Justin's AI Newsletter

The most interesting content, news, and insights in AI each week.

--

Hey friends,

This week was another insane one in the world of AI. It's one of those weeks where it feels like years go by because of how much changes.

Let's discuss.

AI LAUNCHES

🤯 OpenAI's text-to-video model

The future is officially here.

OpenAI launched their text-to-video model Sora this week and the internet is going wild. That's because unlike previous models out there, Sora looks REAL.

Garrett Scott 🕳 @thegarrettscott How it started (1 year ago) vs how it's going: 1:34 PM • Feb 15, 2024 2027 Retweets 13829 Likes Read 245 replies

​

Sam Altman did a whole thread of people sending ideas and him replying with videos—it's worth checking out some.

Sam Altman @sama https://twitter.com/fxturevescent/status/1758203275940155655 FXTUREVESCENT e/acc @fxturevescent @sama Two golden retrievers podcasting on top of a mountain 2:56 PM • Feb 15, 2024 3768 Retweets 43382 Likes Read 727 replies

​

Sam Altman @sama https://twitter.com/catebligh/status/1758198108985389146 cate bligh @catebligh @sama A instructional cooking session for homemade gnocchi hosted by a grandmother social media influencer set in a rustic Tuscan country kitchen with cinematic lighting 2:59 PM • Feb 15, 2024 2075 Retweets 20031 Likes Read 1059 replies

​

OpenAI's new model Sora can:

• Generate videos up to 1 minute long
• Take shorter videos and make them longer
• Take images and turn them into videos

OpenAI hasn't released Sora yet. They're collaborating with red teamers to test the product and ensure it's ready for public use. The risk of misinformation and fabricated videos is understandably high, so it'll be interesting to see how they launch it.

The implications here are massive. What will this mean for creators? Animators? Hollywood? Will YouTube be flooded with AI videos? We will see.

(My TikToks will still be me ... for now 😉)

AI LAUNCHES

🤝 Google's 1M Context Window

On Thursday, Google announced Gemini 1.5 Pro only 1 week after rolling out their Gemini 1.0 Ultra model.

According to Google CEO Sundar Pichai, 1.5 Pro achieves comparable quality to 1.0 Ultra while using less compute—and we all know how valuable compute is these days (why Sam Altman wants to raise $7T).

The context window for standard users is 128,000 tokens, but a limited group of developers and enterprise customers can use a model that can operate with 1 million tokens—the longest context window for any LLM.

What does that mean?

As a reminder, a context window refers to the size of the content you input in your prompt. At 1 million tokens, you can include 1 hour of video, 11 hours of audio, 30,000 lines of code, or over 700,000 words in a prompt. In research, the Google team has even experimented with up to 10 million tokens, which is unprecedented.

Check out the demos of what's possible with the 1.5 Pro model. 👇

Rowan Cheung @rowancheung Gemini 1.5 Pro's 1,000,000+ token context length is incredible. I got early access and spent my Saturday night running tests. Here are 6 impressive capabilities I found: 1:15 PM • Feb 18, 2024 590 Retweets 4661 Likes Read 138 replies

​

This is huge news for Google and the AI world. Very excited to see what people use it for.

AI TOOLS

✈️ The Chatbot Lawsuit

What a story.

Air Canada just lost a lawsuit because their AI chatbot made up a refund policy out of thin air! Check out my video for a full explanation: [Video]​

Until AI chatbots become more reliable and less susceptible to hallucinations, I don't think it's practical for most businesses to use them.

Rather than fully replacing humans with AI and implementing a chatbot on your site, you should develop internal workflows to help automate manual tasks for your customer service team (without introducing the risk of AI hallucinations).

This is precisely why we've been building Cassidy. We believe internal automations hold the most value for businesses right now. In fact, we don't even allow businesses to create external facing chatbots with our product!

Check out how customer service teams have been using Cassidy 👇

​

Want to access the power of Cassidy. You can sign up here.

⭐ AI Quick Hits

1. OpenAI is disrupting malicious use of AI by state-affiliated threat actors. Nope not Hollywood, this is real. [Read here]
2. Apple is working on Copilot for Xcode. Big news for iOS devs.[Read here]
3. Daniel Gross and Nat Friedman invest $100M in Magic.dev, which is building a superhuman AI software engineer. [Read here]
4. Google, Amazon, Adobe, IBM, OpenAI, ElevenLabs, and more signed a pledge to fight against election deepfakes [Read here]
5. The US Patent Office has denied OpenAI's trademark for GPT—which when you think about it, is pretty funny [Read here]

Stay tuned for more next week...

Thanks for reading,

Justin

​

P.S. Want more quick, hot AI takes?

Be sure to follow me on Twitter & LinkedIn! 🔥🤖