#177 - Instagram AI Bots, Noam Shazeer -> Google, FLUX.1, SAM2

Last Week in AI

Safeguarding AI: The Role of Human-Like Judgement

2min Snip

00:00

Play full episode

Summary

Transcript

Episode notes

Combining multiple large language models (LLMs) can enhance AI outputs to better reflect human preferences, but introduces safety concerns. An LLM judge can evaluate and ensure alignment of generative model outputs with human values, potentially mitigating risks from rogue AIs. This judge can intervene to prevent harmful actions by generative AI, effectively acting as a safeguard against malevolent behaviors. While it may not eliminate the risks entirely, the LLM judge offers a mechanism to confine and control problematic AI actions, providing a layer of protection for human interests.

Our 177th episode with a summary and discussion of last week's big AI news!

With guest co-host Jon Krohn from the super data science podcast (https://www.superdatascience.com/podcast)!

If you'd like to listen to the interview with Andrey, check out https://www.superdatascience.com/podcast

If you would like to get a sneak peek and help test Andrey's generative AI application, go to Astrocade.com to join the waitlist and the discord.

In this episode, hosts Andrey Kurenkov and Jon Krohn dive into significant updates and discussions in the AI world, including Instagram's new AI features, Waymo's driverless cars rollout in San Francisco, and NVIDIA’s chip delays. They also review Meta's AI Studio, character.ai CEO Noam Shazir's return to Google, and Google's Gemini updates. Additional topics cover NVIDIA's hardware issues, advancements in humanoid robots, and new open-source AI tools like Open Devon. Policy discussions touch on the EU AI Act, the U.S. stance on open-source AI, and investigations into Google and Anthropic. The impact of misinformation via deepfakes, particularly one involving Elon Musk, is also highlighted, all emphasizing significant industry effects and regulatory implications.

Read out our text newsletter and comment on the podcast at https://lastweekin.ai/

If you would like to become a sponsor for the newsletter, podcast, or both, please fill out this form.

Email us your questions and feedback at contact@lastweekinai.com and/or hello@gladstone.ai

(00:00:00) AI Song / Intro Banter
(00:05:32) Response to listener comments / corrections
Tools & Apps
- (00:10:16) Apple Intelligence to Miss Initial Launch of Upcoming iOS 18 Overhaul
- (00:16:35) Instagram starts letting people create AI versions of themselves
- Lighting round
  - (00:22:49) Runway just dropped image-to-video in Gen3
  - (00:25:41) Midjourney drops surprise v6.1 update — now humans look more real than ever
  - (00:28:07) AI-Powered Necklace Will Be Your Friend for $99
  - (00:30:06) Microsoft is adding AI-powered summaries to Bing search results
Applications & Business
- (00:31:44) Character.AI CEO Noam Shazeer returns to Google
- (00:39:41) Perplexity is cutting checks to publishers following plagiarism accusations
- Lighting round
  - (00:43:30) Nvidia reportedly delays its next AI chip due to a design flaw
  - (00:41:08) Neura shows off humanoid robot 4NE-1
  - (00:46:0) Yes, there are more driverless Waymos in S.F. Here’s how busy they are
  - (00:57:27) Canva acquires Leonardo.ai to boost its generative AI efforts
Projects & Open Source
- (00:59:19) Black Forest Labs Open-Source FLUX.1: A 12 Billion Parameter Rectified Flow Transformer Capable of Generating Images from Text Descriptions
- (01:01:59) Google releases new ‘open’ AI models with a focus on safety
- Lighting round
  - (01:05:09) Stability AI releases super-fast model for 3D asset image generation
  - (01:09:29) OpenDevin: An Open Platform for AI Software Developers as Generalist Agents
Research & Advancements
- (01:12:10) Meta AI Introduces Meta Segment Anything Model 2 (SAM 2): The First Unified Model for Segmenting Objects Across Images and Videos
- (01:19:20) MoMa: Efficient Early-Fusion Pre-training with Mixture of Modality-Aware Experts
- Lighting round
Policy & Safety
- (01:33:03) World's First-Ever AI Law Now Enforced in Europe, Targeting US Tech Giants
- (01:39:12) White House says no need to restrict ‘open-source’ artificial intelligence — at least for now
- Lighting round
  - (01:41:12) With Smugglers and Front Companies, China Is Skirting American A.I. Bans
  - (01:44:03) UK antitrust body probes Google’s ties with AI rival Anthropic
  - (01:45:20) Elon Musk posts deepfake of Kamala Harris that violates X policy
(01:50:10) AI Outro

Get the Snipd
podcast app

Unlock the knowledge in podcasts with the podcast player of the future.

App store banner

Play store banner

AI-powered
podcast player

Listen to all your favourite podcasts with AI-powered features

Discover
highlights

Listen to the best highlights from the podcasts you love and dive into the full episode

Save any
moment

Hear something you like? Tap your headphones to save it with AI-generated key takeaways

Share
& Export

Send highlights to Twitter, WhatsApp or export them to Notion, Readwise & more

AI-powered
podcast player

Listen to all your favourite podcasts with AI-powered features

Discover
highlights

Listen to the best highlights from the podcasts you love and dive into the full episode