Today’s newsletter was brought to you by Mercury.

Deeper Learning

Written by Sarah Wright with Aaron O’Leary

Hey folks! Today is a doozy. It’s a big week for the battle of OpenAI vs. Google. Then I couldn’t even finish this newsletter without news breaking from OpenAI.

I’m also covering a new product that combines the best features of Notion with Miro — but that’s not even the best part about it.

Ok, let’s go deeper.

SPONSORED BY

To keep up with the world of AI, you need the right talent, a minimum lovable product, and a sure path to profitability. More than that, you need the right financial tools.

Mercury offers banking** and credit cards to help ambitious companies simplify their finances. With the IO credit card, you can start spending right away with an industry-low qualifying balance minimum, boost your cash flow with 1.5% cashback, and unlock higher credit limits thanks to our startup-friendly underwriting model.

Plus, get $500 when you sign up for Mercury and deposit $10K in your first 90 days.

TALKING POINTS

🔊 Earphones with cameras: Meta is working on AI-assisted earphones with cameras, but apparently Mark Zuckerberg has yet to see a design from the team that he likes. The thought is that such earphones can be used to identify objects and translate foreign languages.

🛩️ Travel easy: I’ve got to give it to Expedia. Romie is a cute name for a bot, especially one that can help you plan travel. Expedia is using a mix of in-house and OpenAI models to test out AI tools that can help do things like search for hotels and build itineraries.

PRODUCT HIGHLIGHT

OpenAI came out swinging. This time Google was ready.

Open AI held its Spring Update event yesterday, and Google held its annual developer conference, Google I/O, today. Over the past few years, Google has watched OpenAI take the lead and become the darling child of AI. That didn’t really change this week, but Google’s updates were big enough to be worthy of the spotlight.

OpenAI’s announcements

A new model. Despite rumors that OpenAI would launch a search product or GPT-5 at the Spring Update, the big news was actually a new flagship LLM called GPT-4o. GPT-4o is similar to GPT-4 in intelligence but is much faster and improves capabilities across text, vision, and audio, explained CTO Mira Murati.

“Her” is here. It was the voice handling that really stood out during the event. ChatGPT has long had a voice option, but GPT-4o massively improves it. For example, users can ask the model a question and mid-answer, interrupt the model to divert the conversation or correct it. During the demo, the team stressed how much more human this new model is by showcasing its emotional reasoning capabilities and even doubling down on the whole “Her” movie comparisons by demoing an assistant that observes the world around you and responds in real-time. It even sounded similar to Scarlett Johanson’s character in the movie.

The roll-out: GPT-4o is rolling out to all paid and free users of ChatGPT now, plus the company’s API for developers to tinker with. Alongside the model launch, OpenAI also announced a new desktop app that’s rolling out on Mac first — kind of ironic since Microsoft has been bankrolling.

Google Search is getting better, finally

Google announced a slew of updates across its models and products at Google I/O today.

New models. Veo is Google’s answer to OpenAI’s Sora – it’s a new generative AI model for video that outputs 1080p video more than a minute long. Google’s competitor to DALL•E also got an update: Imagen 3. Google says Imagen 3 better understands the intent behind your prompts which helps it to master a range of styles.

Gemini also got an upgrade. Gemini 1.5 is faster and it's context window (we explained this last week here) is nearly double its predecessor. The model interface will go into the sidebar for Docs, Sheets, Slides, Drive, and Gmail.

THE big update to Search. Gemini 1.5 will power the new Google Search, and we’ve been waiting a long time for this. Google had already been testing AI Overviews while other AI browsers sprouted up. Now, Google will start rolling its updates for English language queries in the U.S. Here’s what you’ll see:

• AI Overviews offer AI-generated answers to complex questions, meaning ones that require a mix of inspiration and information. For example, if you type, “Find the best yoga classes in Boston with intro offers that are walking distance to me,” Google will pull you those options and put them on a map.

• Some answers will even come with organization options to further have AI adjust or filter the response, for example, separating restaurant results into categories.

• Google will even help you “plan ahead.” For example, ask Google to create a 3-day, easy-prep meal plan, and it will give you recipes that you can export to Google Docs or Gmail.

• Woah. Maybe the coolest feature announced is being able to use video and voice to make complex queries, like this:

• Womp, Womp. Unfortunately, Google is in hot water because the demo of that very feature produced incorrect answers. And this isn’t the first time Google included a demo of its AI with a mistake.

What does it mean? “This is one of Google’s most obvious advantages over upstart search engines, which don’t have anything close to the troves of reviews, mapping data, or other knowledge that Google has, and may not be able to tap into APIs for real-time or local information so easily,” writes Lauren Goode for Wired.

Woah, again. An AI-Assistant. Everyone in AI is trying to crack the nut of creating a real, Jarvis-like AI assistant and Google is no different. Google revealed Project Astra, a multi-modal AI assistant. In the demo, a woman points her camera across the room asking for all kinds of help — solving equations, identifying objects, and finding her missing glasses. The assistant answers all of it. It’s an early model, but Google swears it’s not a doctored video.

So who wins this round, Google or OpenAI? This one is a toss-up for me, but if pressed, I’d have to say OpenAI because of Google’s fumble with its video Search feature. If we see Project Astra come to life before OpenAI drops a Jarvis assistant, then we’ll talk. What do you think?

Q&A

What questions do you have about AI?

Ask Kitty is where we answer your questions about AI. Kitty will be back next week. In the meantime, below are some of the submitted questions that we’ve received. Let us know what you want to learn about next with your vote.

MORE TOOLS

For Work

• Dover launched a free Rate My JD for hiring managers.

• VoiceCheap lets you dub and translate videos in any language.

• Alphabot works real customer support people into your support flow when the AI bot can’t answer.

For Life

• SofaBrain is an AI tool for visualizing the interior of your new home.

• Stoic is an all-in-one journaling app with AI-powered reflections. 

• Peridot, the adorable AR pet from Niantic, has a new experience called Hello, Dot on Meta Quest 3.

For Developers

• Wejic is a web developer assistant that uses GPT-4o to generate sites.

• Copilot Chat in GitHub's mobile app is now generally available.

Thanks for going deeper with us!

Here via forward? Subscribe here.

Have feedback?