- Modern Chaos
- Posts
- MC.32: AGI within 3-24 months?
MC.32: AGI within 3-24 months?
Unveiling the near-future of AGI with Dario Amodei, Google's breakthrough in long-text memory, and Meta's quest for spatially aware AI.
Samuel ROY
April 18, 2024
In partnership with
Hi everyone,
With the viral arrival of Suno and Udio in the generative music space, I was wondering how music will evolve.
These services are just starting, so the music produced is the worst it will ever be, yet it already does an amazing job. What happened with images and Midjourney will happen in music.
Now, what does it mean for artists? If everyone is able to create music at a professional level, how can you make a place for yourself?
My take: you'll need to get out there and produce experiences for people to live.
You can't do music that will just be listened to. AI can do it and it can do it on a scale unimaginable. It can take the best popular songs produced in the past 10 years and create original tracks from them. But AI won't be able to play that music or explain why it created it. There’s no emotion infused; it’s not grounded with personal history.
If you are an artist, go out there; we need you to break out of our smartphones, live great moments with friends, and be energized.
Cheers,
Sam
PS: Apple folks, we shipped a massive feature on Alter this week; now you can invoke AI actions directly on your files from your Finder. Check our changelog!
In this issue:
🗞️ The New York Times: Ezra Klein Interviews Dario Amodei
🔗 Google: Efficient Infinite Context Transformers
🌪️ Mistral: Mixtral 8×22B instruct version
🏠 AI at Meta: OpenEQA
👁️ X: Grok-1.5 Vision Preview
🍏 Apple: Readies M4 Chip Mac Line, Including New MacBook Air and Mac Pro
🌐 Microsoft: $2.9 billion in AI and cloud infrastructure in Japan
🚀 AI 50: Companies of the Future
🖼️ A.I.Warper on X: Viggle AI Experiment
🏛️ Sweetspot: AI for Government Contracting
🤖 Magna: Sanctuary AI’s humanoid robot test drive
🧑💻 V7Labs: It's here.. Work automation for the multimodal AI era.
Updates & tools
The New York Times: Ezra Klein Interviews Dario Amodei
Dario Amodei, co-founder and CEO of Anthropic, discusses the imminent arrival of AI systems resembling AGI within 3-24 months, primarily through scaling up current technologies. Additionally, the financial investment in training these models is skyrocketing, with future models estimated to cost between $5 billion and $10 billion. MORE
Google: Efficient Infinite Context Transformers
Google's new paper introduces a transformer that can handle super long texts without forgetting the earlier parts, which is a game-changer for understanding large documents or books. Imagine an AI that can remember and use everything it's read—bye bye RAG systems? MORE
Mistral: Mixtral 8×22B instruct version
Mistral AI continues to impress with its contributions to the open-source community. Their latest offering, Mixtral 8x22B, promises enhanced capabilities over the already impressive Mixtral 8x7B, and I am eager to see how the new instruct version performs in practical applications. MORE
AI at Meta: OpenEQA
Meta's new OpenEQA benchmark is here to push the boundaries of how AI understands and interacts with physical spaces, asking questions like "Where did I leave my badge?" It's a big step toward making AI more aware of its surroundings, which today's models are still pretty bad at. MORE
X: Grok-1.5 Vision Preview
Grok-1.5V is setting new benchmarks in understanding our physical world. They've also introduced a new dataset called RealWorldQA that really tests an AI's grasp of spatial realities. MORE
Apple: Readies M4 Chip Mac Line, Including New MacBook Air and Mac Pro
Apple is gearing up to introduce a new line of Macs with M4 chips focused on enhancing AI capabilities. This move could redefine personal computing with more powerful and efficient machines. Personal AI soon? MORE
Microsoft: $2.9 billion in AI and cloud infrastructure in Japan
Microsoft plans a massive $2.9 billion investment in Japan to enhance AI and cloud services, aiming to boost the country's tech skills and cybersecurity. Is there a country where Microsoft hasn’t poured dozens of millions for AI? MORE
AI 50: Companies of the Future
Sequoia Capital's AI 50 list shows us which companies are pushing the boundaries of AI to improve productivity. It’s exciting to see how these companies evolve and what new tools they bring to our everyday lives. MORE
A.I.Warper on X: Viggle AI Experiment
This tool might be the next big thing in digital content creation. It takes a person's photo and applies style transfer to another video, which could unleash a wave of new meme content across the internet. MORE
Sweetspot: AI for Government Contracting
Sweetspot uses AI to streamline the process of finding and bidding on government contracts, making a complicated process much simpler. It's a clever application of AI, targeting a niche use case but with tremendous value. MORE
Magna: Sanctuary AI’s humanoid robot test drive
Sanctuary AI's robots are being tested in car manufacturing, hinting at a future where robots could be common in industries. If these tests succeed, we might start seeing more robots in places we never expected. MORE
V7Labs: It's here.. Work automation for the multimodal AI era.
V7 Go introduces a way to automate repetitive tasks using AI, scaling up operations without losing reliability. This tool could change how we handle bulk tasks, making work life a bit easier. MORE
Have an AI Idea and need help building it?
When you know AI should be part of your business but aren’t sure how to implement your concept, talk to AE Studio.
Elite software creators collaborate with you to turn any AI/ML idea into a reality–from NLP and custom chatbots to automated reports and beyond.
AE Studio has worked with early stage startups and Fortune 500 companies, and we’re ready to partner with your team. Computer vision, blockchain, e2e product development, you name it, we want to hear about it.
Join the conversation