🤖 GPT-4 Vision's newest competitor is FREE

Welcome, AI enthusiasts.

An open-source AI model emerged this week with visual abilities potentially rivaling GPT-4 (and it’s completely free to try).

Can open-source AI compete with the LLM giants? Let’s dive in…

In today’s AI rundown:

• 👀 Open-source LLaVA challenges GPT-4

• 🧠 Anthropic breakthrough makes AI understandable

• 🛠️ 8 New AI tools

• 💻 Create websites from sketches with ChatGPT

• 🤖 3 Quick AI updates

Read time: 3 minutes

LATEST DEVELOPMENTS

OPEN SOURCE AI

👀 Open-source LLaVA challenges GPT-4

The Rundown: Researchers from Stanford, UW-Madison and Columbia introduced LLaVA, a new open-source AI system that could rival GPT-4 for visual and language understanding. It’s open-sourced and completely free to use.

The details:

• LLaVA connects CLIP (an AI system that understands images) with LLaMA, an open-source text AI model.

• In a demo, LLaVA showed it could understand and have convos about images, much like the proprietary GPT-4 system, despite having far less training data.

• Unlike the private GPT-4, LLaVA's code, trained model weights, and generated training data are freely available online.

Why it matters: LLaVA proves the potential of open to push vision-language AI forward. While still likely no match for GPT-4's sheer scale, its accessibility, and strong initial results are impressive.

TOGETHER WITH PLUTO

🤖 How to use AI to invest smarter

The Rundown: With Pluto.fi, you can have your own AI analyst that can provide reports on your portfolio and specific stocks, so you can make smarter investing decisions.

Here’s what over 20,000 investors use Pluto for:

📊 Access personalized real-time market data and analysis

🧠 Team up with your AI to turn market trends into lucrative Automations

💰 Invest in any US stock, fine-tuning with AI support

Start your free trial today at Pluto.fi

ANTHROPIC

🧠 Anthropic breakthrough makes AI understandable

The Rundown: Anthropic just developed a new method to interpret the individual neurons inside LLMs like Claude, helping researchers better understand and decode the model’s reasoning.

The details:

• Neurons usually respond to many unrelated concepts, making them hard to understand in isolation.

• Anthropic found a way to decompose groups of neurons into simpler "features" with clearer meanings.

• For example, they extracted DNA, legal text, and HTML features from a blend of 500 neurons — and stimulating a single feature consistently caused expected model behavior.

• Features also link together in systems for complex functions like producing valid HTML.

Why it matters: This massive breakthrough makes it easier to understand what’s happening when the AI is thinking. Not only is it crucial to understand this from an AI safety perspective, but it will allow much more control over the entire system as a whole.

QUESTION OF THE DAY

TRENDING TOOLS

🤖 VantagePoint A.I. - dual-patented forecasting tool predicts moves on Tesla, Nvidia, Apple, and more(!) with up to 87.4% proven accuracy. Attend FREE Live A.I. Market Training >*

💬 TalkNotes- Turn your thoughts into actionable notes, fast (link)

🗣️ Chatcare- Reduce your customer support volume by 60% (link)

🧾 Airparser- Data extraction powered by GPT-4 (link)

🪄 Framer AI- Generate a website in minutes from prompts (link)

⚡️ Notion AI- Your notes, on steroids. Access the limitless power of AI, right inside Notion (link)

📱 Synthesia- Create professional videos without mics, cameras, or actors using an AI-generated character (link)

📝 Writesonic- ChatGPT alternative that addresses the limitations of ChatGPT (link)

Browse the most popular tools ever featured with our tool database.

* = This is sponsored content

AI TRAINING

💻 Create websites from sketches with ChatGPT 

If you have an idea for a website or app layout, but don’t have any coding experience, you can now use ChatGPT’s new vision capabilities turn your simple sketches into functioning code!

Step 1: Sketch your desired webpage layout. Don’t worry about expert design — the tool is surprisingly competent at interpreting the drawings to generate a rough layout.

Step 2: Upload the image. Use the following prompt: "You are an expert developer. Write HTML and CSS code for the webpage layouts I've drawn."

Step 3: Iterate and add details. ChatGPT will provide the code for your design — test the output in your tool of choice. Then, conversationally talk through any changes or additions you’d like!

It’s never been easier to build and execute with the power of AI — shoutout to Brian Marley on X for the demo.

QUICK HITS

⭐️ Research: LLMs outperforming search engines

Is AI about to replace traditional search engines? A new paper from Google found that using GPT-4 and Fresh Prompt (a method of prompting LLMs with web search results) resulted in more factual responses than both traditional search and standalone LLMs.

✅ Meta shares method to watermark AI images

Meta and Inria just released a new technique to create invisible watermarks on images from open-source AI models. Their Stable Signature method traces generated images back to the original model — and potentially offers a solution to one of the more heavily critiqued elements of AI creations.

🗳️ Alexa spreads election fraud claim

Amazon's Alexa was reportedly spreading misinfo that the 2020 election was "stolen by massive fraud” — citing an unverified source for the false claim of a rigged election. Amazon called it a bug, saying it uses credible sources for election info.

SPONSOR US

🦾 Get your product in front of over 350k+ AI enthusiasts

Our newsletter is read by thousands of tech professionals, investors, engineers, managers, and business owners around the world. Get in touch today.

THAT’S A WRAP

Keep an eye out and click the star emoji hidden somewhere in today’s email for a free entry into our Ray-Ban Meta AI smartglasses giveaway.

Additionally, you’ll get another free entry for every subscriber who joins our newsletter using your unique link: https://www.therundown.ai/subscribe?ref=PLACEHOLDER

You can track how many referrals you’ve made here. Happy hunting!

If you have anything interesting to share, please reach out to us by sending us a DM on Twitter: @rowancheung & @therundownai

❤️ Rundown review of the day