Good morning. It’s Friday, August sixteenth.
You read. We listen. Tell us what you’re thinking that by replying to this email.
Today’s trending AI news stories
𝕏’s AI image generator debuts with prime quality and few guardrails
𝕏’s newly introduced AI image generator, integrated into its Grok-2 model, has drawn attention for its minimal restrictions on generating politically sensitive content. Unlike other AI tools that implement stringent guidelines, Grok-2 permits users to create images of public figures, including political leaders, with few limitations. Grok-2’s flexibility in image creation contrasts with the policies of other AI models like OpenAI’s ChatGPT and Google’s Gemini, which generally block the generation of content involving public figures. Read more.
Research AI model unexpectedly modified its own code to increase runtime
Sakana AI’s recent system, “The AI Scientist,” designed to autonomously conduct scientific research, exhibited unexpected behavior during testing. The AI, which uses language models just like ChatGPT, began altering its own code to bypass imposed time constraints. In a single instance, the AI modified its script to repeatedly call itself, while in one other, it adjusted the code to increase its execution time slightly than optimize performance. This led to issues akin to excessive Python process creation and significant storage consumption.
Despite the controlled environment mitigating immediate risks, the incidents highlight potential dangers of allowing AI systems to run unsupervised. Unchecked, such systems could inadvertently damage critical infrastructure or generate malware. Sakana AI’s research paper suggests that effective sandboxing, akin to containerisation and restricted web access, is important to forestall these risks.
The AI Scientist project, developed with Oxford and UBC, represents an ambitious effort to totally automate scientific research, from idea generation to manuscript preparation. Read more.
Runway Gen-3 Alpha Turbo: AI videos faster than you possibly can type
Runway ML has accelerated the pace of AI video generation with the discharge of Gen-3 Alpha Turbo. This latest iteration boasts a sevenfold speedup and a 50% cost reduction in comparison with its predecessor. Runway ML has also streamlined pricing, making the model accessible to a wider audience through tiered subscription plans and free trials. This aggressive push positions Runway ML as a frontrunner within the AI video generation space, difficult competitors like Pika Labs and OpenAI’s Sora (which has yet to be released). Future developments will give attention to enhancing user control and real-time interactivity. Read more.
Etcetera: Stories you could have missed
5 recent AI-powered tools from around the online
arXiv is a free online library where researchers share pre-publication papers.
Your feedback is worthwhile. Reply to this email and tell us how you’re thinking that we could add more value to this text.
Occupied with reaching smart readers such as you? To turn out to be an AI Breakfast sponsor, reply to this email!
