| |
Welcome to the latest Inside AI newsletter! Here are the latest headlines for today: - ChatGPT experienced a slowdown in monthly visits.
- Nvidia unveiled its latest AI superchip for inference tasks.
- Pope Francis stressed the need for responsible AI ahead of World Day of Peace.
- OpenAI has launched GPTBot, a web crawling bot.
If you find this newsletter useful, please share it with your friends and colleagues. Thank you. Beth
p/beth-duckett | |
| 1 | What happened: ChatGPT's web traffic is slowing down, according to Similarweb estimates. July marked the second straight month of a drop in ChatGPT's monthly visits, which fell by 9.6% to nearly 1.5 billion. What the numbers say: June's 1.6 billion visits reflected a 9.7% decrease from May's peak of 1.8 billion. In the U.S. only, traffic to chat.openai.com fell 15% in June and another 4% in July. In contrast, Google's Bard experienced a 34.5% spike in July visits to 189 million, although it remains significantly below ChatGPT's numbers. ChatGPT also outperformed Microsoft's Bing search engine with 1.47 million visits in July, compared to 1.2 million for bing.com. The bigger picture: The gap between AI chatbots could be shrinking as more people adopt alternatives like Bing Chat, which uses the same OpenAI algorithms. It's uncertain if ChatGPT's traffic dip is also a seasonal pattern tied to school breaks. College-age individuals make up over a quarter of ChatGPT's adult audience, with 28.6% in the U.S. and 27% worldwide, according to Similarweb. 
|     | |
| 2 | | Nvidia unveiled its latest AI superchip, the GH200, specifically tailored for AI inference tasks. AI inference refers to the process where AI models apply their learned knowledge to make predictions or decisions or generate content based on new input data. More: - Nvidia's superchip combines the GPU from Nvidia's top-tier AI chip, the H100, with 141 gigabytes of memory and a 72-core Arm central processor.
- The superchip, an upgraded version of the Grace Hopper combo chip, is specifically designed to enhance inference tasks, where AI models make predictions or generate content.
- With the chip, the inference cost of large language models is expected to "drop significantly," said Nvidia CEO Jensen Huang.
- The chip achieves faster performance with high-bandwidth memory 3 (HBM3e), reaching 5 terabytes per second data access speeds.
- The GH200 will be available in Q2 2024 as standalone chips or integrated server systems.
Zoom out: - While the GH200's exact price is unknown, Nvidia's H100 line has sold for upwards of $40,000.
- Nvidia's move addresses competition from AMD's new MI300X AI chip and efforts by Google and Amazon to create their own custom AI chips for inference.
| | | |
| A message from our sponsor, Revelry. |  | | Big ideas - big opportunities - demand additional support. Technology is evolving faster than ever, thanks to headline grabbers like AI. As your business considers new ways to compete, scale, and succeed in today’s digital world, companies like Revelry offer an alternative to recruiting, hiring, and training new FTEs. Look to them for expertise in: - Custom software / product development
- Engineering staff augmentation
- Digital strategy consulting
- Product organization development and coaching
See how Revelry’s Lean Agile process and Innovation Stack help engineering teams form and deliver 30-40% faster than industry averages. Find out how | |
|
| 3 | | In a statement today, Pope Francis stressed the importance of responsible and ethical AI, highlighting that the technology should be used in the "service of humanity." In a declaration shared by the Vatican, the Pope cited the rapidly growing influence that AI is having on society. Francis said there is an "urgent need to orient" AI in a responsible way so as to avoid "conflicts and antagonism." More: - The Vatican revealed the theme "artificial intelligence and peace" for the upcoming World Day of Peace in the Catholic Church, set for New Year's Day.
- The Pope urged an open dialogue regarding the meaning of AI technologies.
- He highlighted the ethical use of AI in areas like education and law, warning against neglecting vulnerable and marginalized groups.
Zoom out: | | | |
| 4 | OpenAI introduced GPTBot, a web crawling bot that extracts website data for training and improving large language models. Websites can block GPTBot's access by using methods like IP blocking or adjusting their robots.txt files. More: - The GPTBot gathers publicly available data from the internet, excluding paywalled content, personally identifiable information, and policy-violating text, according to OpenAI.
- The data is used to improve current and upcoming models while filtering out restricted sources.
- OpenAI's recent trademark application for GPT-5 hints at a potential successor to GPT-4.
Zoom out: - Despite the launch, legal challenges persist in companies' web scraping practices.
- Last year, a U.S. court confirmed the legality of scraping publicly accessible data, yet AI training using scraped data has faced numerous lawsuits citing copyright infringement and privacy breaches.
| | | |
| A message from our sponsor, Airplane. |  | | Build AI-powered internal tools with Airplane Airplane is the developer platform for building custom internal tools. You can transform scripts, queries, APIs, and more into powerful UIs and workflows. Airplane's newest solution, AI Applications, allows you to seamlessly integrate AI into your internal tools. Use Airplane’s AI built-ins to quickly incorporate large language models into your workflows with minimal configuration. These AI built-ins offer several features: - Chat function to send a single message to an LLM
- Chatbot function to facilitate a conversation with an LLM
- Reusable AI function to convert inputs to outputs by following a set of instructions
Use AI Applications to automate data entry, parse information from documents, and more. Sign up today! | |
|
| 5 | U.S. Sen. Mark Warner (D-Va.) is pressing Google for more transparency regarding the hospital testing of its medical chatbot, Med-PaLM 2. The chatbot is currently undergoing limited testing at medical institutions, including Mayo Clinic, with the goal of offering medical information and support to healthcare providers. More: - Warner's concerns revolve around patient risks stemming from a rush to commercialize the technology.
- In a letter to Google CEO Sundar Pichai, Warner cautioned against premature deployment, citing concerns over inaccuracies and reservations expressed by Google's own senior researchers.
- He stressed that the unproven technology has the potential to damage trust and exacerbate racial disparities and diagnostic errors.
- Warner asked Google leaders for clarification in areas like patient privacy, data retention, and safeguards against over-reliance on AI-generated output.
Zoom out: - Med-PaLM 2, trained on medical expert demos and licensing exam materials, is based on an updated version of Google's PaLM 2 language model.
- A Google spokesperson defended the technology, highlighting its limited rollout and commitment to safety and privacy.
- In a study, Med-PaLM and Med-PaLM 2 had more inaccuracies than doctors but still performed similarly in other areas, including reasoning and comprehension accuracy.
| | | |
| |
| 7 | | QUICK HITS - Wiz partnered with Wiley to create the AWS Security Foundations Guide For Dummies. This free 46-page guide includes techniques for securing AWS resources.*
- Foxconn is set to be Apple's exclusive provider of dedicated servers for AI testing and training. Foxconn already supplies servers to Amazon, OpenAI, and Nvidia and is Apple's main data center server provider.
- Stable Diffusion co-creator Stability AI launched StableCode, an open large language model catering to programming needs.
- Spotify's AI-powered "DJ" is expanding globally to dozens of markets, offering personalized curated music with synthetic voice commentary that references users' listening history.
- According to The Verge, smaller players in generative AI are feeling left out as large tech companies like Microsoft and OpenAI dominate AI policy discussions and influence regulations.
- Join over 5,000 fast-growing companies around the world that trust Vanta to manage risk and prove security in real time. Inside readers get $1,000 off.*
*This is a sponsored post. | | | |
| Upcoming Events | | AUG
5 | | Learn about streamlining SOC 2 compliance with Vanta and AWS. Join the webinar on Aug 17.* | | | | | | | OCT
4 | | Join this free hybrid event in Amsterdam and you'll hear from industry experts at Meta, Opera, AdGuard, Mozilla, Google, and more.* | | | | | | | OCT
16 | | Meet Our Fund 5, Building and Investing in AI | | | | | | | NOV
7 | | Join 300+ CCOs and senior leaders from Amex, Verizon, Hilton, Wells Fargo and more at Reuters Events: Customer Service & Experience East 2023.* | | | | | | * This is a sponsored event | | | |
Term of the Day Display ads: Display ads are a type of online advertisement that is used on websites, apps, and social media platforms to promote a product or service Question of the Week Do you think the 9-to-5 workday still relevant in today's modern workplace? Join the conversation |
INSIDE AI LEADERBOARD (7 DAYS) |
 | AI and technology writer | | Beth is the curator and writer of the AI and Tech newsletters at Inside. With a career in journalism and editing that began in 2007, she has written for publications including USA Today, The Arizona Republic, Arizona Business Gazette, and more. You can reach her at Beth.Duckett@yahoo.com. | | This newsletter was edited by Aaron Crutchfield | |
| |
| |
|
 | Build custom internal UIs and workflows quickly using code with Airplane. | |
|
| |
Comments
Post a Comment