Big publishers are going all-in on AI

OpenAI Cuts Deals with Vox and The Atlantic

Breaking news in the media world: Vox Media and The Atlantic inked licensing deals with OpenAI. Let’s dive into what this means. 

Some context: OpenAI tried to gobble up the entire internet to speed up model training—only to backfire with copyright infringement claims from big publishers like the New York Times. Now? The AI company has been mending rocky relationships with media companies through a different strategy: formal partnerships. 

The new deals: This year, OpenAI secured partnerships with Dotdash, Axios, The Financial Times, and more—gaining full rights to training data to clear the way for AI development. This week, Vox Media and The Atlantic joined the fold with deals of their own. 

• Price tags for these agreements weren't disclosed, but it's safe to assume both publishers were well-compensated for their content—anywhere from $1 million to $5 million a year. 

Both multi-year partnerships allow OpenAI to: 

• License archived and current content to train its AI models.

• Use content to fine-tune accurate responses in OpenAI products, including ChatGPT.

• Include citations and links to the original articles when their content is referenced.

The Atlantic and Vox Media are cashing in their side of the bargain too:

• Atlantic Labs: The Atlantic’s product team will use OpenAI's tech to develop new AI-driven products and features.

• Vox Media’s Forte: Vox will integrate OpenAI’s technology into its first-party data platform, Forte, to boost advertising creative optimization and audience targeting capabilities. 

Why is this happening? These deals provide quicker (and often bigger) financial returns than waiting for lawsuit settlements and immediate access to advanced AI tools. Still, some media outlets are holding out to keep control over their content and protect their site traffic/revenue.

Mistral Launches A New AI Model for Code Generation

Programming just got a lot more accessible. On Wednesday, French AI startup Mistral launched Codestral, its first-ever code-centric AI model. 

The details: Codestral is a 22B parameter, open-weight AI model designed for code generation tasks. It’s trained on a diverse dataset covering over 80 programming languages—including Python, Java, C++, JavaScript, and more. 

• Mistral’s model excels at completing coding functions, writing tests, and filling in partial code snippets.

How it stacks up: Mistral claims that Codestral already outperforms existing code-centric models such as CodeLlama 70B, DeepSeek Coder 33B, and Llama 3 70B. One of its standout features? A context window of 32k tokens that handles long-range code generation tasks more effectively than its competitors.

How to use Codestral: 

• On HuggingFace: Codestral is available for download under the Mistral AI Non-Production License for research and testing purposes.

• Via API: Developers can use one of the dedicated endpoints at codestral.mistral.ai (free during an 8-week beta period) or api.mistral.ai (queries billed per token).

• Le Chat: An instructed version of Codestral is accessible through Le Chat (Mistral’s chatbot), allowing developers to interact conversationally with the model.

Empower Your Business with AI: Your Essential Guide Awaits

Unlock the potential of AI to propel your business forward with HubSpot’s highly anticipated AI For Business Builders guide! Tap into it’s power to:

• Automate tasks

• Elevate decision-making processes

• Give you a competitive edge in your industry

Step into the league of accomplished entrepreneurs who are harnessing AI to drive growth and foster innovation in their ventures. Don't miss out on this opportunity to unlock your business's full potential.

Download HubSpot’s free guide and unlock the capabilities of AI.

What’s Up With Google’s AI Overviews?

Google's AI is in the news for all the wrong reasons...again. Last week, Google launched AI Overviews—and the execution has been anything but smooth.

Some context: AI Overviews, previously known as the Search Generative Experience (SGE), aim to enhance search results by providing AI-generated summaries at the top of Google searches.

So what happened? Google’s AI Overviews have been delivering some…questionable (and downright inaccurate) answers. For instance, one query told the user to eat “at least one small rock a day.” Another query response suggested adding glue to get cheese to stick to pizza. 🤔

• This isn’t the first time Google has faced criticism for AI inaccuracies: Past issues with its AI-generated content (such as its Gemini AI tool depicting Nazi-era German soldiers as people of color) have already raised concerns about the reliability of its search results.

• Iyo thinks its GenAI earbuds can succeed where Humane and Rabbit stumbled.

• Chatbot maker Character.ai is discussing potential partnerships with Meta, xAI.

• With hallucinations waning, AI is diving deeper into scientific research.

• Microsoft’s Copilot is now on Telegram.

• Exactly.ai secures $4M to help artists use AI to scale up their output.

• Palantir lands $480M Army contract for Maven artificial intelligence tech.