- The Daily Update
- Posts
- Google uses Gemini to make robots smarter
Google uses Gemini to make robots smarter
AND OpenAI explains the path toward AGI
Jack Shields
July 12, 2024
Happy Friday!
You won’t regret opening today’s email. We have some incredibly thought-provoking headlines on deck. Let’s dive in.
In today’s Daily Update:
🗞️ Google uses Gemini AI to make its robots smarter
🤖 OpenAI explains the path toward AGI
📸 Judges crown winner of world’s first AI beauty pageant
🚨 AI Roundup: Four quick hits
Read time: 2 minutes
TOP STORY
🗞️ Google uses Gemini AI to make its robots smarter
DALL-E 3
Google DeepMind researchers are training robots with Gemini AI to improve their navigation capabilities and ability to complete tasks.
What you should know:
Gemini 1.5 enables Google’s RT-2 robots to learn about their environment by “watching” video tours of areas like office spaces.
The robots can then complete tasks based on natural prompts like “where can I charge my phone?” DeepMind says Gemini-powered robots have a 90% success rate across over 50 user instructions.
Researchers also found “preliminary evidence” for basic reasoning. When a user with several Coke cans on their desk asks for their favorite drink, RT-2 knows to look for Coke in the fridge.
It takes Gemini-powered robots between 10-30 seconds to process these instructions.
Why it matters: This is a significant milestone for AI robotics. The integration of Gemini into RT-2 is a major step toward developing robots that can understand and act on natural language instructions. DeepMind’s approach here also suggests that robots can potentially be deployed in new environments more quickly and with less manual programming.
AI INSIGHT
🤖 OpenAI explains the path toward AGI
DALL-E 3
OpenAI has identified a scale of five levels to track its progress toward AGI, a hypothetical type of AI system that outperforms humans.
The scale:
Level 1: Chatbots that can interact in conversational language. OpenAI is currently here.
Level 2: Reasoners that can complete basic problem-solving tasks as well as a human with a doctorate-level education. OpenAI believes it is on the cusp of reaching this level.
Level 3: Agents that can spend several days taking actions on a user’s behalf.
Level 4: Innovators that can aid in the advancement of different fields.
Level 5: AI that can do the work of an entire organization on its own.
Why it matters: Despite AI’s surge in popularity since late 2022, there is little consensus on what defines artificial intelligence or hypothetical AGI agents. OpenAI’s new scale provides measurable benchmarks that can help researchers track and report their progress in AI development.
INDUSTRY SPOTLIGHT: INFLUENCER MARKETING
📸 Judges crown winner of world’s first AI beauty pageant
Source: Fanvue World AI Creator Awards via CNN
Kenza Layli was just named the inaugural Miss AI in the world’s first AI beauty pageant.
The details:
Layli is an entirely AI-generated Moroccan lifestyle influencer with over 200,000 Instagram followers and 45,000 on TikTok.
Previous generations of virtual influencers required some input from human copywriters and art directors.
The AI model is dedicated to promoting diversity and inclusivity in the face of technological progress.
Layli placed first in a field of roughly 1,500 entries from AI programmers.
Why it matters: AI tools are going to drastically shake up the $14.2 billion global influencer marketing market. Synthetic faces are becoming increasingly hard to spot, which doesn’t bode well for influencers or social media users.
MORE TRENDING NEWS
🚨 AI Roundup: Four quick hits
Source: Rick Fury / Getty Images
Samsung will launch an upgraded version of its Bixby voice assistant based on its own AI models later this year.
K-pop fans debate AI experiments in the South Korean music scene.
China leads the world in generative AI adoption according to new survey.
German AI defense company raises almost $500 million in funding.
THAT’S ALL FOR THIS WEEK