Claude 3.5 is here and surpasses GPt-4o + OpenAI might become a for-profit company

You're receiving this email because you registered to one of our workshops. You can unsubscribe at the bottom of each email at any time.

We are now live with our new webinar schedule for the next month. You can sign up for all upcoming webinars here. We have events for everyone, from beginners to advanced builders!

This week, Anthropic took the stage and released Claude 3.5 Sonnet + Artifacts, and promised that Claude 3.5 Haiku and Opus are coming later this year. OpenAI is seemingly considering turning into a for-profit company. Runway announced a new generative AI video model, “Gen-3.” NVIDIA briefly became the world’s most valuable company, and Ilya, the former Chief Scientist at OpenAI, started his own AI company named SSI - Safe Superintelligence Inc.

As a reminder, we’re now welcoming partners that want to build AIs for their clients. Sign up for extra support, training resources, and more here.

🗞️ Industry news

1. Anthropic releases Claude 3.5 Sonnet: new default model in Claude chat, and it beats GPT-4o in most evals
   
   

   

   Copyright Anthropic

   
   
   Anthropic has launched Claude 3.5 Sonnet, setting a new standard in AI intelligence. Claude 3.5 Sonnet surpasses competitors like GPT-4o, its own big brother Claude 3 Opus, in numerous evaluations, maintaining the speed and cost efficiency of Claude 3 Sonnet. All while being faster and cheaper.
   
   Claude 3.5 Sonnet operates at twice the speed of Claude 3 Opus, but retains the same old context window set to 200k tokens.
   
   A new feature, Artifacts, has been introduced on Claude.ai, allowing users to generate and interact with content like code snippets and documents in a dynamic workspace, enhancing real-time collaboration.
   
   

   

   Evals posted by Anthropic

   
   
   From my initial testing:

   • Claude 3.5 Sonnet is significantly better than Claude 3 Sonnet. The difference is palpable, and it also doesn’t reject prompts as often as before; 

   • The new model excels at coding tasks for non-coders. While the raw coding abilities of GPT-4o in something like GitHub Copilot might be better, Claude 3.5 Sonnet seems to have a soft spot for non-devs looking to build web apps. It generated forms, mini games, and website blocks for me in one prompt; 

   • Its vision capabilities are EXCELLENT, I’d say the new Claude 3.5 model is much better than GPT-4o at tasks like coding a website or section from a screenshot; 

   • It retains the great creativity and copywriting skills of the Claude family, making it ideal for long-form content writing; 

   • It’s the first Claude model I tried that is pretty much capable of sticking to a certain word count. For example, for the first time, if you ask for a 2,000 words article you might actually get one in 1 prompt.

   
   Anthropic plans to release Claude 3.5 Haiku and Claude 3.5 Opus later this year, alongside new features and enterprise application integrations.
   
   Claude 3.5 is coming to MindStudio later today or next week. Stay tuned!
   
   

2. Runway introduces Gen-3 Alpha: impressive, but still very limited
   
   Runway has launched Gen-3 Alpha, their latest AI model for video generation, offering significant improvements in speed and quality over Gen-2. This new model can create video clips from text descriptions and still images, providing fine-grained control over structure, style, and motion. Gen-3 Alpha:
   

   • Generates expressive human characters with diverse actions, gestures, and emotions.

   • Interprets various styles and cinematic terminology, enabling imaginative transitions and precise key-framing.

   • Supports high-resolution 5- and 10-second clips, generating them in 45 and 90 seconds, respectively.

   
   Despite its advancements, Gen-3 Alpha has limitations, such as maxing out at 10-second footage and struggling with complex interactions.
   
   Gen-3 Alpha is rolling out for Runway subscribers, mostly for enterprise customers and creative partners.
   
   

3. NVIDIA shortly became the world’s most valuable company
   
   

   

   
   Nvidia has shortly been the world's most valuable company, reaching a market valuation of $3.34 trillion after its share price hit an all-time high. The stock has nearly doubled in value this year, ending Tuesday at $136, up 3.5%, surpassing both Microsoft and Apple.
   
   The company has since lost its spot to Microsoft after a massive drop. Analysts from Wedbush Securities foresee a race to a $4 trillion market cap among Nvidia, Apple, and Microsoft.
   
   Nvidia's meteoric rise is remarkable, considering its stock was worth less than 1% of its current value eight years ago. Initially known for its graphics card rivalry with AMD, Nvidia has recently capitalized on the demand for AI chips and the Bitcoin mining boom of 2020.
   
   At the time of writing, NVIDIA’s valuation stands at $3.1t. Considering its earnings and revenue metrics, the valuation is a stellar multiplier, higher than most other tech companies.
   
   What do you think? Do you believe NVIDIA deserves to be the most valuable company in the World? Let us know on Discord!
   

In other news, there are some rumors Sam Altman might be considering a for-profit model for OpenAI, and his old colleague Ilya started his own Superintelligence company aimed at developing safe and ethical AI. And yes, their website is an HTML document with no styling.

🔥 Product Updates

MindStudio turned multimodal last week. Users now have access to new blocks:

• Generate image: Use DALL-E 2 or DALL-E 3 to generate images that stun your audience, without needing your own API key.

• Display text: Use this block to display values stored in variables. This block will replace the current send message block when the sender is set to “system.”

• Generate text: Replaces the current “Send Message” block.

• Text to speech: Use audio models to generate AI outputs that speak to you.

• Analyze image: Tap into the power of vision models like GPT-4 Vision to read images and add the result to your context.

If you haven’t played with the new blocks, I highly suggest you do! You can build great experiences for yourself, your team members, and your users.

If you don’t want to start from scratch, make a copy of our Demo app in your workspace to experience all the new blocks and see how to build complex workflows that include them all.

Coming soon:

• Voice controls for text-to-speech models. Right now, all models are loaded with the default voice and settings. We want to give you access to controls to customize these.

• More intuitive UI to manipulate multiple output types. Right now, our “?” icon for the block “Display Text” will provide more information on what you can add within its borders (more below);

• Refinements to the Chat terminator to work with the new multi-modal capabilities;

• New Claude 3.5 Sonnet model;

• … and Groq! 🔥 The company is finally ready to ship its Enterprise plan. We will be among their first customers, and we’re in touch with them to bring Groq to MindStudio very soon, probably around the first week of July.

💡 Tip of The Week

Previously accessible by custom functions only, GPT-4o Vision and GPT-4 Turbo Vision are now available as a standalone block in MindStudio - no API required.

The “Analyze Image” block takes in two inputs:

1. Prompt: what do you want to check in the image, and what do you want the model to give you? For example, do you want to describe the image, find something in it, or label it?

2. Image URL: this has to be a publicly accessible URL containing your image or a variable in your workflow that links to an image URL. Our new “Image Upload” input type does exactly that: takes in an image and stores it in a URL. You can then use that in the Image URL field of the “Analyze Image” block.

And one output:

1. Output variable: this is where the query result is saved. The response from GPT-4o Vision or GPT-4 Turbo won't show up directly to the user; it will only be stored as a variable.

Remember you need a “Display Text” block afterwards to display the content of the output variable. Otherwise, your user will never see the response.

If you need the response to inform another prompt, then you can use the variable within a “Generate Text” block before sending it over to an LLM.

🤝 Community Events

If you want to hangout with our team, we usually host a Discord event every Friday @ 3PM Eastern. Join our Discord channel to keep up to date with the hangouts - our entire team is active there.

You can register for upcoming events on our brand new events page here.

Our new webinar series is up on there as well, with the following on-demand webinars:

• MindStudio for Partners: Level Up Your Agency With AI Workflows

• Customize AI Applications With Your Data

Plus, we have new weekly and bi-weekly events:

• How To Build Your First AI Assistant (the best webinar for beginners!)

• Which Models to Choose for Your AI

• First Steps to Bring AI Into Your Workflow

These webinars are our new hybrid events. The video itself is pre-recorded for the best possible experience and to make them more concise, while the chat and interactive widgets are live, creating a line of communication with our team, who monitor them in real time during working hours.

Thank you for being an invaluable member of our community, it’s always great to see many of you join multiple workshops 🔥

🌯 That’s a wrap!

Stay tuned to learn more about what’s next and get tips & tricks for your MindStudio build.

You saw it here first,

Giorgio Barilla
MindStudio Developer & Project Manager @ MindStudio