🥟 Chao-Down #350 YouTube makes AI deals with music record labels, Meta creates a watermark for AI-generated speech, AI's insatiable need for energy strains global power grids

Plus, a look at the telltale words to identify AI-generated text.

Jul 02, 2024

Will we soon see a $1 billion price tag for the latest AI models? Based off current trends and scaling laws that have proven to be resilient, the costs to build leading foundation models may soon eclipse that mark.

The folks at Epoch AI charted the data starting from 2017 with early models such as AlphaGo to the latest iteration of GPT-4. They conclude:

Only a few large organizations can keep up with these expenses, potentially limiting innovation and concentrating influence over frontier AI development. Unless investors are persuaded that these ballooning costs are justified by the economic returns to AI, developers will find it challenging to raise sufficient capital to purchase the amount of hardware needed to continue along this trend.

Does this mean AI will only be dominated by big tech? Or can we discover some breakthrough to train cheaper, more efficient models?

-Alex, your resident Chaos Coordinator.

What happened in AI? 📰

Amazon reportedly thinks people will pay up to $10 per month for next-gen Alexa (Engadget)

Figma AI: A New Era Of Design (Raw.Studio)

YouTube is trying to make AI music deals with major record labels (The Verge)

The telltale words that could identify generative AI text (Ars Technica)

AI’s Insatiable Need for Energy Is Straining Global Power Grids (Bloomberg)

Meta has created a way to watermark AI-generated speech (MIT Technology Review)

Always be Learnin’ 📕 📖

AI and the future of game design (Creative Bloq)

Navigating the Promise and Pitfalls of AI (Figma Blog)

Indexing all of Wikipedia, on a laptop (foojay.io)

Projects to Keep an Eye On 🛠

GitHub - khoj-ai/khoj: Your AI second brain. Get answers to your questions, whether they be online or in your own notes. (Github)

AgentOps-AI/tokencost: Easy token price estimates for 400+ LLMs (Github)

GitHub - Nike-Inc/koheesio: Python framework for building efficient data pipelines. It promotes modularity and collaboration, enabling the creation of complex pipelines from simple, reusable components. (Github)

The Latest in AI Research 💡

Augmenting Language Models with Long-Term Memory (arxiv)

Less is More: Accurate Speech Recognition & Translation without Web-Scale Data (arxiv)

BABILong: Testing the Limits of LLMs with Long Context Reasoning-in-a-Haystack (arxiv)

The World Outside of AI 🌎

How thousands of Americans got caught in fintech’s false promise and lost access to bank accounts (CNBC)

EVs still have major quality problems, and it’s mostly about the software (The Verge)

The big idea: can you inherit memories from your ancestors? (The Guardian)

Social-Media Influencers Aren’t Getting Rich—They’re Barely Getting By (WSJ)

How Shein and Temu Snuck Up on Amazon (bigtechnology.com)

Why Men Are ‘Rawdogging’ Flights (GQ)