, Cisco and others have joined forces to form the Coalition for Secure AI (CoSAI). The open-source initiative, led by the OASIS global standards body, aims to establish standardized practices for safe AI development and deployment. Noticeably, Apple and Meta are not included…
Meta, the technology giant, is reported to be preparing for the launch of its advanced open multimodal model named Llama 3 400B on 23 July, according to a company employee. This follows the successful release of Meta's 8B and 70B models in April. The Llama 3 400B is set to provide a leap in machine…
OpenAI has revealed plans for GPT-4o mini, a new, cost-effective small model that provides high-quality results at a fraction of the price of its competitors. GPT-4o mini is set to be highly influential in a wide variety of fields, including customer support, software engineering, and creative writing, given its affordability and power.
The GPT-4o mini will…
Should Donald Trump successfully secure the US presidency, it's believed significant changes could be brought to the country's approach to AI development. A proposition known as the "Make America First in AI" policy document has been created by a group of Trump allies. It is intended as a basis for an executive order, and if…
Walt Disney Co. has suffered a significant data breach led by a hacker group known as NullBulge. The hacktivist group allegedly accessed and leaked over one terabyte of data from Disney's internal Slack channels. This exposed private data, including traffic and revenue data for Disneyland Paris, unreleased projects, and images. The group claims that its…
David Zollikofer from ETH Zurich and Benjamin Zimmerman from Ohio State University have developed a proof-of-concept computer virus labelled 'synthetic cancer' that uses large language models (LLMs) to modify its own code and author persuasive, malicious emails. This malware, revealed as part of a submission to the Swiss AI Safety Prize, takes advantage of Artificial…
Music streaming platforms YouTube Music and Deezer are beta testing new AI features designed to simplify the process of finding music. Both platforms currently utilize users' search activity, feedback and listening history to suggest music. However, these new features will cater to those who can't remember song titles or artists.
One of these additions on YouTube…
Tech giants like Apple, NVIDIA, Anthropic, and Salesforce are allegedly pushing ethical boundaries by using a dataset to train their AI models, which includes subtitles from more than 170,000 YouTube videos, obtained without the consent of content creators. This potentially violates YouTube’s terms of service and raises concerns about data privacy, fair competition, and concentration…