GPT-5 News, GPT-4 IMPROVES 900%, AI Crashes The Stockmarket ,Microsoft Announcement,[AINEWS#6]

AI Landscape Overview

This section provides an overview of the different things that occurred in the AI landscape.

Deep Fakes

• Deep fakes are becoming more perfected with stable diffusion controller and epsonth and fusion.

• Facial reconstruction and deep fakes are scary as they can be used to create realistic fake videos.

Economic Impact of Fake News

• An AI-generated image of an explosion at the US Pentagon surfaced, causing a 500 billion dollar market cap swing on a fake image.

• This incident shows how AI tools can impact the economy, and it raises concerns about what could happen when these tools get even more sophisticated.

Scams Using Artificial Intelligence

• A clip from 60 Minutes showed how easy it was to scam someone using artificial intelligence in real-time.

• The clip highlights concerns around privacy and how this technology is going to affect many verification tools that we currently use.

Wearables That Let AI See the World

• A user on Twitter created a wearable that lets AI see the world using opening eyes whisper, an image-to-text model, and 11 Labs human-like voice model.

• Humane is building artificial intelligence-powered wearables that try to assist us in our daily lives and replace phones.

Sketching Environments and AI Clones

In this section, the speaker discusses the ability to sketch out environments and control their layout using AI. They also talk about an influencer who launched an AI girlfriend voicebot that made $72,000 in revenue in just one week.

Sketching Environments with AI

• The ability to sketch out environments and control their layout with initial sketches is creative and interesting.

• This kind of technology will power the next generation of creative individuals.

• An influencer launched an AI girlfriend voicebot that made $72,000 in revenue in just one week by cloning her voice with 11 labs.

Any-to-Any Generation via Composable Diffusion

In this section, the speaker talks about any-to-any combinations from video image audio and text via composable diffusion. They discuss how large language models are evolving into multimodal capabilities involving image, video, text, and audio.

Any-to-Any Generation via Composable Diffusion

• Researchers can generate any-to-any combinations from video image audio and text via composable diffusion.

• Large language models are evolving into multimodal capabilities involving image, video, text, and audio.

Scientific Progress in Image and Text Reconstruction

In this section, the speaker discusses scientific progress in image and text reconstruction from fmri research. They mention Adobe's fascinating AI tool for manipulating images.

Image and Text Reconstruction from fmri Research

• There has been incredible scientific progress in image and text reconstruction from fmri research.

• Researchers are refining techniques using AI to get higher quality images.

Adobe's Fascinating AI Tool

• Adobe has released a fascinating AI tool for manipulating images.

• The tool allows users to expand and manipulate images with generative fill using AI.

Microsoft Announces Three Insane Things

In this section, the speaker talks about Microsoft's announcement of three new things, including an update for Bing with chair TPT and many other integrations across Bing.

Microsoft's Announcement

• Microsoft announced three insane things, including an update for Bing with chair TPT and many other integrations across Bing.

Instacart Shopping and Windows Copilot

The section discusses the productivity benefits of using Instacart shopping and introduces Windows Copilot, a tool that allows users to ask their computer questions and execute commands.

Instacart Shopping

• Users can take ingredients from a recipe page and add them to an Instacart shopping cart with one click.

• This feature provides incredible productivity benefits for people who want to save time on grocery shopping.

Windows Copilot

• Microsoft has introduced Windows Copilot, which allows users to ask their computer questions and execute commands.

• This tool is similar to Bing for Windows, where users can ask their computer anything they want.

• It can actively run and execute different programs, commands, and help navigate the computer if the user is confused.

• It has many applications because many people don't know where everything is located on their computer.

Future Modalities: Video vs Text

The section discusses future modalities in AI training models.

Video as a Modality

• Altman mentions that video is going to be his focus for future modalities in AI training models.

• He believes that text is good but cannot achieve true AGI with text alone.

• Using video will be faster compared to using text alone.

Language Model vs Video

• There's a debate about whether language models can get all the way to AGI or if video representation is necessary.

• Altman thinks it's possible but the fastest way would be to have other representations like video in these models as well.

Adobe's Text-to-3D Image Composer

The section discusses Adobe's new tool that converts text into 3D models.

Text-to-3D Image Composer

• Adobe is working on a text-to-3D image composer that converts text into 3D models.

• It seems to be really effective and could take market share from Photoshop.

• The tool allows users to manipulate images as if they're 3D models, which is incredibly efficient.

• However, it requires a wide array of images to be trained on in order for the manipulation to be effective.

Figure Raises $70 Million for Humanoid Robots

The section discusses Figure, a company that raised $70 million for humanoid robots.

Figure's Humanoid Robots

• Figure raised $70 million for their humanoid robots.

• They are advancing rapidly and one of the first humanoid robots backed by OpenAI is expected to debut in summer 2023.

Working on Robots and Improving AI Reasoning

In this section, the speaker discusses the design of robots and their potential advancements in the future. They also mention a new paper that improved GPT4's reasoning by 900.

Robot Design

• The robots being worked on have a sleek, slanted cylinder-shaped face.

• It will be interesting to see where these robots are in a year from now and what kind of AI systems power them.

Improving AI Reasoning

• A new paper was released discussing the tree of thought and how they were able to improve GPT4's reasoning by 900.

• The speaker provides a link to a video that covers this topic in depth.