- Everyday AI
- Posts
- Balancing AI Productivity and Human Intelligence in Everyday Work
Balancing AI Productivity and Human Intelligence in Everyday Work
Judge challenges Meta on copyrighted books, Amazon Alexa+ hits 100,00 users, Microsoft to host Grok on Azure and more!
š Subscribe Here | š£ Hire Us To Speak | š¤ Partner with Us | š¤ Grow with GenAI
Outsmart The Future
Today in Everyday AI
7 minute read
š Daily Podcast Episode: AI makes us more productive, but it can make us dumber, too. Yikes. Go listen.
šµļøāāļø Fresh Finds: OpenAIās sycophancy study, Google NotebookLM app on the way and a new AI scientist agent. Read on for Fresh Finds.
š Byte Sized Daily AI News: Judge challenges Meta on copyrighted books, Amazon Alexa+ hits 100,00 users and Microsoft to host Grok on Azure. Read on for Byte Sized News.
š§ Leverage AI: So you wanna use LLMs as much as possible yet keep your brain power? Hereās your guide, from a leader at Notion. Keep reading for that!
ā©ļø Donāt miss out: Did you miss our last newsletter? We talked about Google widening access to AI mode, NVIDIA and Anthropic clash on U.S. chip restrictions, Meta Ray-Bans making voice recording default and the future of the AI workforce. Check it here!
Balancing AI Productivity and Human Intelligence in Everyday Work
You're outsourcing your brain to AI š§ .
Bad idea?
AI can write your SQL queries. Build your dashboards. Even brainstorm your next big idea.
Itās saving you hours. Maybe days.
But hereās the catchāit's also stealing your critical thinking. Making you reliant.
Maybe even... dumber.
Sumit Gupta knows this first-hand. Heās built data strategies at Notion, Snowflake, and Dropbox.
And, heās here to break down how AI is both supercharging productivity and quietly eroding our problem-solving skills.
Are we trading our brains for convenience?
Letās find out.
Also on the pod today:
How AI can cost more than brainpower šø
Studies that show cognitive decline š§
How to stay sharp in an AI-first world š¦¾
Itāll be worth your 33 minutes:
Listen on our site:
Subscribe and listen on your favorite podcast platform
Listen on:
Hereās our favorite AI finds from across the web:
New AI Tool Spotlight ā SupaMetas helps you up your RAG game, Intelswift is an AI agents marketplace, Strella uses AI to give you faster insights and clips from videos.
OpenAI ā OpenAI published a deep dive on their findings with GPT-4o being more sycophantic.
Weāve spent the last few days doing a deep dive on what went wrong with last weekās GPT-4o update in ChatGPT.
Expanding on what we missed with sycophancy and the changes weāre going to make in the future:
ā OpenAI (@OpenAI)
3:06 PM ⢠May 2, 2025
Google ā Google has announced a new effort to train 100,00 electrical workers in the United States.
The NotebookLM app is almost here and you can now preregister.
AI Startups ā FutureHouse has released publicly available AI scientist agents.
Future of Work ā Airbnb has quietly rolled out an AI customer service bot in the U.S.
AI in Education ā This Texas pre-K through 8th grade school teaches an entire day of academic lessons in 2 hours with AI.
AI Tech ā GigaIO is showcasing its next-gen AI fabric technology.
1. Judge Questions Metaās Use of Copyrighted Books for AI Training š§āāļø
In a pivotal court hearing, U.S. District Judge Vince Chhabria challenged Metaās claim that training its Llama AI model on copyrighted books qualifies as fair use, highlighting concerns over the potential market impact on original authors. The judge acknowledged the transformative nature of the AIās use but stressed that flooding the market with AI-generated content could harm the value of the original works.
Metaās lawyers argue that requiring licenses would stifle innovation in the rapidly growing AI industry, while plaintiffs insist Metaās actions amount to copyright infringement. This case could set a major precedent for how AI companies handle copyrighted material, influencing creators and businesses navigating content rights in AI development.
2. Amazonās Alexa+ AI Assistant Rolls Out to 100,000 Users š£
Amazon CEO Andy Jassy revealed on Thursday that Alexa+, the companyās new generative AI-powered digital assistant, has reached over 100,000 users since its February debut. While still far from the installed base of 600 million Alexa devices, the rollout is progressing steadily, though key features like third-party app integration and creative tasks remain absent for now.
Jassy acknowledged the technology is āprimitiveā with current AI agent accuracy between 30% and 60%, aiming for a 90% accuracy benchmark with their Nova Act web-browsing agent.
3. Microsoft Prepares to Host Elon Muskās Grok AI on Azure āļø
Microsoft is instructing engineers to prepare for hosting Elon Muskās Grok AI model on its Azure AI Foundry platform, aiming to make it accessible to developers and Microsoftās own apps. This move, driven by CEO Satya Nadellaās ambition to position Azure as the premier AI infrastructure, comes amid growing tensions with OpenAI, Microsoftās primary AI partner, especially after Muskās public fallout with OpenAI and legal disputes.
While Microsoft focuses on hosting Grok rather than training future modelsāan area Musk decided to handle internally after canceling an Oracle server dealāthis deal could be announced at Microsoftās Build conference on May 19th and may strain internal dynamics.
4. Anthropic Ups AI Game with New Integrations and Research Tools š ļø
Anthropic just rolled out two major updates for its Claude chatbot: Integrations, which lets apps like Atlassian and Zapier connect directly to Claude for smarter workflow automation, and Advanced Research, a tool designed to deliver deeper, more thorough reports by crawling internal and external sources.
Available now in beta for select plans, these features push Claude closer to rivals like Googleās Gemini and OpenAIās ChatGPT by enhancing both context awareness and research depth.
5. AI2 Unveils Olmo 2 1B, a Small But Mighty Model š
Nonprofit AI research institute AI2 has just released Olmo 2 1B, a 1-billion-parameter model that outperforms similar-sized offerings from Google, Meta, and Alibaba on key benchmarks like arithmetic reasoning and factual accuracy. What sets Olmo 2 1B apart is its open-source nature, with full code and datasets available, making it accessible for developers running on everyday laptops rather than high-end servers.
This launch underscores a growing trend toward smaller, more efficient AI models that balance power with accessibility, critical for hobbyists and smaller companies looking to leverage AI without massive infrastructure.
6. MITās AI Breakthrough Shrinks Diagnostic Guesswork by 30% š§¬
A new study from MIT researchers reveals a smarter way for AI to assist clinicians by trimming down the size of diagnostic prediction sets without losing accuracy. Using a clever combo of test-time augmentation and conformal classification, this method makes AI predictions more reliable and manageable, helping doctors focus on fewer, more probable diagnoses.
This advance could speed up medical decisions and reduce uncertainty in fields like chest X-ray analysis, where conditions often look alike.
š¦¾How You Can Leverage:
Sumit Gupta legit wrote a book on Tableau.
Now, he forgets basic queries that he ACTUALLY wrote the book on.
Sumit is Notionās Business Intelligence lead, and he joined the Everyday AI show today for some hot takes onā¦.. intelligence.
Like ā are we losing our intelligence as we use LLMs? Is GenAI making us more productive, but dumber?
Itās happening everywhere.
Sumit revealed that AI slashes his dashboard creation time from weeks to hours ā a 5,000% productivity boost. Sounds amazing until the cognitive bill comes due.
At Berkeley, professors are watching student test scores crater to decade-lows while questions during class have virtually disappeared.
Even more alarming?
An AI-generated SQL query that should've cost a few dollars to execute recently racked up a $10,000 bill at Notion because it lacked efficiency optimizations.
These aren't hypothetical scenarios.
These are real-world warning flares from someone who lives on AI's bleeding edge.
Todayās convo is an insightful one. Hopefully itāll make you think.
(You know⦠letās use our brains together!)
Before or after you dive in, hereās three big takeaways from our convo. š
1. AI Can Quietly Eat Your Brain šØ
AIās biggest danger isnāt obvious mistakes. Itās subtle, gradual mental erosion.
Sumit says before AI, complex SQL queries kept his mind sharp.
Manually building dashboards forced him to think critically, analyze deeply, and develop robust solutions. Now? ClaudeAI quickly pumps out perfect calculations, so he rarely engages in deep thinking.
The dangerous result?
He notices a decline in his cognitive agility, becoming slower when facing complex issues. AI hasnāt just streamlined tasksāitās quietly making it harder
(Same.)
Try this
Strategically limit AI use on complex tasks. Force manual problem-solving weekly. Analyze your cognitive agility quarterly by attempting challenging tasks without AI. Verify AI-generated work manually, no matter how reliable it appears. Reflect monthly on tasks reclaimed from AI to ensure mental sharpness.
A recent 2025 Microsoft study revealed heavy AI reliance can deteriorate users' analytical skills within months.
When using GenAI, the study showed that knowledge workers perceived that critical thinking activities generally required less effort compared to when they didn't use AI. For instance, 72% reported "much less effort" or "less effort" for recalling information
2. Donāt Let AI Hijack Your Strongest Skills š”
Sumit warned bluntly: delegating your strongest skills to AI risks losing them permanently.
His SQL expertise faded dramatically after repeatedly outsourcing tasks to AI tools. Remembering basic commands grew harderāan alarming wake-up call.
Sumit argues AI should complement, never replace, your key strengths.
Short-term efficiency gained from AI outsourcing can severely degrade your core skills, creating long-term vulnerability.
Try this
Regularly test core skills without AI support.
Implement firm boundaries around your strongest expertise, keeping them AI-free. Analyze skill retention quarterly through manual task assessments.
Verify periodically that your reliance on AI isnāt creeping upward unnoticed. Reflect monthly on maintaining the right balance between efficiency and essential skill retention.
Proactively guard your expertise to stay sharp.
3. AI Blindspots Cost Companies Big šø
Sumit detailed a shocking reality: blind trust in AI can lead to enormous hidden costs.
He cited a frightening scenario where AI-generated code inadvertently leaked AWS keys, resulting in a huge and costly security breach.
Another costly mistake involved AI-written SQL queries running inefficiently, incurring thousands in unnecessary expenses.
Yikes.
Unchecked AI usage isnāt just risky for your common sense. It can be risky for your dollar bills.
Try this
Regularly test AI-generated solutions for hidden vulnerabilities.
Implement rigid financial controls for AI-driven projects. Analyze weekly spending to catch AI-induced anomalies quickly. Verify security measures rigorously to prevent costly errors. Reflect monthly on the true ROI of your AI investments.
A Y Combinator report showed that 95% of startups rely on AI-driven code.
Sumit's advice? Use AI to amplify productivity. But keep your skills sharp and donāt let an over-reliance on LLMs diminish your brainpower.
Reply