- Everyday AI
- Posts
- Deep Research Throwdown: Perplexity vs. Google vs. OpenAI
Deep Research Throwdown: Perplexity vs. Google vs. OpenAI
Deep Research comparison, Google’s AI co-scientist system, Microsoft’s AI gaming model, Meta’s global AI infrastructure plan and more!
Everyday AI
February 19, 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: Deep Research tools are on the rise and for good reason. We break down some of the top Deep Research tools to see which is best for you. Give it a listen.
🕵️♂️ Fresh Finds: SSI to raise $1 billion, Apple’s new AI iPhone and Gemini Deep Research now available on mobile. Read on for Fresh Finds.
🗞 Byte Sized Daily AI News: Google’s AI co-scientist system, Microsoft’s AI gaming model and Meta’s global AI infrastructure plan. For that and more, read on for Byte Sized News.
🚀 AI In 5: This ChatGPT feature in DALL-E 3 can help change the way you make AI images! See it here
🧠 Learn & Leveraging AI: Wondering what all the Deep Research buzz is about? We break down how Deep research works and compare the best options out there. Keep reading for that!
↩️ Don’t miss out: Did you miss our last newsletter? We talked about xAI releasing Grok 3, Google boosting Chrome security with AI and OpenAI’s plans to avoid takeovers. Check it here!
Deep Research Throwdown: Perplexity vs. Google vs. OpenAI 🔍
Deep Research tools are gonna be the second wave of GenAI.
Granted, they'll be a MUCH smaller wave than the ChatGPT moment of 2022. Yet, before AI Agents take off flying, we're going to see a mass adoption to these new Deep Research Tools.
But which ones are good?
Is Google's the winner? Is OpenAI's in-depth researcher untouchable? Or is Perplexity's free and speedy option the way to go?
We go in-depth on each.
Join the conversation and ask Jordan questions on Deep Research here.
Also on the pod today:
• Tasks Suitable for Deep Research Tools 🕵
• Comparison of Deep Research Tools 🥊
• Challenges in Using the Internet for Research 🌐
It’ll be worth your 52 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 – Polymet is an AI product designer, Magic Inspector is an AI web test automation platform and Nuvio provides AI-powered financial management.
Trending in AI – Ilya Sutskever’s Safe Superintelligence AI startup is close to raising $1 billion.
Apple – Apple’s new iPhone 16e adds AI and is launching February 28th.
Google - Gemini Deep Research is now available for mobile users.
Google Meet’s Gemini-powered note-taking will now generate a checklist of next steps.
Future of Work – Fiverr is now letting freelancers sell access to AI models trained on their work.
AI Models - Replicate has made its Flux code open source and is adding some updates.
AI Law – U.S. law firm Morgan & Morgan has warned its lawyers about AI's potential to generate fake case law.
1. Google Unveils AI Co-Scientist System 🧑🔬️
Google Research, DeepMind, and Google Cloud AI have unveiled an AI co-scientist system, designed as a collaborative tool for researchers. According to their recent report, this multi-agent AI system assists scientists in generating novel research hypotheses and experimental protocols, significantly enhancing the speed and quality of scientific discovery.
Validated through real-world experiments in areas like drug repurposing and antimicrobial resistance, the AI co-scientist demonstrates profound potential to transform biomedical research.
2. Microsoft Unveils Muse AI for Gaming 🎮
Microsoft has just announced a major breakthrough in AI for gaming with its new Muse AI model. This pioneering generative AI can create game environments by interpreting visuals and player actions, offering a glimpse into the future of game development. While it’s not ready to generate full games, Microsoft envisions Muse enhancing creativity and preserving classic games for modern hardware.
The company emphasizes that AI is here to support, not replace, human artistry in game creation, a crucial reassurance amid industry concerns about job security.
3. Meta's AI Cable Infrastructure 🛠️
Meta has revealed its ambitious Project Waterworth, a colossal endeavor to lay over 50,000 kilometers of subsea cable. This multi-billion-dollar project aims to bolster AI infrastructure globally, with a route stretching from the U.S. coasts to Brazil, South Africa, India, and Australia. Designed to endure depths of up to 7,000 meters and withstand maritime hazards, this will be the world’s longest undersea cable.
According to Meta, this infrastructure leap could revolutionize data connectivity, potentially accelerating AI advancements and offering new opportunities for tech-driven growth worldwide.
4. EU's AI Act Sparks Copyright Concerns 🧑⚖️
The European Union is facing criticism over its AI Act, which cultural organizations argue leaves creatives vulnerable due to a "devastating" copyright loophole. Architect of EU copyright law, Axel Voss, expressed frustration at the lack of strong copyright protections for writers and musicians, highlighting that the exemption for text and data mining in the AI Act was meant for limited private use, not large-scale corporate exploitation.
Despite calls from cultural groups for more transparency and protection, the European Commission has yet to respond with concrete measures. This gap places individual creatives at a disadvantage in protecting their work from being used by AI systems, posing a systemic risk to their industry.
5. Humane's AI Pin Bows Out as HP Steps In 🔄
Humane has announced the sale of most of its assets to HP for $116 million, effectively ending sales and support for its AI Pin. Starting February 28th, the AI Pin will lose cloud-dependent features, leaving users with offline capabilities like checking battery levels.
The AI Pin faced harsh criticism upon its launch and struggled with low sales, prompting Humane to seek a buyer. This acquisition will see Humane's team, including founders Imran Chaudhri and Bethany Bongiorno, join HP to spearhead AI integration in a new division called HP IQ, focusing on advancing AI across HP's product line.
6. Mira Murati Launches New AI Venture, Thinking Machines Lab 👀
Former OpenAI CTO Mira Murati has unveiled Thinking Machines Lab, a startup aiming to revolutionize AI customization and understanding. The lab seeks to bridge gaps in current AI capabilities by developing multimodal systems that can adapt to human expertise across various domains.
With AI safety as a core focus, the company plans to share best practices and support external research, ensuring responsible AI use.
7. U.S. AI Safety Institute Faces Major Staff Cuts ✂️
The Trump administration is set to make significant cuts at the US AI Safety Institute (AISI), with 497 roles reportedly on the chopping block, according to Axios. This shake-up comes on the heels of the AISI being sidelined from a major AI summit and the resignation of its director, Elizabeth Kelly, under political pressure.
The cuts are expected to heavily impact AI safety and semiconductor production efforts, raising questions about the administration's strategy to maintain an AI edge over China. The timing of these actions highlights a shift in priorities, focusing less on regulation and more on achieving "AI dominance."
ChatGPT's in painting feature in DALL-E
ChatGPT’s in painting image editing feature inside of DALL-E is a great way to edit your generated images to your desire.
We’re breaking down how the feature works and putting it to work to see if it’s any good.
Check out today's AI in 5.
🦾How You Can Leverage:
Three AI giants.
Three identical product names.
Back to back to back.
Yikes.
Google, OpenAI and Perplexity each launched "Deep Research" tools that promise to end tab-hopping hell forever.
The timing? That’s gotta mean this is a shockingly obvious next area of GenAI investment.
One hallucinated NEURAL IMPLANTS and QR CODE MIND CONTROL.
Another found a basketball reference from TWENTY YEARS AGO mentioned once online.
The third (kinda) costs $2 PER SEARCH but never lies.
On today’s show, we gave the Deep Dive treatment to all things Deep Research.
Let's get surgical.
1 – OpenAI’s Truth Machine (That Costs a Fortune) 😶🌫️
OpenAI's deep research runs on a specialized o3 finetune that makes ChatGPT look like a toy.
Goo-goo-chat-chat.
It takes forever (5-30 minutes). Costs a ridiculous $200/month for 100 queries for the Pro plan. (Although the plan comes with a lot more than that.) And it only? visits 15-30 websites in real-time.
But guess what?
It’s like never wrong. A weird anomaly in the world of hallucinations everywhere AI. It’s like a truth machine at times.
When researching our podcast, it found details from obscure interviews nobody remembers giving.
Zero hallucinations.
The reasoning is terrifyingly good - it connects dots across disconnected sources like some kind of digital savant.
Like that research intern you always wanted to hire, but they just quit week 2.
Before starting, OpenAI’s DR asks clarifying questions to prevent wasting precious query allowance.
Smart.
For high-stakes research where mistakes would be catastrophic?
Worth every penny.
Try This:
Run a test query about your company with information you know exists online but is hard to find. When it passes that test, try a mission-critical query worth $2. You'll never go back to manual research for decisions that actually matter.
TBH, we think watching this episode covering OpenAI’s Deep Research will be the best 28 minutes you spend today. (If you watch on 2x! lolz)
2 – Google’s Source-Devouring Crawler 🎯
Google took a completely different route.
Brilliant move.
Instead of visiting sites in real-time, Gemini 1.5 Pro leverages Google's ENTIRE CACHE of pre-crawled pages. (Yeah, they kinda built that thang up after dominating search for 20 years.)
Result? It processes 50-500 sources (sometimes 1,300+) in 3-10 minutes flat.
Hot dang Gina.
Every session starts with an editable research plan. The reports are MASSIVE - typically 2,000+ words of densely cited analysis.
Our test found a twenty-year-old basketball reference mentioned exactly once on the internet. It made just two minor errors, inventing a fictional co-founder and briefly making us college professors.
(Pop Quiz coming soon?)
No file uploads.
No clarifying questions.
But for competitive intelligence requiring MASSIVE breadth? Untouchable.
Try This:
Create this exact template: "Analyze [Competitor] focusing on: 1) Revenue model evolution, 2) Recent tech stack changes, 3) Customer acquisition shifts, 4) Executive backgrounds, 5) Product roadmap signals from public statements."
Run monthly on competitors to spot patterns invisible through conventional research.
We’re HUGE fans of this tool.
We already gave Google’s version the full run-through, and named it one of our top AI tools/features of 2024.
3 – Perplexity’s Lighting-Fast Fiction Generator 🧍
Speed: INSANE.
Cost: FREE!
Accuracy: YIKES. (sometimes)
Perplexity delivers results in 1-3 minutes. Offers five free queries daily, unlike the paid-only approaches from OpenAI and Google.
Processes 20-200 sources with an undisclosed reasoning model.
But it makes stuff up. Constantly.
It claimed our episodes contain "eerie vocal modulation" at exactly "the seventeen minute mark." Created fictional conference presentations with "subliminal messaging in slide decks." Even hallucinated a "neural lace interface" in our "leaked internal roadmap."
Yo bro where did you even come up with that?
Pure fiction delivered with absolute confidence.
The company promises improvements. Our follow-up tests showed slight progress. But not enough.
For anything requiring factual reliability?
Hard pass.
Try This:
If you're desperate for speed, use this prompt: "Research [topic] with these requirements: Only include verifiable information with specific sources. Mark anything uncertain as 'Unverified.' Prioritize official documentation."
Then fact-check EVERYTHING before using it.
Also, check our full Perplexity Deep Research Episode.
Our take?
We get how free AI tools can be all the rage.
So Perplexity’s 5 free Deep Research queries a day can be tempting. Before you use them for any business use-case, rigorously test it first on an area you’re a legit Einstein in. Otherwise you risk copying-and-pasting nonsense.
While we can’t recommend Perplexity’s Deep Research now, it has a high ceiling if it can be trained on the truth.
Google’s Deep Research is a legit gem and produces fantastic reports.
But our fave?
OpenAI's Deep Research is already human-level analyst. We see future versions of this tool shaking up the management consulting industry.
(Yeah, it’s already better than most junior analysts in the hands of someone that knows what they’re doing.)
For now?
We're using OpenAI when accuracy matters most, Google for broad competitive scanning, and approaching Perplexity with extreme caution.
The tab nightmare is ending. Knowledge work is changing forever. Adapt or get left behind.
⌚
Numbers to watch
$75 Million
Augury has raised $75M at $1B+ valuation for its AI that detects malfunctions in factory machines.
Now This …
Let us know your thoughts!
Vote to see live results
How many AI newsletters do you read? |
Reply