- The Rundown AI
- Posts
- AI whistleblowers speak out
Sign Up | Advertise | Tools | AI University
Welcome, AI enthusiasts.
The AI safety discourse is boiling over — and a new ‘Right to Warn’ petition has both current and former researchers from top firms on the offensive.
With the chorus only growing louder as AI advancements continue to accelerate, the industry feels on the brink of a major reckoning. Let’s investigate…
In today’s AI rundown:
AI researchers demand ‘Right to Warn’
Musk diverts AI chips from Tesla to X
Explain images in real-time with GPT-4o
OpenAI researcher: AGI coming by 2027
5 new AI tools & 4 new AI jobs
More AI & tech news
Read time: 4 minutes
LATEST DEVELOPMENTS
OPENAI
Image source: Right To Warn AI
The Rundown: Current and former employees from top AI labs, including OpenAI, Anthropic, and DeepMind just published an open letter calling for companies to expand whistleblower protections so workers can raise the alarm about potential AI dangers without fear of retaliation.
The details:
The ‘Right to Warn AI’ petition was crafted by current and former employees of OpenAI, Anthropic, and Google DeepMind.
The open letter was also endorsed by AI visionaries Yoshua Bengio, Geoffrey Hinton, and Stuart Russell.
The statement pushes for AI firms to agree to several principles:
Eliminating non-disparagement clauses concerning AI risk
Establishing and facilitating anonymous channels for raising concerns
Expanding whistleblower protections and anti-retaliation measures
Several researchers posted threads on their experience, with Daniel Kokotajlo revealing he quit OpenAI after ‘losing hope’ the company would act responsibly.
Why it matters: The AI safety discourse is reaching a boiling point, and there is clearly a major industry divide that transcends any one AI firm or researcher. The proposed principles seem reasonable and necessary — but it remains to be seen if the top AI leaders will actually listen.
TOGETHER WITH AE STUDIO
The Rundown: Imagine a tool that transforms your business challenges into actionable solutions. AE Studio's new AI ideas tool does just that — helping you tackle your biggest problems efficiently.
Here’s how it works:
Answer 3 questions about your business needs
The AI churns out proven solutions
Try AI ideas by AE Studio for $0 right now and tackle your biggest problems today.
ELON MUSK
Image source: Nvidia
The Rundown: Newly-leaked internal emails to CNBC revealed that Elon Musk ordered chipmaker Nvidia to prioritize shipments of its AI processors to both X and xAI, delaying Tesla’s awaited $500M order by months.
The details:
Internal Nvidia emails show Musk redirected 12,000 H100 GPUs in December to X.
The move contradicts public statements that Tesla would increase H100 purchases from 35,000 to 85,000 and spend $10B on AI infrastructure.
Musk clarified that Tesla had nowhere to install the chips, with the automaker still planning to spend $3-4B in 2024 on Nvidia hardware.
Why it matters: While this is likely a nothingburger in the long run, and Elon is still clearly intent on spending massively on AI across his companies, moves like this only fuel critics who remain skeptical of his startup’s entanglements — particularly as Tesla pushes to deliver on lofty self-driving goals.
AI TRAINING
The Rundown: This hack lets you access GPT-4o directly from an iPhone via a custom shortcut, allowing you to analyze and explain images in real time.
Step-by-step:
Head over to OpenAI’s platform to obtain an API key. Make sure you have some credits in your account.
Go to your iPhone Shortcuts and click the “+” button.
Add different actions before your OpenAI API call, e.g., your desired prompt as “Text”, your API key as “Text”, and “Take a screenshot”.
Call GPT-4o by using “*https://api.openai.com/v1/chat/completions”* and obtain a description of what you are seeing!
Bonus: You can also simply download the full Shortcut in The Rundown University here 🎉
PRESENTED BY COVALENT
The Rundown: Looking for a simpler way to develop and deploy your AI applications? Covalent allows you to effortlessly train, fine-tune, and serve AI models in minutes.
The details:
Train, fine-tune, & serve AI models with only a few additional lines of Python
Covalent handles all of the underlying infrastructure, providing serverless H100s, A100s & more
Only pay for compute used — no reservations or minimum spend required
AI RESEARCH
Image source: Situational Awareness
The Rundown: Former OpenAI researcher Leopold Aschenbrenner just released a new essay series detailing his view on AGI, saying the rate of AI progress will be the most intense and volatile events in human history.
The details:
Aschenbrenner says that ‘nobody is pricing in’ what is coming in AI, and to expect another GPT-2 to GPT-4 level jump by 2027 (that would take us to AGI).
The researcher predicts that hundreds of millions of AGI would then rapidly accelerate progress, compressing decades of progress into a year.
He also discussed the economic and military advantages that will come with AGI, calling it a national security issue that needs its own ‘Manhattan Project’.
Aschenbrenner reiterated these views on the Dwarkesh Podcast, also revealing he was fired from OpenAI after raising AI security concerns.
Why it matters: As an insider at OpenAI, Aschenbrenner's analysis carries weight — and his projections paint a striking picture of how radically AGI will reshape the world. His commentary on the firing also raises even more drama, given OpenAI’s current media battle with former researchers and the board over safety concerns.
NEW TOOLS & JOBS
✂️ Underlord by Descript - An AI video editor that works like docs and slides
📺 Showrunner - Generate AI TV shows and episodes
🎮 Rosebud AI Gammaker - Create, share, and play AI-powered games
🍻 Crewting - Turn isolated remote workers into connected teams with AI-matching
🚗 Sēkr - AI-enhanced road trip planning and community for vanlife travelers
QUICK HITS
Cohere is set to raise $450M in fresh funding from Nvidia, Salesforce, Cisco, and others, valuing the company at $5B.
Shutterstock announced a new deal with Reka AI, with the company’s AI licensing business reportedly generating $104M last year from agreements with major AI firms.
A Stanford University AI team apologized after it was revealed its Llama 3-V architecture was nearly identical to Chinese lab ModelBest’s open-source MiniCPM-Llama3-V 2.5 model.
Major AI platforms suffered outages at various points on Tuesday, with ChatGPT, Claude, Gemini, and Perplexity all experiencing intermittent issues.
Cisco launched a new $1B global AI investment fund, revealing initial investments in Cohere, Mistral AI, and Scale AI.
Nvidia announced a $50M investment in Twelve Labs Inc., a startup focused on AI search and analysis of video content.
THAT’S A WRAP
SPONSOR US
Get your product in front of over 600k+ AI enthusiasts
Our newsletter is read by thousands of tech professionals, investors, engineers, managers, and business owners around the world. Get in touch today.
FEEDBACK
How would you rate today's newsletter?Vote below to help us improve the newsletter for you. |
If you have specific feedback or anything interesting you’d like to share, please let us know by replying to this email.
Reply