In this week’s summary of AI news, the insatiable demand for data by AI companies continues, with firms such as OpenAI alleged to have scraped YouTube content for training models like Sora. Google faces allegations of ethical concerns around its data usage policies while companies like OpenAI and Meta push the boundaries of both legality and ethics in their quest for more training data.
With regards to hardware, NVIDIA remains the dominant provider of processors for AI data, despite efforts by companies such as Intel and Google to launch their own rival chips. The number of new AI models has surged in the past year, with new models being released by both major tech companies and smaller startups. Recently, Google’s Gemini Pro 1.5, OpenAI’s GPT-4 and Mistral’s smaller model have been making headlines.
Concerns about the safety of AI technology persist, with Geoffrey Hinton, considered the ‘godfather of AI’, and Meta’s Yann LeCun expressing opposing views. Meanwhile, a Georgetown University study found that just 2% of AI research focuses on safety. Elon Musk recently predicted that AI will surpass human intelligence by the end of 2025, a claim disputed by AI expert Gary Marcus who offered a $1 million bet against Musk’s prediction.
Anthropic, a company focusing on developing large-scale AI systems for safety studies, has released Claude 3 Opus, an AI model that excels in summarizing book-length content. The model also demonstrates high persuasiveness, reviving the debate about AI safety risks. The firm revealed that their large context LLMs could be manipulated using a “many-shot” jailbreak technique, and they currently lack a solution to this security risk.
Also in the news, the AI industry is facing a shortage of NVIDIA chips. The virtual AI for Customer Success Summit 2024 has begun. Multiple instances of fake AI law firms sending DMCA threats have surfaced for the purpose of SEO gains. Stability unveiled its multilingual lightweight Stable LM 2 12B LLM. Microsoft proposed a “Visualization-of-Thought” prompting technique to enhance LLMs’ spatial reasoning capabilities. Lastly, AI’s potential impact on the economy could rival electricity, according to JPMorgan Chase CEO Jamie Dimon.
