New AI Models in MindStudio + Conditional Logic in Prompts

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

MindStudio is on a roll. This week, we worked on a new learn page, a batch of over 10 models to ship in the platform, and conditional logic. The new models will be available on Monday. Conditional logic is available right now, in all accounts.

The industry wasn’t sleeping either. Microsoft is working on a new AI model, OpenAI announced an official partnership with Stack Overflow, some rumors hint at OpenAI releasing an alternative to Perplexity, and Mistral AI is raising a new fund at $6b valuation.

Keep reading more, and don’t forget next week we’ll hear from Google at Google I/O 2024 😉

🗞️ Industry news

1. Microsoft is working on a new proprietary AI model, MAI-1

   Microsoft is stepping up its game in the AI field with the development of a new AI language model, MAI-1, aimed at competing with major players like Google and OpenAI. Although the specific applications of MAI-1 are yet to be determined and will depend on its performance, Microsoft may showcase this model at the upcoming Build developer conference.
   

   The MAI-1 model is a significant leap from Microsoft's previous efforts, being described as "far larger" than earlier, smaller, open-source models. The new model is expected to have around 500 billion parameters, compared to OpenAI's GPT-4 with one trillion parameters and Microsoft's own Phi-3-mini with 3.8 billion parameters.
   

2. OpenAI partners with Stack Overflow to help developers ship faster
   OpenAI and Stack Overflow have entered into a new API partnership designed to enhance the capabilities of developers by integrating OpenAI's advanced language models with the vast technical knowledge base of Stack Overflow.
   
   The partnership will see OpenAI use Stack Overflow's OverflowAPI to refine its AI models and improve user experience, while Stack Overflow will leverage OpenAI's models in developing its OverflowAI. The first set of integrations and new capabilities resulting from this partnership is expected to be available in the first half of 2024.
   

3. Is OpenAI looking to compete with Google Search?
   Speculation is mounting that OpenAI is poised to unveil a new ChatGPT-powered search engine in May, challenging Google's search dominance. The domain search.chatgpt.com was recently registered, and various online clues, including a cryptic "Not found" message on the subdomain rather than a standard error, suggest an imminent launch.
   
   It’s worth noting AI search is nothing new at this point. Perplexity AI, a search unicorn, and Bing Search have been offering AI citations, AI search, and more for the last year. New browsers like Arc also include AI search in features like voice search.
   
   However, OpenAI undoubtedly boasts a certain authority in the field, and has one of the most used tech products of all time (ChatGPT) to foster adoption.
   

4. Mistral is raising at a $6b valuation. SoftBank confirms they won’t be joining the round
   Mistral spearheaded the open source AI revolution. With their headquarters in France, they already broke most records and managed to get to Unicorn status in very little time. This is especially hard in the EU, where securing funding for AI startups is much harder.
   
   The team appears to be looking for $600M, but is finding it harder than usual to close. According to TechCrunch, DST, along with General Catalyst and Lightspeed Venture Partners, are all looking to be a part of this round.

Also, Google I/O is just around the corner. We’ll probably see many, many AI updates during the event, and possibly some updates to the Gemini Pro models. Is it time for an even larger context window?

After all, Google said they were working on a 10m context window when Gemini 1.5 Pro was first released… 🤔

🔥 Product Updates

This week, MindStudio released several new features and is preparing to launch even more next week.

• Conditional Logic: we've introduced conditional logic in Send Message blocks, allowing you to request optional user inputs and utilize them based on specific conditions. For instance, you can ask for the user and company name, then display only one of them based on a condition you define in the prompt;

• New Embed Options: you can now choose to minimize the thread panel or disable threads entirely when embedding a MindStudio application on your website;

• New Onboarding: there’s a new platform and editor onboarding sequence for new users. This makes it easier to pick up the product and start building your AI-powered applications;

Together with Conditional Logic, we’re also shipping new AI models next week. The list is massive and includes:

• Gemini 1.5 Pro: one of our top requested model since release, Gemini 1.5 pro with its massive 1m context window will soon pop up in your model settings;

• Llama 3 8b & 70b: the new Llama 3 family shocked the world. The models are incredibly capable, very cheap, and can match or surpass performance of similar models in their category - including closed source models;

• Cohere Command & Command Plus: set to be the fastest models in our platform, these rank very high in the chatbot arena and can output tokens at lightning speed;

• Perplexity AI: many of you asked for live search, and the API from Perplexity is a step in that direction. The models have up-to-date information about the world thanks to their backend, one of the most popular AI search engine at the moment.

That’s not all. In the coming weeks, you’ll get access to multimodal capabilities in your workflow. The new batch comes with a full restructuring on our end, which makes adding new models on our end a breeze.

Soon, you will be able to use:

• DALL-E 3: image generation without an API. Simply select the model, add the prompt, and get an HD image;

• Whisper: we heard you… you want speech-to-text capabilities. We’re working on audio transcription so you can interact with the models with more inputs than just text;

• GPT-4 and Gemini Vision: you’ll be able to unlock the full power of these multimodal frameworks with image and file inputs.

• Eleven Labs: let your AI workflow speak its mind with human-sounding, hyper-realistic AI voices.

While these might take more troubleshooting and testing on our end, we’re prioritizing the image models and plan to release DALL-E 3 ASAP, likely this month. GPT-4 Vision and Gemini Pro Vision will be next.

It’s an exciting time for MindStudio, and we're sure you’ll find creative ways to use the new models and conditional feature.

💡 Tip of The Week

How you structure your prompt in the send message block makes a massive difference in the output you can expect from the model.

I suggest structuring your prompt like this:

• Act as … → this is usually your Main prompt (for the flow, not the send message block). Here, you can specify what the AI should act as, e.g. an accountant, a teacher, a marketer;

• Scope → the first line in the send message prompt should outline the scope. What do you want the AI, acting as the persona in the main prompt, to do? Should it summarize a doc, simulate a sales call?

• Formatting → how should the AI format the response? Add markdown details, what titles and talking points should be considered, and general guidelines after the scope;

• Example → provide a demo output, or multiple if your model accepts a larger context window.

Following these steps, you might find that a lower-intelligence model like Claude 3 Haiku is all you need. While more intelligent models like Sonnet outperform Haiku easily in multi-shot prompting or in understanding, this kind of prompt engineering limits the amount of understanding the base model needs to do, and focuses on the output.

Remember, the clearer you are in the prompt, the better your response will be.

🤝 Community Events

You can now register for all our upcoming workshops on our redesigned event home page. You can find it here.

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

I also worked on a set of new workshops we will publish very soon. We will start with a new batch from next week and go from there.

There’s a kicker for these new workshops. We took the learnings from the previous ones to deliver a better, more streamlined experience optimized to make you learn more in less time. I’m confident you will enjoy the new format.

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