#211 - Claude Voice, Flux Kontext, wrong RL research?

Our 211th episode with a summary and discussion of last week's big AI news!

Recorded on 05/31/2025

Hosted by Andrey Kurenkov (twitter.com/andrey_kurenkov) and Jeremie Harris (twitter.com/jeremiecharris).

Feel free to email us your questions and feedback at contact@lastweekinai.comand/or hello@gladstone.ai

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

Join our Discord here! (discord.gg/nTyezGSKwP) discord.gg/nTyezGSKwP

In this episode:

• Recent AI podcast covers significant AI news: startups, new tools, applications, investments in hardware, and research advancements.

• Discussions include the introduction of various new tools and applications such as Flux's new image generating models and Perplexity's new spreadsheet and dashboard functionalities.

• A notable segment focuses on OpenAI's partnership with the UAE and discussions on potential legislation aiming to prevent states from regulating AI for a decade.

• Concerns around model behaviors and safety are discussed, highlighting incidents like Claude Opus 4's blackmail attempt and Palisade Research's tests showing AI models bypassing shutdown commands.

Timestamps + Links:

• (00:00:10) Intro / Banter

• (00:01:39) News Preview

• (00:02:50) Response to Listener Comments

• Tools & Apps

• (00:07:10) Anthropic launches a voice mode for Claude (techcrunch.com/2025/05/27/anthropic-launches-a…)

• (00:10:35) Black Forest Labs’ Kontext AI models can edit pics as well as generate them (techcrunch.com/2025/05/29/black-forest-labs-ko…)

• (00:15:30) Perplexity’s new tool can generate spreadsheets, dashboards, and more (techcrunch.com/2025/05/29/perplexitys-new-tool…)

• (00:18:43) xAI to pay Telegram $300M to integrate Grok into the chat app (techcrunch.com/2025/05/28/xai-to-pay-300m-in-t…)

• (00:22:42) Opera’s new AI browser promises to write code while you sleep (theverge.com/news/675406/opera-neon-ai-agentic…)

• (00:24:17) Google Photos debuts redesigned editor with new AI tools (techcrunch.com/2025/05/28/google-photos-debuts…)

• Applications & Business

• (00:25:13) Top Chinese memory maker expected to abandon DDR4 manufacturing at the behest of Beijing (tomshardware.com/pc-components/ddr4/top-chines…)

• (00:30:04) Oracle to Buy $40 Billion Worth of Nvidia Chips for First Stargate Data Center (pymnts.com/artificial-intelligence-2/2025/orac…)

• (00:31:47) UAE makes ChatGPT Plus subscription free for all residents as part of deal with OpenAI (businesstoday.in/technology/news/story/uae-mak…)

• (00:35:34) NVIDIA Corporation (NVDA) to Launch Cheaper Blackwell AI Chip for China, Says Report (finance.yahoo.com/news/nvidia-corporation-nvda…)

• (00:38:39) The New York Times and Amazon ink AI licensing deal (techcrunch.com/2025/05/29/the-new-york-times-a…)

• Projects & Open Source

• (00:41:11) DeepSeek’s distilled new R1 AI model can run on a single GPU (techcrunch.com/2025/05/29/deepseeks-distilled-…)

• (00:45:19) Google Unveils SignGemma, an AI Model That Can Translate Sign Language Into Spoken Text (gadgets360.com/ai/news/google-signgemma-ai-mod…)

• (00:47:08) Open-sourcing circuit tracing tools (anthropic.com/research/open-source-circuit-tra…)

• (00:49:42) Hugging Face unveils two new humanoid robots (techcrunch.com/2025/05/29/hugging-face-unveils…)

• Research & Advancements

• (00:52:33) PANGU PRO MOE: MIXTURE OF GROUPED EXPERTS FOR EFFICIENT SPARSITY (arxiv.org/abs/2505.21411)

• (00:58:55) DataRater: Meta-Learned Dataset Curation (arxiv.org/abs/2505.17895)

• (01:05:05) Incorrect Baseline Evaluations Call into Question Recent LLM-RL Claims (safe-lip-9a8.notion.site/Incorrect-Baseline-Ev…)

• (01:10:17) Maximizing Confidence Alone Improves Reasoning (arxiv.org/abs/2505.22660v1)

• (01:11:00) Guided by Gut: Efficient Test-Time Scaling with Reinforced Intrinsic Confidence (arxiv.org/abs/2505.20325v1)

• (01:11:44) One RL to See Them All (arxiv.org/abs/2505.18129)

• (01:15:05) Efficient Reinforcement Finetuning via Adaptive Curriculum Learning (arxiv.org/abs/2504.05520)

• Policy & Safety

• (01:17:58) Trump's 'Big Beautiful Bill' could ban states from regulating AI for a decade (usatoday.com/story/news/politics/2025/05/27/tr…)

• (01:24:31) Researchers claim ChatGPT o3 bypassed shutdown in controlled test (bleepingcomputer.com/news/artificial-intellige…)

• (01:30:10) Anthropic’s new AI model turns to blackmail when engineers try to take it offline (techcrunch.com/2025/05/22/anthropics-new-ai-mo…)

• (01:31:09) Anthropic Faces Backlash As Claude 4 Opus Can Autonomously Alert Authorities (wccftech.com/anthropic-faces-backlash-as-claud…)

• (01:35:37) Claude helps users make bioweapons (x.com/ARGleave/status/1926138376509440433)

• (01:35:49) The Claude 4 System Card is a Wild Read (ignorance.ai/p/the-claude-4-system-card-is-a-w…)

See Privacy Policy at art19.com/privacy and California Privacy Notice at art19.com/privacy#do-not-sell-my-info.