- Everyday AI
- Posts
- Gemini 2.0 – Google's Logan Kilpatrick gives inside scoop on Gemini updates
Gemini 2.0 – Google's Logan Kilpatrick gives inside scoop on Gemini updates
Gemini 2.0 update explained, Altman seeks European Stargate, Google tests Search AI mode, Musk’s DOGE to create an AI chatbot and more!
Everyday AI
February 07, 2025
👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI
Outsmart The Future
Today in Everyday AI
7 minute read
🎙 Daily Podcast Episode: We have Logan Kilpatrick, Senior Product Manager at Google DeepMind, break down the latest updates with Gemini 2.0. Give it a listen.
🕵️♂️ Fresh Finds: Google confirms Gmail attacks, Meta studies human and robot collaboration and Mistral’s in-car AI assistant. Read on for Fresh Finds.
🗞 Byte Sized Daily AI News: Sam Altman seeks European Stargate, Google tests Search AI mode and Musk’s DOGE to create an AI chatbot. For that and more, read on for Byte Sized News.
🚀 AI In 5: The Join Live feature in NotebookLM is straight bonkers. See it here
🧠 Learn & Leveraging AI: One of AI’s brightest minds gave us an exclusive breakdown on how Google’s new Gemini 2.0 updates change the future of work. Keep reading for that!
↩️ Don’t miss out: Did you miss our last newsletter? We talked about Musk's DOGE raising AI concerns with U.S. Education Department, U.S. Gov. wanting to ban DeepSeek federal use and OpenAI unveiling o3-mini thought process. Check it here!
Gemini 2.0 – Google's Logan Kilpatrick gives inside scoop on Gemini updates 👀
One of the smartest leaders in AI is taking us to Gemini school.
Google just released its highly anticipated Gemini 2.0 updates. Logan Kilpatrick is the Senior Product Manager at Google DeepMind and is widely considered one of the leading voices in AI development.
What better way to learn about Google’s groundbreaking model update than straight from the source?
Join the conversation and ask Jordan and Logan questions on Google AI here.
Also on the pod today:
• Rapid Progress in LLM Capabilities 🚀
• Multimodal Features of Gemini 2.0 🔀
• Google's Agentic AI Projects 🧠
It’ll be worth your 32 minutes:
Listen on our site:
Click to listen
Subscribe and listen on your favorite podcast platform
Listen on:
Here’s our favorite AI finds from across the web:
New AI Tool Spotlight – pipl.ai gives you AI-powered cold outreach, Stylar is a controllable AI image editor, and Tablize lets you create easy AI dashboards for your data.
Safe AI — OpenAI Co-founder's startup is fundraising at a $20 billion valuation.
Trending in AI – Ilya Sutskever’s startup is in talks to fundraise at a valuation of at least $20 billion.
Amazon — Amazon is reportedly investing $100 billion into its future AI efforts. Yeah, with a B.
Google – Google has confirmed the Gmail attacks and has urged users not to ignore the FBI warning.
Meta – Meta is studying how humans and robots can collaborate on housework.
AI Tech - Mistral AI and Stellantis are developing an AI-powered in-car assistant.
AI Governance - More than 100 organizations have published an open letter calling for the AI industry to regulate and mitigate AI’s harmful environmental impacts.
AI in Media – Google’s Gemini Super Bowl ad highlights Gemini as an AI tool that you can use everyday.
1. Sam Altman Wants Stargate Europe 🇪🇺
OpenAI's CEO Sam Altman expressed eagerness to establish a Stargate-like AI venture in Europe, marking a significant expansion of AI infrastructure. Speaking at the Technical University of Berlin, Altman emphasized Europe's pivotal role in setting its own AI regulations while announcing OpenAI's new office in Munich.
This expansion is part of a broader European strategy, following earlier openings in Dublin, London, Paris, and Brussels. With an AI summit in Paris on the horizon, where Altman is set to meet European leaders, this initiative could drive substantial growth and innovation across the continent.
2. Google's Search AI Mode Reportedly Started Internal Testing 🧑🔬️
Google is gearing up to enhance its Search capabilities with "AI Mode," an innovative feature reportedly being tested internally. This new tool, integrating with the Gemini chatbot interface, aims to tackle more open-ended questions, promising to revolutionize how users interact with search results.
While AI Mode takes up significant space on the search page, offering comprehensive responses and linking to source sites, it may stir concerns among publishers about traffic impact.
3. Musk's DOGE Dives into AI for GSA 🇺🇸
Elon Musk's Department of Government Efficiency (DOGE) is pushing to create "GSAi," a custom AI chatbot for the US General Services Administration, reports WIRED. This initiative aims to enhance productivity for 12,000 GSA employees and leverage AI for analyzing extensive contract data.
While discussions with Google about its Gemini offering didn't materialize, DOGE is exploring alternatives like Microsoft's GitHub Copilot for coding assistance.
4. SoftBank Invests $40 Billion into OpenAI 💰
SoftBank is on the brink of finalizing a massive $40 billion investment in OpenAI, pushing the startup’s valuation to an eye-popping $300 billion, according to CNBC’s David Faber. This deal is set to unfold over the next 12 to 24 months, positioning SoftBank as the leading backer of OpenAI, surpassing even Microsoft.
Part of this funding will fuel Stargate, a major AI infrastructure project in collaboration with Oracle. This strategic maneuver underscores the escalating competition in the AI sector, as OpenAI continues to rival giants like Microsoft and Google while also addressing challenges from emerging players like China’s DeepSeek.
5. Meta Partners with UNESCO to Boost AI Language Models 🗣
Meta is teaming up with UNESCO to launch the Language Technology Partner Program, aiming to enhance AI speech recognition and translation models by collecting diverse language datasets. This initiative will focus on underserved languages, with partners like Nunavut's government contributing Inuktut language materials.
In a bid to improve transparency and collaboration, Meta is also releasing an open-source machine translation benchmark via Hugging Face, supporting seven languages.
6. U.S. AI Safety Team Misses Paris Summit 🇫🇷
The U.S. delegation to the AI summit in Paris next week won't include the AI Safety Institute, according to sources close to Washington's plans. Vice President JD Vance will lead the group, but key technical voices from the institute, which has collaborated with OpenAI and Anthropic, will be absent.
As global discussions on AI safety continue without them, this decision could impact how America navigates AI innovation amid rising competition from China.
7. Pentagon's AI Leap Forward 🏛
The Pentagon's AI acceleration hub has partnered with BigBear.ai to prototype the Virtual Anticipation Network (VANE), designed to monitor adversarial media and deliver predictive insights. This $1.3 million contract marks a significant step from research to operational use, enhancing the DOD's data-driven decision-making capabilities.
With VANE's deployment across multiple military echelons, the initiative aims to provide strategic context in areas like grey-zone warfare and information operations.
This NotebookLM update is worth your time
Click Image To Play Video 👆
What could make NotebookLM any better?
How about if you could interact with the ‘Deep Dive’ hosts in real-time to learn more from your Notebooks.
Yeah. You’ll wanna learn this one.
Check out today's AI in 5.
🦾How You Can Leverage:
Remember when everyone thought AI hit a wall two months ago?
(Sorry Gary.)
Google DeepMind Senior Product Manager Logan Kilpatrick joined Everyday AI today and straight up shattered conventional wisdom about AI's limitations.
And let’s dig beneath the title for Logan for a second.
After helping launch OpenAI’s developer relationships, he’s now doing the exact thing for Google.
And since he’s joined Google and the Deepmind team, they’ve been on a hot streak.
Today, we not only talked about what’s new in Google’s just-released Gemini 2.0 family of models, but unraveled how it all might impact how we work.
The secret?
It's not about models anymore. It's about the collapse of the cost barrier.
When AI computing drops from $30 to 10 cents per million tokens while getting dramatically better, entire industries reshape overnight.
Slo what’ next?
Let’s dive in for the 3 big takeaways. 👇
1 – The end of traditional software dev? ⌨️
Could AI eliminate the need to write code from scratch?
He revealed that Google's internal debates now center on whether they should even test traditional coding skills in interviews anymore.
(Oh you still team LeetCode or somethin?)
The focus is shifting from writing code to directing AI to build exactly what you envision.
Try This: Next time you need to build something, don't start by writing code.
Start by describing your ideal solution in plain English to an AI like Gemini.
Then iterate on the results. Logan shared that he builds entire features this way now, only jumping in to troubleshoot when things break.
Begin with a small project - maybe that Chrome extension you've been putting off. Seriously. It'll take minutes, not days.
2 – Why Your AI Agent Still Acts Like a Caffeinated Toddler 👶
Could AI agents actually run your entire workflow by next quarter?
Google and other AI tech giants have the individual ingredients for a future of work where you just talk to an AI and everything gets done.
Yet he doesn't have AI agents running his life – yet.
Why?
The blockers weren't what everyone thought. It wasn't about processing power or better prompts.
The real problem?
AI can’t think like humans think. At least not yet.
Logan broke down the three missing pieces: true multimodal processing (seeing and hearing like humans do), advanced human reasoning capabilities (not just pattern recognition), and proactive behavior patterns.
The biggest shift?
When taking questions from our audience, Logan said he hopes to soon wake up to an AI-generated list of high-value tasks his agents want to tackle – not just auto-responses to spam emails.
Try This
Map your task DNA.
Document every repeating task for one week.
But here's the twist – note which tasks require genuine understanding versus simple automation.
Example: Auto-replying to emails? That's basic.
Understanding that an urgent client email at 2 AM needs immediate escalation? That's intelligence.
Group your tasks by complexity level. Those top-tier tasks? They're about to become AI-eligible. And keep your eye on Google’s Project Mariner agent. That dream may not be too far off.
3 – Small Models May Eat Their Parents 👨👩👧👦
Remember when bigger meant better in AI?
Logan detailed how some of Gemini’s new 2.0 Flash models are changing that.
The gap between frontier models and their smaller cousins isn't shrinking by accident. Logan revealed it's actually the biggest AI success story nobody's talking about.
Here's the reality check: Moving from 80% to 85% accuracy sounds borrrrring. But Logan explained that final 5% unlocks an exponentially larger set of real-world applications.
And now those capabilities are trickling down to models that cost pennies to run.
The evidence?
Logan highlighted how domain-specific tasks like coding are already showing better results on optimized smaller models than their resource-hungry parents.
Try This:
Time for a model arbitrage play.
Take your three most expensive AI workflows. Port them to smaller, optimized models.
But here's the key insight Logan shared – don't just test for accuracy. Test for specific capability thresholds your use-case actually needs.
A model that's 98% accurate but costs 10x more than the model that’s 95%? Unless you're doing brain surgery, that extra 3% might be pure waste.
(Yeah, humans in the loop still gotta spot the hallucinations like whack-a-mole)
The future of AI isn't about bigger models or fancier features.
It's about the democratization of capability.
When world-class AI costs less than your Spotify subscription, everything changes. Welcome to the revolution. Hope you brought your keyboard shorties.
Now This …
Let us know your thoughts!
Vote to see live results
What are your thoughts? |
Reply