- Everyday AI
- Posts
- Musk says Grok 3 today, OpenAI lays out plans for GPT-5 & more
Musk says Grok 3 today, OpenAI lays out plans for GPT-5 & more
EU launches open source LLM project, South Korea bans DeepSeek downloads, the NYT unveils new AI tools and more!
Everyday AI
February 17, 2025
👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI
Outsmart The Future
Today in Everyday AI
8 minute read
🎙 Daily Podcast Episode: OpenAI just had a slew of new model updates including GPT-5. Grok 3 is here? And so much more! We’re covering the AI news that you missed last week. Give it a listen.
🕵️♂️ Fresh Finds: Sam Altman speaks on next open source project, Apple Intelligence’s image generation might have bias and researches teaching AI animal emotions. Read on for Fresh Finds.
🗞 Byte Sized Daily AI News: EU launches open source LLM project, South Korea bans DeepSeek downloads and the NYT unveils new AI tools. For that and more, read on for Byte Sized News.
🚀 AI In 5: We’re breaking down the 5 things you need to know about ChatGPT’s Canvas. Actually in 5 minutes. See it here
🧠 AI News That Matters: From Claude’s new hybrid model to updates on Apple and Amazon’s voice assistants, we break down the AI news that matters. Keep reading for that!
↩️ Don’t miss out: Did you miss our last newsletter? We talked about OpenAI board’s unanimous no to Musk, Meta’s AI humanoid robot plans and Siri and Alexa hitting another delay. Check it here!
AI News That Matters - February 17th, 2025 📰
Sam Altman laid out plans for GPT-5.
Elon Musk says Grok-3 is dropping today.
We finally have plans about Claude's new 'hybrid' model.
AI news doesn't slow down. Neither should your company's AI adoption.
Don't waste time. Join us on Mondays as we bring you the AI News that Matters.
Join the conversation and ask Jordan any questions on AI here.
Also on the pod today:
• Grok 3 launch 🚀
• YouTube Veo 2 Integration 🎥
• Apple and Amazon voice assistants 🗣
It’ll be worth your 51 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 – Kvistly provides AI quizzes for better training, Cove is a visual AI workspace for thinking and Eddie AI is ChatGPT for video editing.
OpenAI – Sam Altman posed a question on X about OpenAI’s next open source project.
for our next open source project, would it be more useful to do an o3-mini level model that is pretty small but still needs to run on GPUs, or the best phone-sized model we can do?
— Sam Altman (@sama)
1:53 AM • Feb 18, 2025
Apple – Apple Intelligence’s AI image generator seems to have some bias issues.
AI in Society – South Korea is looking to buy 10,000 GPUs for its national AI computing center.
AI Research – Researchers are training AI to interpret animal emotions.
AI Governance - Here’s what the US’ first AI copyright ruling could mean for IP law.
AI Models – A study shows that older AI models seem to show signs of cognitive decline.
1. Europe's AI Ambitions Spark with OpenEuroLLM Initiative 🇪🇺️
Europe has launched the OpenEuroLLM project, aiming to create "truly" open-source language models covering all EU languages, according to TechCrunch. Co-led by Jan Hajič and Peter Sarlin, this initiative brings together over 20 organizations to bolster Europe's AI capabilities, despite its relatively modest budget of €37.4 million.
The project builds on previous efforts like the High Performance Language Technologies (HPLT) project and aims for initial model releases by 2026.
2. South Korea Halts DeepSeek Downloads Over Privacy Concerns 🚫
South Korea has pulled the plug on new downloads of China's DeepSeek AI chatbot, citing privacy law compliance issues, according to the country's Personal Information Protection Commission. Despite skyrocketing to app store fame with a million weekly users, DeepSeek now faces global scrutiny and bans, including from Taiwan and Australia, over privacy and security concerns.
Current users remain unaffected, but the app's swift rise has raised alarm bells worldwide, challenging America's AI dominance and sparking debates on data protection.
3. NY Times Embraces AI Tools for Newsroom Innovation 📰
The New York Times is diving into the AI waters, rolling out new tools like Echo to help its editorial team streamline tasks, according to Semafor. This move aims to enhance journalistic capabilities by generating SEO headlines, summarizing articles, and brainstorming interview questions. However, the Times is treading carefully, outlining strict guidelines to avoid copyright issues and protect source confidentiality.
This development comes amid a legal spat with OpenAI over alleged copyright violations, highlighting the complex dance between innovation and intellectual property rights.
4. Apple Intelligence Heads to Vision Pro Soon 🕶️
Apple is gearing up to introduce its Apple Intelligence suite, including Writing Tools, Genmoji, and Image Playground, to the Vision Pro headset by April, reports Bloomberg's Mark Gurman.
This eagerly awaited update could revolutionize the mixed reality experience by enriching it with AI capabilities, though a revamped Siri isn't part of the initial rollout due to ongoing engineering challenges. For developers, a sneak peek at the visionOS 2.4 software update might be available as soon as this week.
5. Mistral Unveils Arabic AI Model Saba ✨
Paris-based AI startup Mistral has launched its latest model, Mistral Saba, aimed at excelling in Arabic interactions. This specialized model, with 24 billion parameters, surpasses its general-purpose counterpart, Mistral Small 3, in handling Arabic content and even performs well with South Indian languages like Tamil and Malayalam.
This strategic move highlights Mistral's focus on the Middle East market, potentially attracting regional investors amid geopolitical shifts. Accessible via API and deployable on-premise, Saba is poised to support industries needing localized conversational AI, marking a significant step for Mistral in the global AI arena.
ChatGPT Canvas: 5 things to know in 5 minutes
ChatGPT’s Canvas mode is kinda underrated.
We’re going to tell you how it compares to Claude Artifacts, the bigger AI picture that Canvas plays a role in and 5 things you need to know about this ChatGPT mode.
We keep it actually short this time shorties.
Find out in today's AI in 5.
New AI models. ✨
New AI models everywhere.
Just because it’s a U.S. holiday doesn’t mean the AI updates stopped.
OpenAI, Anthropic and xAI are all on the brink of new AI models, Meta and Apple are going robotic, and Perplexit’s playing the Deep game.
Don’t spend hours a day trying to keep up with AI developments and what they mean for your biz.
Instead, we do the heavy lifting daily and on (most) Mondays, we bring you the AI news that matters.
Let’s get it poppin y’all 👇
1 – OpenAI Rejects $97.4B Musk Takeover Bid 🙅
The OpenAI board unanimously shot down Elon's acquisition attempt faster than you can say "non-profit status." Their statement on X emphasized keeping AGI beneficial for humanity.
Sam Altman's response was chef's kiss perfect: offering $9.74B for Twitter/X – which is actually closer to its real value than Musk's $44B purchase price.
Lol
What it means:
This whole circus is Elon's way of complicating OpenAI's non-profit to for-profit transition. The "$97.4B bid" is just theater to establish a market value OpenAI may have to deal with later.
Not real.
2 – Grok 3 Launches Tonight 👀️
The latest update brings reasoning and deep research modes.
XAI's latest model runs on a behemoth: 100,000 NVIDIA H100 GPUs creating the world's largest AI training center.
That's 10x more computing power than previous versions.
They've raised $6B from NVIDIA and AMD, using synthetic datasets and real-time self-correction mechanisms.
What it means:
No serious business will touch this with a 10-foot pole. Making X/Twitter data a big part of your LLM is like building a security system trained on TikTok pranks.
Won’t be taken seriously. Though, we do think it’ll benchmark well, possibly as a Top 2-3 model on LM Arena.
3 – Apple and Amazon’s Voice AI Delays 🗣️
Want smarter AI voice assistants? Keep waiting.
Siri's big AI upgrade? Pushed to May or later. The bugs are winning. Simple features like app control and screen context understanding are giving Apple fits.
Amazon's LLM-powered Alexa is stuck in testing until March 31st. They're struggling with features ChatGPT Voice knocked out months ago.
What it means:
The voice AI pioneers are now dead last in the race. They pioneered this tech a decade ago and now can't even match what OpenAI and Google are doing now.
Awkward.
4 – Perplexity Launches Deep Research 🔍
Perplexity has launched a new feature in Deep Research they’re hoping can compete with OpenAI and Google’s Deep Research tools.
Their new research mode processes complex queries in 2-5 minutes. Free users get 5 daily queries, Pro users get 500 (though you'd need to be a research cyborg to hit that limit).
They're already negotiating revenue deals with Time and Fortune while other media companies are lawyering up.
What it means:
THIS is the AI feature that'll convert enterprise skeptics. (Whether that’s from Forget pretty pictures – this is actual business value that makes companies money.
5 – YouTube Veo 2 Slow Release 🐢
Google's beast of a video model is live... but only in YouTube Shorts.
Limited to US, Canada, Australia, and New Zealand users.
Access it through the Shorts camera's green screen option for AI backgrounds or standalone clips.
What it means:
Google's got a Ferrari but only lets you drive it in the parking lot.
Meanwhile, every AI company in China is pushing video boundaries.
Are you gonna use Veo2?
6 – GPT-4 Gets Ninja Update 🥷
OpenAI just stealth-dropped major improvements to GPT-4o. No big blog post. No changelog. NADA.
Just a Sam Altman tweet.
we put out an update to chatgpt (4o).
it is pretty good.
it is soon going to get much better, team is cooking.
— Sam Altman (@sama)
5:33 PM • Feb 15, 2025
But users are reporting spicier writing style, less censorship, and way better benchmarks.
The proof?
GPT-4o immediately shot back up to a three-way tie for first place in the LM Arena, going head-to-head with Gemini 2.0 Pro and Flash Thinking.
What it means:
While other companies write blog posts about every minor tweak, OpenAI's out here dropping game-changing updates like they're nothing.
Also, thousands of developers probably woke up to broken systems this morning because SOMEONE doesn't believe in changelogs. 👀
7 – OpenAI Drops GPT-5 Bombshell 🤯
Speaking of low-key tweets, Sam Altman just casually revealed their GPT plans on Twitter (codename: Orion).
OPENAI ROADMAP UPDATE FOR GPT-4.5 and GPT-5:
We want to do a better job of sharing our intended roadmap, and a much better job simplifying our product offerings.
We want AI to “just work” for you; we realize how complicated our model and product offerings have gotten.
We hate… x.com/i/web/status/1…
— Sam Altman (@sama)
7:17 PM • Feb 12, 2025
But here's the REAL strawberry juice: GPT-4.5 will be their final non-chain-of-thought model. And, they won’t release o3 as a standalone model.
Instead of the current "pick your model" confusion fest, they're building a unified AI system under the GPT-5 umbrella. So, o-series reasoners and GPT level models working together.
Free tier gets unlimited GPT-5 access, paid users get higher intelligence levels, and pro users get the REALLY good stuff.
Apparently.
Plus, they just added Google-style autocomplete suggestions to ChatGPT. Interesting timing, right?
What it means:
OpenAI isn't playing "catch up with Claude" or "beat DeepSeek."
Stop that.
Every single move – free ChatGPT search, autocomplete, unified AI system – is aimed straight at Google's.
This isn't about AI leadership; it's about controlling where people start their day.
8 – Claude’s New Hybrid Model ✨
Not like a car.
Anthropic's dropping a new model in weeks that switches between deep reasoning and quick responses. According to reports, early tests show it crushing OpenAI's o3-mini high on coding tasks.
Developers get a sliding scale to manage those beefy reasoning costs.
What it means:
Like 7 months without a major update is like nine years in AI time. Great for developers, but without real-time internet access, it's still bringing chopsticks to a gunfight.
9 – Meta and Apple’s Robot Adventures 🤖
According to reports, both Meta and Apple are making a hard pivot into AI humanoids.
Yah?
Meta's building Android-for-robots software in their Reality Labs. Apple's working on some mysterious tabletop robotic arm thingy.
Both companies poached self-driving car talent to lead these projects.
What it means:
They can't make Siri understand "remind me to buy milk" but sure, let's build robots.
10 – Adobe’s Commercial Video AI Debut 🎥
Adode’s new AI video generator is publicly available in beta, months after it was teased.
Their Firefly app now includes text-to-video and image-to-video at 24fps. Current clips max out at 5 seconds, with faster/lower-res and 4K versions in development.
Training is reportedly mainly on licensed content. Plans range from $9.99 to $29.99 monthly.
What it means:
Finally, an AI video tool that won't give your legal team heart palpitations. Or at least out of the box.
The 5-second limit though……
Now This …
What are your thoughts?
Vote to see live results
How much trust do you have in our reporting? |
Reply