Good morning from OWITH.ai, the podcast that gives you only what's important to hear in the AI and tech world.
In the latest issue of Lore, several significant developments in AI and tech have been highlighted. Anthropomic has restored access to its Fable 5 model after refining cybersecurity measures, though access will be limited to paid plans until July 7. This demonstrates a commitment to secure AI accessibility. Meanwhile, OpenAI is previewing its next-generation GPT-5.6 models, showcasing advanced capabilities in coding, biology, and cybersecurity. These models come with innovative features such as max reasoning effort and ultra mode, available in various tiers for different price points. A limited preview is currently available for trusted partners, with a broader rollout planned. Notably, OpenAI's GPT-5.6 models are restricted to government-approved users while safeguarding against misuse, highlighting the growing intersection between AI development and regulatory oversight.
Transitioning smoothly to another groundbreaking development, Meta has advanced non-invasive brain-to-text technology with Brain2QWERTY v2, capable of decoding full sentences from raw MEG brain signals in real-time. This open-sourced breakthrough aims to accelerate neuroscience research. In a similar vein of innovation, Etched has unveiled custom AI inference hardware after significant funding and contract acquisitions. The focus is on low-voltage inference and cluster-scale memory to enhance throughput on large models.
In the world of investment, 8VC has closed a $1.5 billion fund targeting AI infrastructure, underscoring the rapid pace of innovation and the firm's focus on execution. Additionally, Anthropic has launched Claude Sonnet 5, its most robust sonnet model yet, enhancing reasoning, tool use, coding, and knowledge work. Anthropic also introduced Claude Science—a beta workbench designed specifically for scientific tasks like data analysis or manuscript writing—as part of their broader strategy aimed at expanding professional applications within academia or industry sectors alike.
Switching gears to collaborative tools, Notion has introduced native HTML blocks for interactive pages, allowing users to embed fully functional HTML directly inside pages for real-time collaboration on various projects. Meanwhile, Google has released faster and cheaper image and video models—Nano Banana 2 Lite for quick image generation and Gemini Omni Flash for video editing—available through the Gemini API and AI Studio.
Continuing with corporate strategies, OpenAI has reportedly proposed donating 5% of its equity to a U.S. sovereign wealth fund as part of discussions about sharing AI's economic benefits with the public. This move is aligned with OpenAI’s broader discussions about offering a 5% equity stake to the U.S. government ahead of its potential IPO, aiming at sharing AI's economic benefits publicly while strengthening governmental ties.
The AI landscape is filled with noise from various media sources. DeepLearning.AI, led by Andrew Ng, prioritizes learners by providing quality AI education without accepting payments from partners for course creation. This initiative collaborates with leading companies like OpenAI, Google, and Microsoft to develop courses focused on fundamental AI engineering skills.
In parallel developments, Sakana AI released "Fugu" models that orchestrate tasks across various benchmarks. This approach reduces dependency on any single provider or nation amid access restrictions on advanced models.
Microsoft's venture into developing its own reasoning model marks a shift toward proprietary solutions integrated within their product ecosystem. In robotics advancements, researchers have developed "Roboreward," a vision-language model family improving reward estimations for training robots across different tasks.
The U.S. government recently lifted export controls on Anthropic's AI models Fable and Mythos—an essential move for cybersecurity experts utilizing Mythos in identifying security flaws. Despite this positive step, U.S. AI policy continues to grapple with transparency issues regarding licensing regimes. To streamline model release processes amidst these challenges, voluntary standards for frontier AI models are under discussion between leading labs and the government—focusing particularly on cybersecurity aspects.
Meanwhile Meta’s stock surged following reports regarding plans launching cloud computing business selling excess computing capacity—potentially offsetting significant investments made thus far into infrastructure but also posing competitive threats existing providers due increased competition within this space overall too!
Finally, potential changes in UK policy horizon advisors are developing strategies focused on reducing dependence on large companies and enhancing British sovereignty in the technology field. Future endeavors likely under Prime Minister Andy Burnham's administration will approach moving forward accordingly based on current trends observed globally today already evident elsewhere too similarly aligned interests therein shared common goals objectives ultimately pursued together collaboratively over time ahead!
Thanks for listening! Follow us on Twitter, Instagram and Linkedin
