Nvidia's Blackwell chips are significantly accelerating AI training speeds, outperforming previous Hopper chips by twofold, according to recent data and benchmarks. CoreWeave's Chetan Kapoor has observed a shift towards utilizing smaller chip groups for AI training. Simultaneously, concerns around AI and copyright are growing, with ministers offering concessions on copyright to avoid a defeat in the House of Lords regarding the use of copyrighted material for AI model training. The government plans to release reports on AI and copyright regulation within nine months, as stakeholders like Beeban Kidron advocate for transparency in AI training data. In the US, labor unions are actively seeking to mitigate the potential negative impacts of AI on jobs through legal restrictions and contract negotiations. Johnson is defending a 10-year freeze on state AI laws within a megabill. Deloitte's Kieran Norton emphasizes the importance of security and governance in AI deployment, highlighting AI's role in cybersecurity and the need for responsible AI usage. Google Gemini's Veo 3, an AI video generator, raises concerns about online truth due to its ability to create realistic videos. Experts are also warning about the potential dangers of an AI arms race, including dehumanization and the increased risk of AI-powered warfare. Samsung is considering integrating Perplexity AI into its devices, potentially including a pre-built app and AI assistant features. Research indicates that the way the brain uses dopamine to anticipate rewards is informing new AI adaptation methods. WPP CEO Mark Read anticipates that AI will reshape the advertising workforce, leading to job reductions but also creating new opportunities.
Key Takeaways
- Nvidia's Blackwell chips double AI training speed compared to previous Hopper chips.
- AI copyright concessions are being offered to address concerns over copyrighted material use in AI training.
- Transparency in AI training data is being pushed by stakeholders in the EU, U.S., and UK.
- US labor unions are fighting to limit AI's negative impact on jobs through legal and contractual means.
- Johnson defends a 10-year freeze on state AI laws.
- AI deployment requires strong security and governance to ensure responsible use.
- Google Gemini's Veo 3 AI video generator poses a threat to online truth due to its realistic video creation capabilities.
- An AI arms race could lead to dehumanization and increased risk of AI-powered warfare.
- Samsung is exploring integrating Perplexity AI into its devices.
- Brain dopamine research is inspiring new AI adaptation methods.
- AI is expected to reshape the advertising workforce, reducing some jobs while creating others.
Nvidia's Blackwell chips boost AI training speed
New data shows Nvidia's Blackwell chips are faster at training AI systems. The chips are more than twice as fast as the previous Hopper chips. In tests, 2,496 Blackwell chips completed a training task in 27 minutes. CoreWeave's Chetan Kapoor notes a trend toward using smaller chip groups for AI training.
Nvidia's Blackwell chips boost AI training speed
Nvidia's newest chips are improving AI system training, according to new data. MLCommons released data showing Nvidia and AMD chips training AI systems like Llama 3.1 405B. Nvidia's Blackwell chips are twice as fast as older Hopper chips. 2,496 Blackwell chips finished a training test in 27 minutes. CoreWeave's Chetan Kapoor says AI training is shifting to smaller chip groups.
Nvidia Blackwell chips double AI training speed
Nvidia's Blackwell chips are making AI training faster and more efficient. The chips reduce the number needed for training large language models like Llama 3.1 405B. Benchmarks show Blackwell chips are twice as fast as previous Hopper chips. CoreWeave's Chetan Kapoor notes a trend towards smaller chip groups for training. These improvements highlight Nvidia's lead in AI training technology.
AI copyright concessions offered to avoid Lords defeat
Ministers are offering concessions on copyright to avoid a defeat in the House of Lords. The data bill is facing opposition over AI companies using copyrighted material to train models. The government will publish reports on AI and copyright regulation within nine months. The bill aims to generate \u00a310 billion in economic benefits by updating data protection law. Beeban Kidron is opposing the government's changes to copyright law.
AI training data transparency what it means for content owners
AI companies use lots of data, including copyrighted content, to train their AI models. There's a push for AI developers to reveal what data they use. The EU, U.S., and UK are considering rules for training data disclosure. Beeban Kidron is advocating for transparency in the UK's Data Bill. Transparency could lead to licensing and lawsuits for copyright infringement.
US labor unions fight AI disruption
US labor unions are working to limit the negative effects of AI on jobs. Automation could replace many blue-collar and white-collar jobs. The Teamsters union is pushing for laws to restrict automation. The Department of Labor encourages companies to be transparent about AI use. Some unions have negotiated AI protections in their contracts, but more progress is needed.
Johnson defends megabill's AI law freeze
Johnson defends megabill's 10-year freeze on state AI laws.
AI deployment security and governance are key
Kieran Norton from Deloitte discusses AI's impact on cybersecurity and business. AI can help detect network problems and phishing. Companies must use AI responsibly, considering privacy and data security. AI requires updated governance, secure systems, and new specialists. Norton suggests starting with small AI projects and being practical about AI's use.
Gemini's Veo 3 AI video generator threatens online truth
Google Gemini's Veo 3 can create realistic videos, but it could harm online truth. Veo 3 can render small objects in motion and create realistic audio. The lack of an image upload feature is the only thing holding it back. Once Veo 3 gets the image upload feature, it's all over. Google's moderation isn't strict enough to stop the worst from happening.
AI arms race could destroy humanity
The rapid development of AI by companies like Meta and Google is creating an arms race. Meta loosened safety rules for its AI bots, even with risks to underage users. AI can dehumanize people, making them dependent on AI services. AI-powered weapons could make wars easier to start. Experts warn that AI could pose a significant threat to humanity.
Samsung may integrate Perplexity AI into phones
Samsung and Perplexity are discussing a partnership to integrate AI into Samsung devices. Perplexity could have a pre-built app on Galaxy phones. Samsung might use Perplexity's AI assistant in its web browser. Samsung is expected to invest in Perplexity. The companies are also working on an AI operating system and AI agents app.
Brain dopamine informs AI adaptation
Researchers found that dopamine neurons in the brain map possible future rewards. These maps adjust to different situations and help us assess risks. This brain process is similar to recent AI advances. It could inspire new ways for AI to predict and adapt to changing environments. The findings show the brain uses a complex system to anticipate rewards.
AI will change advertising workforce says WPP's Mark Read
WPP CEO Mark Read says AI will reduce the number of jobs in advertising. However, AI will also create new types of jobs. Read compares AI's impact to the rise of social media. WPP is investing heavily in AI, including its WPP Open platform. Read believes the creative industries will be well-defended against AI.
Sources
- Nvidia chips make gains in training largest AI systems, new data shows
- Nvidia chips make gains in training largest AI systems, new data shows
- Nvidia’s Blackwell chips double AI training speed, MLCommons benchmark shows
- Ministers offer concessions on AI and copyright to avoid fifth Lords defeat
- What AI Training Data Transparency Means for Content Owners
- US labor unions fight to contain AI disruption
- Johnson defends megabill’s 10-year freeze on state AI laws
- The modern ROI imperative: AI deployment, security and governance
- Gemini’s Veo 3 AI Video Generator Is Just One Step Away From Decimating Truth on the Internet
- The AI arms race could destroy humanity as we know it
- Samsung and Perplexity reportedly ready for mobile AI integration
- Many possible futures: How dopamine in the brain might inform AI that adapts quickly to change
- WPP's Mark Read Says AI Will Upend the Advertising Workforce