Open AI's latest model, GPT-40, is a groundbreaking multimodal AI assistant that can handle a wide range of inputs and outputs. Some key highlights:

• Multimodal Capabilities: GPT-40 can understand and generate content in various formats, including text, audio, images, and video. This allows for more natural and contextual interactions.

• Free Access for All: The advanced features of GPT-40, such as internet browsing, code interpretation, and data analytics, are now available to all free ChatGPT users. Paid ChatGPT Plus members get additional benefits like faster response times and higher output limits.

• Conversational Abilities: GPT-40 can engage in human-like conversations, with the ability to understand tone, provide emotional support, and even tell stories with expressive delivery.

• Visual Understanding: The model can interpret visual information, such as solving math problems by analyzing images, and generate images based on text descriptions.

• Desktop Integration: Open AI has released a desktop app that allows users to access GPT-40 directly on their computers, with the ability to share screen content and get contextual assistance.

Overall, GPT-40 represents a significant leap forward in AI capabilities, blending advanced language understanding with multimodal interaction. This opens up new possibilities for how humans can collaborate with and leverage AI assistants in their daily lives and work.

Open AI's new GPT-40 model is a powerful and versatile language model that goes beyond just text generation. Here are some of the key capabilities of GPT-40 that were showcased:

Multimodal Abilities

GPT-40 is a multimodal model, meaning it can handle and understand different types of media like audio, video, and images in addition to text. This allows it to perform tasks that combine multiple modalities, like describing the contents of an image or video.

Advanced Conversation

The model demonstrated impressive conversational abilities, engaging in back-and-forth dialogue and even taking on different emotional tones and personas. It was able to understand context and provide relevant and coherent responses.

Step-by-Step Problem Solving

When presented with a math problem, GPT-40 didn't just provide the final answer. Instead, it walked through the step-by-step process to solve the problem, explaining its reasoning along the way.

Customizable Voice Output

GPT-40 can generate speech output with customizable tone, emotion, and expressiveness. This allows it to sound more natural and human-like when conversing.

Image Generation

In addition to text, GPT-40 has the ability to generate images. The examples shown included creating detailed images with legible text, as well as generating consistent character designs across multiple scenes.

3D Object Synthesis

The model can take 2D images and generate 3D reconstructions, animating them and placing logos or other elements onto 3D objects.

Overall, the capabilities demonstrated by GPT-40 showcase its versatility and the rapid progress being made in large language models. The ability to fluidly combine different modalities and perform complex, multi-step tasks points to the transformative potential of this technology.

Ilia Sutskever, one of the original founders of OpenAI, has decided to step away from the company. After almost a decade at OpenAI, Sutskever has made the decision to leave the company.

Sutskever was part of the board that made the decision to fire OpenAI's CEO, Sam Altman, in November 2023. However, Sutskever later regretted this decision and publicly apologized, stating that it was a mistake to get rid of Altman.

While Sutskever's reasons for leaving are not entirely clear, it seems that he may not have been fully aligned with the direction OpenAI is headed. As a researcher and academic, Sutskever is likely more interested in the science and technology behind AI, rather than the monetization and commercialization of the technology.

In his farewell message, Sutskever expressed confidence in OpenAI's leadership under Altman, Greg Brockman, and Mira Murati, and stated that he is excited for what's to come next in his own personal project, which he will share details about in due time.

Sutskever's departure is a significant loss for OpenAI, as he was one of the company's founding members and a guiding light in the field of AI. However, the company seems to be moving forward with its ambitious plans, including the recent release of the powerful GPT-4 model.

It remains to be seen how Sutskever's departure will impact OpenAI's trajectory, but it's clear that the company is undergoing a significant transition as it continues to push the boundaries of artificial intelligence.

According to reports, several key members of the super alignment team at OpenAI have quit the company. This includes Yan Lecun, Leopold Ashenbrener, and William Saunders.

These individuals were part of the team responsible for ensuring that AI systems developed by OpenAI, like GPT-4, remain safe and beneficial. Their departure is concerning, as it suggests potential issues or disagreements within the company around the direction and safety of their advanced AI models.

The super alignment team plays a critical role in trying to mitigate the risks of powerful AI systems. Their exit could signal internal tensions or a shift in priorities at OpenAI that prioritizes rapid development over robust safety measures.

This news comes shortly after the departure of Ilya Sutskever, one of OpenAI's co-founders, who announced he was leaving the company to pursue a "personally meaningful" new project.

The loss of these key figures, especially those focused on AI safety, is a worrying development that bears close watching. It raises questions about the future direction and priorities of OpenAI as they continue to push the boundaries of large language models and other advanced AI capabilities.

The biggest announcement from Google IO 2024 was the introduction of the Gemini AI models. Gemini 1.5 Flash and Gemini 1.5 Pro are the new large language models from Google.

Gemini 1.5 Flash is a faster model, while Gemini 1.5 Pro is designed for the best possible output. Both models have a 1 million token context window, with plans to increase it to 2 million tokens in the future. This allows for input and output of around 1.5 million words.

Another highlight was Project Astra, which allows a mobile phone to see what the camera is looking at and answer questions about it. The demo showed the phone remembering details like the location of a pair of glasses, and the presenter was able to continue interacting with the system using a pair of augmented reality glasses, hinting at future Google Glass-like capabilities.

Google also showcased their new text-to-image model, Imagine 3, which is approaching the realism of models like Midjourney. They demonstrated a new video generation model called Veo, which can create 1080p videos over 1 minute long, though it doesn't quite match the quality of Anthropic's Sorai.

Other announcements included upgrades to Google Search, Gmail, and other Google Suite tools, adding AI-powered features like multi-step reasoning, automatic email organization, and photo context understanding.

Overall, Google's IO event was packed with a wide range of AI-powered announcements, showcasing the company's continued push to integrate AI across its products and services.

Starting with Anthropic, the company has hired Instagram's co-founder Mike Krieger as its new Head of Product. Krieger, who was one of the co-founders of Instagram as well as the co-founder of the news app Artifact, will be tasked with engineering good user experiences to get more people excited about and using Anthropic's tools like Claude.

Anthropic has also released a new prompt generator feature in their console. Users can now generate production-ready prompts by describing what they want to achieve, and the system will use prompt engineering techniques like Chain of Thought reasoning to create more effective, precise, and reliable prompts.

Moving on, the AI company Hume has released a new tool called Chatter - an interactive podcast experience. Chatter is a podcast that allows you to steer the conversation, asking the AI host questions and getting responses tailored to your interests, in this case focusing on the latest AI news.

Finally, a clip from Bumble founder Whitney Wolfe Herd went viral last week, where she speculated about the future of dating involving AI dating concierges. The idea is that your personal AI concierge would go on dates with other people's AI concierges to determine compatibility, before introducing the real people. While this sounds like a plot from Black Mirror, it highlights how AI could potentially play a role in future dating experiences.

Overall, the AI world continues to evolve rapidly, with companies like Anthropic, Hume, and even dating apps exploring new ways to leverage this technology. It will be interesting to see how these developments unfold in the coming months.

The past week has been a whirlwind of AI news, with major announcements from both Google and OpenAI.

OpenAI unveiled their latest language model, GPT-4, which is a powerful multimodal system capable of handling a variety of inputs like audio, images, and video. The most impressive aspect is that GPT-4 will now be available to all free ChatGPT users, giving them access to advanced features previously reserved for paid subscribers.

Google, on the other hand, took a different approach at their I/O event - bombarding the audience with over 100 AI-related announcements. Highlights include the new Gemini language models, the impressive Project Astra that can visually understand a scene, and advancements in text-to-image and video generation.

While Google may have overwhelmed with the sheer volume of updates, both companies demonstrated significant progress in making AI more accessible and capable. The race for AI supremacy is heating up, and it will be exciting to see how these technologies evolve and impact our daily lives in the coming years.

As the AI event season continues, with upcoming showcases from Microsoft, Cisco, Qualcomm, and Apple, there will be no shortage of innovative AI developments to look forward to. This is a thrilling time for the AI industry, and I'm eager to continue sharing the latest news and insights with you.