AI Current Events: Breaking Down the Latest in AI News and Legislation

It’s time for your next AI news update and familiarization!

Arguably the most significant innovation to emerge in February from the AI sphere is Sora. Sora is an AI-generated video program from the makers of ChatGPT. While this technology is still in its infancy, many impressive videos are being created solely from written prompts. Please take some time to check out the Sora information page to view some of the visually stunning videos and learn about some of the limitations of the current state of this remarkable technology.

Below you will find summaries of important AI news as well as links to the original articles.

EU countries give crucial nod to first-of-a-kind Artificial Intelligence Law

February 2nd, 2024 (Updated February 15th, 2024)

• EU ambassadors unanimously approved the world's first comprehensive rulebook for Artificial Intelligence, following a political agreement reached in December.
• The AI Act, aimed at regulating AI based on its potential to cause harm, faced technical refinement for over a month after the agreement.
• Reservations among member states were lifted with the adoption of the AI Act on 2 February, although some countries initially opposed parts of the deal.
• France, Germany, and Italy sought a lighter regulatory regime for powerful AI models but compromised with the European Parliament's demand for hard rules.
• France agreed to support the text "with strict conditions," aiming to balance transparency and protection of trade secrets.
• EU countries will have room to influence the implementation of the AI law, as the Commission will issue around 20 acts of secondary legislation.
• The European Parliament's Internal Market and Civil Liberties Committees are set to adopt the AI rulebook on 13 February, with a plenary vote provisionally scheduled for 10-11 April.
• The AI Act will enter into force 20 days after publication in the official journal, with bans on prohibited practices starting after six months and obligations on AI models starting after one year.

OpenAI is adding new watermarks to DALL-E 3
February 6th, 2024

• OpenAI's image generator DALL-E 3 will add watermarks to image metadata following standards from the Coalition for Content Provenance and Authenticity (C2PA).
• Watermarks will be applied to images generated on the ChatGPT website and the DALL-E 3 API, with mobile users receiving them by February 12th.
• Watermarks will include both invisible metadata and a visible CR symbol in the top left corner of each image.
• Currently, only still images can carry the watermark, not videos or text.
• The C2PA, including companies like Adobe and Microsoft, promotes the use of the Content Credentials watermark to identify content provenance, distinguishing between human and AI-made content.
• Meta (formerly Facebook) announced plans to add tags to AI-generated content on its social media platforms.
• While identifying AI-generated content aligns with directives from the Biden administration, watermarking is not foolproof against misinformation as metadata can be easily removed, especially on social media platforms.
• OpenAI emphasizes the importance of adopting methods to establish provenance and educating users to recognize these signals to increase the trustworthiness of digital information.

Google Expands AI Training Courses & Launches AI Fund for Businesses

February 12th, 2024

• Google is expanding its Google Certificates Program by adding more resources and AI training courses.
• Learning materials will be available in 18 additional languages, aiming to make AI education accessible to millions more individuals.
• The company is investing €25 million in the AI Opportunity Initiative for Europe to equip workers with the skills needed in an AI-powered economy.
• Google plans to add new resources to its Career Certificates program, allowing learners to discover how AI is used in various fields and gain hands-on experience.
• While many AI foundational courses in Google's library are free, learners pursuing Google Career Certificates may need to pay a monthly fee on platforms like Coursera.
• Google also offers funded scholarships for certificates across Europe, the Middle East, and Africa, allowing learners to access AI training courses for free.
• Learners can explore other free AI training courses online to develop specialized skills in areas like coding, advertising, and prompt engineering.

Tech giants including Meta, Microsoft, Google and OpenAI are working on a pact to jointly crack down on AI content intended to deceive voters ahead of crucial elections around the world this year

February 13th, 2024

• Tech companies are negotiating an "accord" on deepfakes and other deceptive content, set to be announced during the Munich Security conference.
• Meta, Google, Adobe, Microsoft, OpenAI, TikTok, and others are collaborating on the accord to combat the deceptive use of AI targeted at voters.
• The companies will work on developing ways to identify, label, and control AI-generated images, videos, and audio aimed at deceiving voters.
• They plan to use a common watermarking standard to tag images generated by their AI applications, such as ChatGPT, Copilot, or Gemini.
• Recent examples of convincing AI deepfakes, including a robocall impersonation of U.S. President Joe Biden and AI-generated speeches by Pakistan's former prime minister Imran Khan's party, have heightened concerns.

Nvidia’s Chat with RTX is a promising AI chatbot that runs locally on your PC

February 14th, 2024

• Nvidia is releasing an early version of Chat with RTX, a demo app that runs a personal AI chatbot on your PC.
• The app allows users to feed it YouTube videos and documents to create summaries and get relevant answers based on their own data.
• It runs locally on a PC and requires an RTX 30- or 40-series GPU with at least 8GB of VRAM.
• Chat with RTX can handle YouTube videos by searching transcripts for specific mentions or summarizing entire videos, but there are some bugs in the early demo.
• It can be useful for data research, such as analyzing collections of documents or fact-checking data.
• Installation takes around 30 minutes and the app is nearly 40GB in size, with some limitations and known issues.
• While not a polished app, Chat with RTX demonstrates the potential of AI chatbots running locally on PCs for tasks like data analysis without the need for cloud-based services.

Treating a chatbot nicely might boost its performance

February 24th, 2024

• Phrasing requests nicely or meanly can yield better results with chatbots like ChatGPT, suggesting emotive prompts influence AI behavior.
• Studies by academics and vendors reveal that generative AI models perform better when prompted with urgency or importance.
• Emotive prompts can manipulate a model's underlying probability mechanisms, leading to different outputs.
• However, emotive prompts can also be exploited for malicious purposes, potentially causing models to ignore safeguards.
• The reasons behind the success of emotive prompts remain unclear, but objective misalignment and mismatches between training data and safety datasets are possible factors.
• Despite the potential for manipulation, understanding and using emotive prompts effectively remains a challenge, with continued research needed to improve model understanding and context.

NVIDIA RTX 500 and 1000 Professional Ada Generation Laptop GPUs Drive AI-Enhanced Workflows From Anywhere

February 26th, 2024

• NVIDIA is introducing the new RTX 500 and 1000 Ada Generation Laptop GPUs for powerful AI-accelerated mobile workstations.
• These GPUs, based on the Ada Lovelace architecture, include both a neural processing unit (NPU) and NVIDIA RTX GPU for AI processing, delivering up to 682 TOPS of AI performance.
• The RTX 500 GPU offers significant improvements in generative AI performance, photo editing speed, and graphics performance compared to CPU-only configurations.
• These GPUs enable professionals to streamline tasks such as video editing, graphic design, and video conferencing with AI effects while on the go.
• NVIDIA also offers RTX 2000, 3000, 3500, 4000, and 5000 Ada Generation Laptop GPUs for advanced rendering, data science, deep learning, and development projects.
• The RTX 500 and 1000 GPUs will be available in mobile workstations from Dell, HP, Lenovo, MSI, and other global manufacturing partners in spring.