The Great AI Slowdown (Or Is It)?

In partnership with

AI slowing down? Not too sure about that yet…

AI’s Math - Big Models, Small Gains

The AI industry is reportedly hitting a wall. Despite throwing massive amounts of data and computing power at large language models, companies like OpenAI are seeing diminishing returns.

The solution? Teaching AI to work smarter, not harder.

Instead of just supersizing their models (looking at you, GPT-5 or Orion, whatever you’re going to be called), they're trying to teach these digital brains to think more like humans.

For example, OpenAI is getting creative with "test-time compute." Their new model, o1, approaches problems differently. The model:

• Generates multiple solutions in real-time

• Evaluates options step-by-step, similar to human reasoning

• Uses feedback from PhDs and industry experts

• Builds on existing models base models like GPT-4

So What?

We're witnessing a pivotal moment in AI development. As traditional scaling hits its limits, companies are being forced to innovate by enhancing the inference phase (meaning the time at which the model gives back a response).

OpenAI's new approach that suggests the next breakthrough in AI might come not from more data, but from teaching AI to think more efficiently.

It's like the industry finally realized that memorizing the entire internet isn't as useful as knowing how to actually think. Who knew?

There’s a reason 400,000 professionals read this daily.

Join The AI Report, trusted by 400,000+ professionals at Google, Microsoft, and OpenAI. Get daily insights, tools, and strategies to master practical AI skills that drive results.

Sign up now for free and work smarter, not harder.

Ollama Says “Eye Do” to Local Vision

Guess what? Ollama just got its vision prescription filled with Meta's Llama 3.2 Vision model (in both 11B and 90B flavors).

This integration allows practitioners to process and interpret visual data efficiently, locally, and completely for free.

Here’s how you can do it yourself:

1. Install the latest version of Ollama.

2. Pull and run the model:

ollama pull llama3.2-vision:11b

ollama run llama3.2-vision:11b

3. Process images! Here is the prompt I used 👇️ 

What is happening in this image? [PATH TO IMAGE]

How to Use Claude’s New Visual PDF Feature

Claude can now read PDFs like a human, analyzing charts and diagrams with ease.

The chat interface now offers visual interpretation of PDFs, going beyond simple text extraction to recognize layouts and visual elements like charts and diagrams.

Imagine being able to interpret intricate documents such as anatomy diagrams or graphic-heavy reports, just by prompting Claude out of the box; I’m glad to say we’re at that point now.

To give this a spin:

1. Navigate to Claude, go you your profile icon on the bottom left.

2. Click on Feature Preview.

3. Toggle Visual PDFs to be On.

4. Drag and drop your most complex PDF into the chat window and try the prompt below! 👇️ 

I'm sharing a [TYPE OF DOCUMENT] about [TOPIC]. Please help me by:

1. First, give me a high-level summary (2-3 sentences) of what this document is about.

2. Analyze any visual elements you see:
   - Describe key charts/graphs and their main takeaways
   - Explain any diagrams or illustrations
   - Note any important visual patterns or trends

3. Point out the 3 most important insights from combining both the text and visuals.

4. If you find any [SPECIFIC ELEMENTS YOU'RE INTERESTED IN], please highlight those.

Stay updated with our favorite highlights, dive in for a full flavor of the coverage!

Sam Altman hints that AGI is just around the corner in his latest interview with Y Combinator.

Ai-Da, the short-haired, realistic robot whose self-portraits and artworks are now the stuff of modern legend, has become the first humanoid robot to sell a work at an auction for $1M.

Baidu is preparing to unveil a pair of glasses with a built-in AI assistant rivaling Meta‘s Ray-Bans that have seen success in AI-powered hardware.

Amazon is considering increasing its investment in OpenAI rival Anthropic.

The Vatican and Microsoft unveiled a digital twin of St. Peter’s Basilica that uses AI to explore one of the world’s most important monuments.

Trending Tools

We’re adding a (small) new section to each new newsletter on AI tools to try out. No fluff, just genuinely cool tools we think you should know about!

⭐️ Grok - Grok is xAI’s LLM that is now accessible to X users on the free plan.

🎙️ Sona - Sona captures your conversations and provides insights that matter most to you.

📹️ RenderLion - Transform links, words, and photos into animations.

P.S. Our Tools Database gets updated weekly with new tools. Refer just ONE friend to Digestible AI and get access to it.

You can find your custom link at the end of this newsletter!

Sure, there's talk about AI progress hitting a wall, but I'm not fully buying it.

While traditional model scaling might be showing its limits, new techniques are constantly emerging to push boundaries.

And let's be real – we're still in the early innings of actual AI adoption across most industries.

But we’re curious, what’s your take? 🧐 

If you have any suggestions or specific feedback, simply reply to this email or fill out this form. Additionally, if you found this insightful, don't hesitate to engage with us on our socials and forward this over to your friends!

You can find our past newsletter editions here!

This newsletter was brought to you by Digestible AI. Was this forwarded to you? Subscribe here to get your AI fix delivered right to your inbox. 👋