Narrow encounter with AI-generated voice, scammers leave family with a scare
The Trapp family narrowly avoided an unsuccessful attempt to scam them. Recently a surge in in AI-generated voice scams has resulted in major concern for the safety of families. The Trapp family almost fell victim to this predicament when a replicated voice of their son called them claiming that they had been detained on charges of assault and reckless driving. The replicated voice of a police officer subsequently demanded bail of $15 000 for their son's freedom
Thankfully, the parents independently contacted their local police department and uncovered the plot to scam them with their son being safely at home.
This incident is only one of many in which the outcome may not have been as desirable as this. In light of the simplicity with which such deceptions can be carried out, doubt is cast on the potential benefits of advancements made in generative AI
Innovation in the field of LLM research
The Field of Large Language Models(LLM) has made massive strides with Meta’s new Self Rewarding Language Models (SRLM). This new method is characterized by less reliance on human preference data through the continuous update of the reward model during LLM alignment
SRLMs can evaluate new examples, adding relevant information to their training sets in the process. The results of the experiments reveal that SRLMs improve the performance of a Llama 2 70B model. Outside of its features, the SRLM is remarkable because it surpasses superior LLMs on AlpacaEval without the use of any human supervision
This innovation is a substantial contribution towards the autonomous language model improvement. It shows the potential of self-rewarding approaches in advancing artificial intelligence.
Zuckerberg's ambitions for AGI
Mark Zuckerberg has expressed his interest in venturing into the exploration of Artificial General Intelligence (AGI). Mark Zuckerberg is determined to develop AGI in light of its significance in product development. Meta is relocating its AI research group to enhance closer collaboration with generative AI product teams. Meta has set out to possess over 340,000 Nvidia H100 GPUs by 2024. Despite the current surge in AI interest, Zuckerberg describes the development as gradual.
As the AGI control debate unfolds, he leans towards an open-source approach, emphasizing Meta’s commitment to both the metaverse and AGI for improving the future of humanity.
America’s AI research facilitator: The NAIRR
Joe Biden has issued an executive order launching a pilot program named National Artificial Intelligence Research Resource(NAIRR). Multiple government agencies among them the National Science Foundation, NASA, and the Department of Defense, partner with 15 private sector entities like Amazon Web Services, Google, Microsoft, and Nvidia to democratize access to Artificial intelligence. The NAIRR provides US-based AI researchers with various resources to facilitate their research. Training, software, computation power, and more are provided fostering creativity and innovation
Microsoft contributes $20 million in Azure credits and access to models, while Nvidia provides access to its DGX Cloud and AI Enterprise software. The NAIRR pilot operates in four key areas; The enablement of diverse AI resource access, the support of secure AI research, the enhancement of interoperability between AI research, and the education of communities in AI usage. This project serves as a medium to strengthen America's position as the global innovation leader.
Opmerkingen