- AI Fire
- Posts
- 💾 AMD Replaces Nvidia in AI Chip?
Plus: Musk buys Character AI?
Read time: 5 minutes
Have a great day!
For years, people have said data is like oil, one of the world’s most valuable resources. Just as big oil reserves gave wealth and power to oil-rich countries and companies in the 20th century, having lots of data can do the same in the 21st century.
What are on FIRE 🔥
AI REPORT
📊 47% of US Homes Familiar with AI Tech
Source: Parks Associates
New research shows 40% of US households regularly use AI tools for various purposes.
Key Takeaways:
47% of US internet households are familiar with AI technologies like ChatGPT and Lensa.
40% of US internet households use AI tools for personal, professional, or educational purposes.
69% of households have data and privacy concerns regarding AI use.
AI tools enhance efficiency and productivity, benefiting media creators and marketers.
Non-users cite lack of perceived value and privacy concerns as barriers to AI adoption.
Why it matters: AI will make work faster and easier as it gets better. It will handle boring tasks, but humans will still need to check the work. Media companies will use AI to make videos faster and better. Easy-to-use AI will get more people to try it, and custom chatbots will become common, like apps. Being able to change and adapt is important because people who don’t use AI will fall behind. Human skills like thinking, creativity, and understanding emotions are still very important to manage AI safely.
TODAY IN AI
AI HIGHLIGHTS
💻 CEO Lisa Su reports AMD's data center business doubled in a year. The Verge's Sean Hollister highlights AMD's plan to release new AI chips annually, competing with Nvidia's H100.
🧠 At Fortune Brainstorm AI Singapore 2024, Minister Josephine Teo discusses Singapore's AI plans. Unlike oil, data can be reused. Singapore adopted AI strategy in 2019, leveraging its position as trade hub.
🛍️ Using 'artificial intelligence' in product descriptions reduces purchase intent. AI mention lowers emotional trust, especially for high-risk products like expensive electronics and medical devices.
Do you feel less interested in a product if it's labeled as created by AI? |
📈 Facebook and Instagram parent Meta sees 7% YoY growth in daily active users, reaching 3.27B in June.
📰 SEO professionals concerned about AI Overviews' impact on traffic. Visibility drops from 15% to 7% of queries.
DAILY AI FUNDRAISING
Sybill raised $11M for its AI assistant that saves salespeople over 5 hours weekly by recording calls, summarizing conversations, drafting emails, and updating CRM systems, boosting sales efficiency.
AI SOURCES FROM AI FIRE
NEW EMPOWERED AI TOOLS
💽 GitStart AI Ticket Studio is AI to write engineering-ready tickets
🕹️ Jamie is human-quality meeting summaries without a bot
🧮 Table grows and maintains your network effortlessly with AI
🗂️ Outlit is AI agent for your SaaS deals
☑️ CopyFrogAI is an AI-powered content creation platform
AI QUICK HITS
1. 🎓 AI cheating threatens the integrity of Australian universities (Read more)
2. 🌮 Taco Bell to expand AI in drive-thrus (Read more)
3. 🚗 SoundHound AI launches voice assistant in Al (Read more)
4. 💾 US considers restricting China’s access to AI chips (Read more)
5. 🚫 Musk denies xAI plans to buy Character.AI (Read more)
AI CHART
Where do these numbers come from, and can they be trusted?
Benchmarks drive progress and help users choose the right model but are often too easy or flawed, leading to inaccurate real-world performance.
Initially, models scored around 25%, but now top models score between 88% and 90%, making it hard to see differences in their abilities, known as "saturation."
New, more challenging benchmarks like MMLU-Pro, GPQA, and MuSR are being developed to address these issues.
MMLU also highlights two other problems.
Incorrect Answers: Many questions have errors, with some having no correct answer or multiple correct answers.
Contamination: Models may "cheat" by having seen benchmark questions during training, making scores unreliable.
The greatest trick AI ever pulled
Anthropic funds new benchmarks to assess risks like cyber-attacks, while Dr. Liang at Stanford and Hugging Face develop standardized testing systems like HELM and EleutherAI Harness. New benchmarks like GAIA and NoCha ensure models are evaluated consistently and accurately for real-world applications.
AI TUTORIAL
Turn Text Into Sora-like AI Videos
Kling AI's text-to-video tool lets you make amazing videos like Sora just by using simple text prompts. This gives you new and easy ways to create high-quality visuals.
Visit Kling AI and create a free account.
On the main dashboard, click on "AI Videos."
In the "Prompt" section, describe the video you want to make.
Adjust settings such as creativity level, video quality, length, and aspect ratio.
Click "Generate" and watch your text turn into a video!
AI JOBS
We read your emails, comments, and poll replies daily
How would you rate today’s newsletter?Your feedback helps us create the best newsletter possible |
Reviews of the day
 |  |
LAST WORD 👋
How are we doing?
We love hearing from readers and are always looking for feedback. How are we doing with AI Fire? Is there anything you'd like to see more of or less of? Which aspects of the newsletter (or AI Fire Academy) do you enjoy most?
Hit reply and say Hello – we'd love to hear from you!
Like what you're reading? Forward it to friends, and they can sign up here.
Cheers,
The AI Fire Team
Reply