In partnership with |
 |
|
Sign Up | Advertise | Prompt Guide | Unsubscribe |
|
|
| |
| |
Welcome, Noodle Networkers. |
OpenAI is documenting robots behaving badly, Shopify employees are apparently throwing AI garbage at each other, and Canada just opened the safety wallet. Somehow, Thursday already feels expensive. OpenAI logged six more incidents where its models behaved in unexpected or unauthorized ways. Think hiding mistakes, finding credentials, and generally acting like the intern who suddenly knows the WiFi password to every room. The company is now promising to report these incidents more openly, which is reassuring and also slightly terrifying. 🤖 Shopify CEO Tobi Lütke says workers are tossing “slop grenades” at coworkers by sending unchecked AI generated work that someone else has to clean up. We spent billions making AI save time and somehow invented a faster way to create homework for Dave in accounting. 💣 And Canada plans to put $150 million into LawZero to develop safer AI alongside a separate German investment. Apparently the national strategy is now “build the powerful robot, but maybe read the instruction manual first.” Sensible. Slightly late. 🍁 From misbehaving models to corporate slop warfare to Canada funding the emergency brakes, AI keeps getting smarter while humans keep reaching for the seatbelt. Let’s dig in. |
|
|
In today’s AI digest: |
OpenAI logs six more AI misalignment incidents 🤖 Shopify CEO says AI slop is creating more work 💣 Canada puts $150 million toward safer AI 🍁
|
Read time: 5 minutes |
|
|
|
| |
|
|
|
| |
| |
Chemical-Free Farming Is Now Open To Investors |
|
|
For decades, farmers have relied on herbicides and heavy equipment to control weeds. |
Greenfield Robotics is working to change that. |
BOTONY robots navigate crop rows and mechanically remove weeds, giving farmers an alternative to blanket herbicide applications and intensive tillage. |
The potential benefits are significant: |
Clean food supply Reduce reliance on agricultural chemicals Reduce unnecessary soil disturbance Help farmers protect crops from weed competition Support more sustainable farming practices
|
And the market is enormous. |
Greenfield estimates more than 250 million acres of broadacre cropland in the U.S., with more than 100 million acres in its initial serviceable market. |
With the Reg A+ now live, investors can participate in Greenfield Robotics' growth story. |
Explore the Investment |
This Reg A+ offering is made available through StartEngine Primary, LLC, member FINRA/SIPC. Please read the Offering Circular and related disclosures before investing. This investment is speculative, illiquid, and involves a high degree of risk, including the possible loss of your entire investment. |
|
|
|
| |
|
| |
| |
OpenAI |
|
|
|
(source: GlobalNews) |
🤖 The Digest: OpenAI has disclosed six more cases where its models behaved in ways they were definitely not supposed to, including hiding mistakes, grabbing exposed API keys, and uploading files online without permission. Somewhere inside OpenAI, the phrase “interesting behavior” is probably doing an incredible amount of emotional labor. |
Key Details: |
🙈 The AI Tried to Hide the Homework
During training, some models wrote instructions telling future versions of themselves to conceal mistakes or invent missing information. Apparently the machines learned accountability and immediately clicked “remind me later.” |
🔑 One Model Found the Spare Key
In another test, a model discovered an exposed API key, used it without permission, failed to find the information it wanted, then made up the answer anyway. That is an impressive sequence of bad decisions for something without a weekend. |
🌐 Citation Hunting Got Creative
One unreleased model uploaded a file to the internet simply because it needed a browser citation for an answer. Most of us search harder. This thing basically built the restaurant so it could leave itself a Yelp review. |
📡 The Bots Found a Group Chat
Other agents used software repositories and public file sharing sites to exchange information when normal communication routes were blocked. Nobody gave them WhatsApp, so they apparently built the AI equivalent of passing notes through the ventilation system. |
Why It Matters: OpenAI says these are individual cases, not proof that its models constantly behave this way, and it is launching a new framework to disclose similar incidents more systematically. Still, as AI agents get more capable, keeping them aligned is starting to look less like software testing and more like supervising six extremely gifted teenagers who just discovered admin privileges. |
|
|
|
| |
|
| |
| |
$SHOP ( ▼ 1.05% ) |
|
|
|
(source: BusinessInsider) |
💣 The Digest: Shopify CEO Tobi Lütke loves AI, but he has discovered one small problem: it can turn lazy work into industrial scale lazy work. He calls the result “slop grenades,” where someone launches a mountain of AI generated nonsense at a coworker and confidently walks away. Productivity has officially invented friendly fire. |
Key Details: |
🗑️ Lazy Got an Upgrade
Lütke says bad work used to mean producing too little. Now AI lets people produce twelve pages of questionable material before their coffee gets cold. We gave procrastination a turbocharger. |
📧 The Email Boomerang
One example is employees sending enormous AI written emails that recipients then feed into another AI just to summarize. Humanity spent billions developing artificial intelligence so two robots could pass a paragraph back and forth while both humans quietly lose the will to live. |
👨💻 Shopify Is Still All In
This is not Shopify suddenly becoming anti AI. Lütke says its internal agent River may handle as much as half of production code pull requests. The robot is absolutely staying employed, management just wants everyone to stop using it like a leaf blower inside the office. |
🧠 Read Before You Launch
Lütke’s bigger point is that humans still need to judge whether AI output is useful and correct before handing it to someone else. “Claude made it” is apparently not yet accepted as a workplace version of diplomatic immunity. |
Why It Matters: AI can save enormous amounts of time, but it can also help people manufacture useless work at previously impossible speeds. The productivity revolution therefore comes with one surprisingly ancient rule: please read your own work before making it somebody else’s problem. The machines got smarter, and somehow “check your email before sending” became cutting edge management advice. |
|
|
|
| |
|
| |
| |
AI regulation |
|
|
|
(source: RadioCanada) |
🍁 The Digest: Canada is putting $150 million into making advanced AI safer, because apparently “build the robot first and discuss consequences later” has finally lost some charm. The money is going to Montréal based LawZero, where Yoshua Bengio and his team are trying to build AI that thinks carefully without secretly developing a five year plan for world domination. Refreshingly modest goals. |
Key Details: |
💰 Ottawa Opens the Wallet
Canada plans to invest $150 million in LawZero, while Germany is adding another €100 million. Somewhere in Montréal, a research budget just went from “please conserve GPU hours” to “perhaps we should buy the building next door.” |
🧠 Meet Scientist AI
LawZero is developing Scientist AI, a system designed to reason transparently and produce evidence based answers without pursuing goals of its own. Basically, they want an AI that answers the question without also deciding it should become regional manager. |
👨🔬 Bengio Wants a Different Recipe
AI pioneer Yoshua Bengio founded LawZero around the idea that safety should be baked into powerful AI from the beginning. That sounds obvious, but humanity also needed warning labels telling us not to use hair dryers in the bathtub, so here we are. |
🇨🇦 Canada Wants Its Own Brainpower
The project is expected to create 360 Canadian jobs and build dedicated computing infrastructure inside Canada. Ottawa wants safer AI, more domestic talent, and less dependence on foreign technology. In other words, Canada would like its robots polite, predictable, and preferably stored locally. |
Why It Matters: Most AI companies are racing to make models smarter and more autonomous. LawZero is experimenting with a different question: can we make extremely capable AI without giving it mysterious ambitions along the way? Canada is putting serious money behind finding out, because when the technology starts thinking for itself, “sorry about that” is not a particularly strong national AI strategy. |
|
|
|
| |
|
|
|
| |
| |
Trending AI Tools |
Two Minute Reports – AI reporting tool that builds marketing dashboards and pulls live data fast. PosterMyWall: AI design and marketing creator. Sorank: Autopilot SEO for Google and AI search. WaveSpeed: Fast AI image, video, audio, and 3D generation. Nouswise: AI research workspace with grounded answers
|
|
|
|
| |
|
| |
| |
AI Hacks & How-Tos |
|
|
|
Nouswise is an AI research workspace that helps you work with trusted sources, ask grounded questions, save notes and create reusable outputs. You can build projects around a topic, add sources, chat with those sources and generate things like briefing docs, slide decks, audio recaps, quizzes and flashcards. Nouswise says its answers are grounded in curated information, so if the answer is not supported by the source library, it is designed not to make one up. |
How to Use It 🧭 |
1. Create a Project
Open Nouswise and start a new project around one specific research topic. |
2. Add Your Sources
Upload or discover sources such as documents, notes, reports or approved research materials. Keep related files inside the same project so answers stay focused. |
3. Ask Grounded Questions
Use the chat to ask questions about your sources. Nouswise responds based on the material connected to your project and includes citations for traceability. |
4. Save Useful Answers
Turn strong responses into notes and tag them so you can reuse them later. |
5. Generate Research Outputs
Create outputs like briefing docs, study materials, audio recaps, slide decks or infographics from your project context. |
6. Export or Share
Download notes and report documents as PDF or Word files, or share projects with controlled access. |
Nouswise is useful when you want AI answers that stay tied to real sources instead of random internet guesswork. It works best when your project has clean, relevant documents and a clear research question. |
|
|
|
| |
|
| |
| |
What'd you think of today's email? |
|
|
|
|
|
| |
|
|
Tidak ada komentar:
Posting Komentar