Learning Library

← Back to Library

Llama 3.1, EU AI Act, VPC Sandbox

Key Points

  • Meta unveiled Llama 3.1, an open‑source multilingual model family (8B, 70B, and a groundbreaking 405B parameter version) that rivals top proprietary LLMs and offers extensive tuning flexibility for developers.
  • The EU AI Act took effect on August 1, introducing a risk‑based regulatory framework that bans high‑risk AI practices, sets standards for high‑risk systems, and governs general‑purpose AI models to promote trustworthy AI in Europe.
  • IBM launched a free, non‑production IBM Cloud VPC sandbox that lets customers benchmark the performance of Intel Xeon 2nd‑ and 4th‑gen processors using workloads like Monte Carlo simulations, Hugging Face inference, Presto data lakes, or their own applications, with full cost reimbursement via IBM Cloud credits.

Full Transcript

# Llama 3.1, EU AI Act, VPC Sandbox **Source:** [https://www.youtube.com/watch?v=qXxFLf7oz98](https://www.youtube.com/watch?v=qXxFLf7oz98) **Duration:** 00:03:25 ## Summary - Meta unveiled Llama 3.1, an open‑source multilingual model family (8B, 70B, and a groundbreaking 405B parameter version) that rivals top proprietary LLMs and offers extensive tuning flexibility for developers. - The EU AI Act took effect on August 1, introducing a risk‑based regulatory framework that bans high‑risk AI practices, sets standards for high‑risk systems, and governs general‑purpose AI models to promote trustworthy AI in Europe. - IBM launched a free, non‑production IBM Cloud VPC sandbox that lets customers benchmark the performance of Intel Xeon 2nd‑ and 4th‑gen processors using workloads like Monte Carlo simulations, Hugging Face inference, Presto data lakes, or their own applications, with full cost reimbursement via IBM Cloud credits. ## Sections - [00:00:00](https://www.youtube.com/watch?v=qXxFLf7oz98&t=0s) **Untitled Section** - ## Full Transcript
0:00llama 3.1 the EU AI act and a new IBM 0:05Cloud VPC sandbox all on this episode of 0:08IBM Tech 0:12now on July 23rd meta announced the 0:15launch of the Llama 3.1 collection of 0:18multilingual large language models llama 0:213.1 comprises both pre-trained and 0:24instruction tune text in text out 0:26open-source generative AI models these 0:29models come ins size of 8B 70b and for 0:32the first time 405b parameters now this 0:36instruction tuned llama 3.1 405b figures 0:40to be the largest and most powerful 0:42open-source language model available 0:44today even making it competitive with 0:46the best proprietary models on the 0:48market perhaps even more importantly 0:50llama 3.1 provides the global AI 0:52Community with an open state-of-the-art 0:55model family and development ecosystem 0:57to build experiment and respon POS L 1:00scale new ideas and approaches IBM and 1:03meta launched the AI Alliance in 2023 1:05with the goal to help shape the 1:07evolution of AI to best reflect the 1:09needs and complexity of our societies 1:12llama 3.1 takes an important step in 1:14this direction while also providing an 1:16unprecedented opportunity to tune and 1:18tailor state-of-the-art generative AI 1:20models to your specific use case to 1:23learn all about the new release check 1:24out the link in the description of this 1:27video next up the EU AI act entered into 1:31force on August 1st making the European 1:33Union one of the first major markets to 1:35Define new rules around AI as one EU 1:38official said the aim is to turn the EU 1:41into a global hub for trustworthy AI the 1:44AI act takes a risk-based approach 1:47meaning that it categorizes applications 1:49according to their potential risk to 1:50fundamental rights and safety some of 1:53the most important Provisions are that 1:54the ACT prohibits certain AI practices 1:57that are deemed to pose unacceptable 1:58risk outline standards for developing 2:01and deploying certain high-risk AI 2:02system and provides rules for general 2:05purpose AI models according to IBM's 2:07brunoo it's a form of regulatory 2:10pragmatism that aims to adapt 2:11constraints with the level of risk the 2:14EU AI act will follow a very fast-paced 2:17schedule for the full details on what 2:19that entails hit the link 2:21below to wrap things up today 2:23performance testing is crucial to 2:25understanding the efficiency of complex 2:27applications inside your Cloud hosting 2:29environment 2:30that's why we're proud to introduce a 2:32new Sandbox that uses IBM Cloud virtual 2:34servers for VPC we're inviting customers 2:37into a non-production environment to 2:39test the performance of second gen and 2:41fourth gen Intel xon processors across 2:44various applications you can choose to 2:47Target benchmarks with a Monte Carlos 2:49simulation hugging face inference 2:51application or Presto data Lake 2:53application or if it's more useful you 2:55can even bring your own custom 2:56application to test for a limited time 2:59you can even try try the ibmcloud VPC 3:01sandbox and get 100% of your cost 3:03reimbursed with ibmcloud credits to 3:06learn more hit the link below thanks so 3:09much for joining me today for this 3:10episode of IBM Tech now if you're 3:12interested in learning more about the 3:13topics I've covered make sure you 3:15explore the links in the description of 3:16this video and of course please don't 3:18forget to subscribe to our channel to 3:20stay up to date on what's going on in 3:22Tech now