OpenAI unveils GPT-4o (Her), Facebook's Camerabuds, Google Claps Back at OpenAI, and More!

As the tech giants unveil their product roadmaps, the race to dominate the AI landscape is more thrilling than ever 🚀. With the top spot still up for grabs, companies are pulling out all the stops to claim leadership in this dynamic space. Apple has yet to officially enter the race, adding an air of mystery to their AI ambitions 🤔, but we won't have to wait too long to learn what they are up to as WWDC is just a few weeks away.

Just ahead of Google I/O, OpenAI revealed its most advanced model yet, GPT-4o, but Google quickly responded with a demo of what's to come. The competition is fierce, and the innovations are game-changing 💥

Bonus points if you can spot the subtle jab OpenAI took at Google right at the top of the GPT-4o announcement page without scrolling any further! [link]

There is lot to cover so let’s dive right in!

ThisWeek in AI - OpenAI GPT-4o (HER)

GPT-4o: Let me start by saying, this is the most impressive model I’ve interacted with to date! It’s not just the responses that are remarkable, but the way it responds, offering the most human-like interaction you’ll experience with an LLM. Is it exactly like talking to a human? Of course not, but it’s closer than any other LLM out there.

Watch this incredibly impressive demo showcasing how GPT-4o effortlessly facilitates a fun and engaging game of rock-paper-scissors, complete with humor and ease.

Ok one more! The humanness of these interactions is seriously impressive

With GPT-4o, ‘o’ for ‘Omni’, as the model is capable of accepting text, audio and visual inputs and also producing the same in outputs unlike it predecessor which leveraged multiple models to handoff information to one another until the desired output was achieved. This update gives GPT a more human personality and reduces the response time for audio inputs to just 320 milliseconds (down from 5.4 seconds in GPT-4), similar to how humans interact with each other.

I watched almost all the demos released yesterday and here are the use cases this new update unlocks.

• Travel and Accessibility guide: Whether it is recognizing landmarks, translating in real-time or reading signs and menus. GPT-4o can do it all!

• All-knowing Tutor/Helper: Whether you are trying to learn math, cooking a new dish or putting IKEA furniture together. GPT-4o can help you with all of it and more!

• AI working with AI: AI’s can now work together to make progress on a specific task. AIs negotiating with each other, sharing information, or as Bumble founder put it, your AI concierge and go on a date with another AI concierge to check for compatibility before you match.

• True Work Partner: What if you could just share your screen with GPT-4o while working on a presentation, a document or solving a coding problem and let GPT-4o help you out in real-time giving you suggestions, unblocking you, etc.. or sending GPT-4o to attend a meeting on your behalf and report back with a meeting summary.

• Personal Assistant: GPT-4o isn't your ultimate personal assistant yet, but it's on the path with its latest update. It can be your brainstorming partner, style guide, business advisor, and more. While it can't perform all tasks like a human assistant, such as finding the best restaurant and making reservations or building business plans from random ideas but we're getting closer to that reality. The future where GPT-4o handles these tasks is exciting, and it's only a matter of time before it reaches that level of functionality.

More highlights from the announcement:

Launch of ChatGPT Desktop App and Updated UI:

• OpenAI introduced a new desktop version of ChatGPT, complete with a refreshed user interface that promises more intuitive and natural interactions.

Advanced Features and Greater Accessibility:

• The new voice mode operates seamlessly within GPT-4o, minimizing latency and making conversations flow more naturally.

• Enhanced vision capabilities allow users to upload and interact with images, screenshots, and documents containing text.

• Memory features provide continuous context across sessions, making interactions more personalized and helpful.

• Real-time browsing and advanced data analysis tools enable users to access current information and analyze complex data sets with ease.

Real-Time Conversational Speech and Emotion Detection:

• Live demonstrations highlighted real-time speech capabilities, where users can interrupt and interact without any delay.

• The model can detect and respond to emotions, enhancing the empathy and natural feel of conversations.

Expanded Language Support and API Availability:

• GPT-4 now offers improved performance in 50 different languages, making it accessible to a broader global audience.

• Developers can now access GPT-4o via API, allowing them to create cutting-edge AI applications with double the speed, half the cost, and five times the rate limits compared to GPT-4 Turbo.

Watch this mind-blowing demo of GPT-4o playing a tutor

Are you excited, yet?! Well, I am! Not only to get my hands on this latest update but also to see how developers innovate with this new model! It’s quite impressive! Watch the full announcement video below and more impressive demos here.

Oh, and before wrapping up this section, I have to share something amusing. Remember when Google released a video last year demonstrating its vision capabilities, only to later reveal that it was edited to cut out the latency? Well, OpenAI cheekily called them out on their GPT-4 homepage 😆

Google didn’t take these announcements lying down. They showcased their own technology in a demo, offering a glimpse of what’s to come at Google I/O. Check out the demo below.

Excited about an AI startup or a product? Let us know at [email protected]

ThisWeek in AI - Art

Randy Travis team released a song that used AI-cloned voice debuted at #45 on Country Airplay chart. Listen to it on the Spotify link below!

ThisWeek in AI - Interesting Reads