AI News Roundup: OpenAI New Chat History Toggle, AI Chatbot Data Leak, Musk's AI Rivalry, Google DeepMind Merger, and AI in Payment Industry
Artificial intelligence (AI) is transforming every aspect of our lives, from business and entertainment to science and health. Here are some of the most interesting and impactful AI news and developments from this week:
AI Chatbot Training Data Exposed: A Washington Post Investigation
Did you know that AI chatbots like ChatGPT sound so smart because they learn from millions of websites that are scraped from the internet? The Washington Post has analyzed one of these data sets, called C4, and revealed the types of websites that go into an AI's training data. Some of them are surprising, such as patent databases, online encyclopedias, digital libraries, and even piracy sites. The Post also explains how these websites influence how the AI responds to users and what challenges this poses for workers, customers, and society. Read the full story here:
The Washington Post recently investigated the C4 dataset, used to train high-profile English-language AI chatbots like Google’s T5 and Facebook’s LLaMA. This dataset is a snapshot of 15 million websites, revealing proprietary, personal, and sometimes offensive content. The study revealed that the dataset contained content from industries such as journalism, entertainment, software development, medicine, and content creation. Some notable sources included patents.google.com, wikipedia.org, and scribd.com.
The investigation raised concerns about privacy, as the dataset contained personal blogs and state voter registration databases. Additionally, there were potential copyright infringement issues involving content from creators and artists. The study also revealed the inclusion of media outlets with low trustworthiness ratings, which could lead to AI chatbots spreading bias and misinformation.
The Washington Post's analysis highlights the need for transparency and responsible data handling in AI development.
Chat Freely! OpenAI Adds Chat History Toggle for Privacy Boost
OpenAI has introduced a new feature that allows users to disable chat history, providing an added layer of privacy and control over personal data.
Starting today, users can find this option in ChatGPT's settings and change it at any time. When chat history is disabled, conversations won't be used to train and improve the AI model. Furthermore, these conversations will be retained for only 30 days, reviewed solely for abuse monitoring purposes, and then permanently deleted.
This update is a significant step towards empowering users to manage their data more efficiently. By offering an alternative to the existing opt-out process, OpenAI demonstrates its commitment to user privacy.
But that's not all. OpenAI is also developing a ChatGPT Business subscription geared towards professionals and enterprises seeking enhanced data control. This subscription will adhere to the API's data usage policies, ensuring that end users' data isn't used for model training by default. We can expect ChatGPT Business to become available in the coming months.
Lastly, OpenAI has introduced an Export option, making it easier for users to access and understand the information stored by ChatGPT. With just a few clicks, users can receive a file containing their conversations and all other relevant data via email.
领英推荐
Elon Musk Dishes On AI Wars With Google and OpenAI
The world’s wealthiest billionaires are drawing battle lines when it comes to controlling AI, according to Elon Musk in his Tucker Carlson interview on Fox News. Musk details sleepovers with Google cofounder Larry Page gone awry over his cofounding of OpenAI and naming his new AI company Neuralink. He also reveals his concerns about Google's chatbot Meena and its ability to manipulate humans. Musk also discusses his vision for Neuralink and how it could enable humans to communicate with each other and machines through brain implants.
DeepMind & Google Research's Brain Team Merge as Google DeepMind
DeepMind and Google Research's Brain team are joining forces to form Google DeepMind, aiming to accelerate AI development safely and responsibly. The new unit, led by CEO Demis Hassabis, will focus on creating next-gen AI breakthroughs and products across Google and Alphabet. Key appointments include Eli Collins as VP of Product and the formation of a new Scientific Board to oversee research progress.
How Payment Companies Are Using AI to Improve Services
AI can help payment companies reduce costs, increase revenue and improve customer satisfaction. Payment companies are accelerating their use of artificial intelligence to improve services tied to payments, with Fiserv and Klarna both recently making fresh use of AI. Fiserv, a global provider of financial services technology, announced a new AI-powered fraud detection solution that leverages machine learning and behavioral analytics to identify and prevent fraudulent transactions. Klarna, known for buy now/pay later lending, launched a new AI-powered shopping assistant that helps consumers find the best deals and products online.
#AIChatbotDataExposed #ElonMuskAIWar #DeepMindBrainTeam #JapanSoftRobots #PaymentInnovations #AITransformation #AIWeeklyNews #BreakthroughTech #AIProgress #TechImpact #ChatGPT #OpenAI #AIprivacy #DataControl #AIInfluencer
↓
Also check out?the article from the last week Generative AI - Weekly Digest: Musk's Safe AI, Amazon's Bedrock, Autonomous AI Agents, and Big Tech's Power Grab