Josh,Teng Yan
October 5, 2024

Weekly AI Edge #15

23andWe DataDAO launches to aggregate genetic data from....23andMe

GM!

In this edition of our weekly AI Edge, we cover:

  • Crypto AI stumbles from the selloff
  • 23andMe’s valuable genetic data up for grabs
  • Our favourite tweets on X

While this made us laugh, we hope the Crypto AI projects don’t start following suit.

Introducing AverageEditorAI, the open source AI powered code editor, making Development process 6.9x faster and reducing debugging efforts by 420%

It's a fork of  BlueberryAI, which is fork of PearAI, which is a fork of Continue, which is a fork of VSCode  

Investors please no… x.com/i/web/status/1…

— Average Engineer (@AverageProMax)
7:54 PM • Sep 29, 2024

🦍 State of the Market..

Source: Coingecko

What a rollercoaster!

The Crypto AI market had a rough week, dropping 21.2% to $24.8B, down from $31.5B last week. It was one of the worst performers, especially compared to DeFi (-8.7%) and Memes (-8%).

News of escalating tensions between Israel and Iran, coupled with the looming election season, has the market on edge.

Surprisingly, TAO has shown remarkable strength. Its price remains at $559, a 4% rise from last week. The Bittensor community has strong hands, clearly. 🤚🤚

The rest of the Crypto AI market stumbled, with tokens like NEAR, NMT, and ASI all suffering 15% drops.

One token that has received a lot of attention is Morphware’s token XMW.

It’s skyrocketed 158% in the last seven days, showing how quickly low liquidity tokens can experience sharp gains. Morphware is an AI project built as a Layer 3 on Starknet, offering compute at lower prices thanks to access to Paraguay’s renewable energy.

There’s been a wave of influencer shills on X, likely driving the recent pump. After diving into the documentation, we didn’t find much to get excited about yet—so tread carefully if you’re trading it!

📊 Chart of the Week

AI Is Taking Over Our Jobs

Source: The Short Term Effects of Generative AI on Employment

In August, researchers from NYU dropped a fascinating study that digs into the ripple effects of generative AI on the freelance market, specifically using data from UpWork, a major freelancing platform.

The findings paint a clear picture: freelancers are feeling the squeeze.

According to the study, freelancers have seen a 2% dip in the number of monthly jobs (Panel A) and a 5% drop in monthly revenue (Panel B) over the past year. These are real financial pain points if you’re making your living through freelance work.

As AI continues to improve, everyone is going to feel the impact, and it’s not just the creatives.

The pace of change will only accelerate.

🌴 23andMe….Or 23andWe?

Over the past week, many people have been discussing 23andMe’s decline and what it means for 15 million users’ genetic data. Most disconcerting is that those 15 million users have no control over what happens to their data next.

Rumours have swirled about various SPACs being spun up to acquire the company. But the latest news is that CEO Anne Wojcicki plans to take the company private and is not considering third-party offers.

How about setting up a Bittensor subnet to train AI on this data?

Someone should acquire @23andMe (stock is valued at $150M today), open source the data and bootstrap a genetics subnet on Bittensor with the newly unlocked DNA database of 15 million human genotypes. The incentive mechanism would reward contributions to this open database. 🧬

— JJ (@JosephJacks_)
6:13 PM • Sep 28, 2024

This week, a Data DAO called 23andWe launched on Vana, offering users a chance to upload their 23andMe data files in exchange for tokens. These tokens give users governance rights over how their data is used.

The potential for aggregating healthcare data directly from consumers is massive, but there’s a huge friction point: personal health data is notoriously difficult to access. Just try requesting your old X-ray from a hospital.

If a team can crack the code—whether through engineering or clever incentives—the opportunity in healthcare data is staggering.

🏆 Caught Our Eyes..

Source: Mind Network

Fundraising

  • Mind Network secures $10 million in Pre-A funding to develop FHE layer for AI and PoS Networks

Project Updates

  • Talus introduces Nexus, an onchain framework for building and managing AI Agents
  • Grass explains Live Context Retrieval (LCR) and its integration into language model training
  • Hyperspace crosses 4000+ connected nodes, becoming a top global P2P network for AI
  • Nous Research shares clips from the recent NousCon event featuring speakers from Haize Labs, Hugging Face, Arcee AI and .txt AI
  • Prime Intellect reports ~4.4x monthly growth as more AI developers and startups turn to Prime Intellect for compute resources
  • 0G labs pivots to build a decentralised AI operating system
  • Spectral Labs showcases creating a copytrading agent to track one of the most profitable addresses on BRETT
  • Theoriq teases how users can participate in their testnet
  • Panthalia announces a verification layer for low-cost P2P compute scaling to large ML models
  • AO users claim 750 relics, which are part of a limited AO NFT collection. Collect all 5 relics to unlock your own Omega Dumdum NFT

Incentives / Rewards Programs

  • Spheron Network introduces Fizz Node, offering FN point rewards for providing compute on their platform

Bittensor

  • Bittensor TAO allocation in Grayscale AI Fund increases from 2.6% to 27.6%
  • BitMind AI releases offline DeepFake Detection Benchmark with CAMO framework
  • Highlights of the Bittensor x Google event from the Dippy team

Hyperbolic

  • Hyperbolic’s inference service reaches over 1 billion LLM tokens daily across 22 AI models
  • Hyperbolic’s deep dive on their product and features

🧠 Web2 AI Corner

Liquid models achieve state of the art performance

Major AI developments this week!

  • Liquid, a spin-off from MIT, has released its first series of foundational models built on a novel non-transformer architecture. This new architecture significantly outperforms transformer-based models like Llama 3.2 for the first time. It shows how much scope for improvement there is on the architecture side.
  • OpenAI releases Canvas, a fresh interface for GPT, along with new tools like vision fine-tuning and a real-time API for speech-to-speech applications.
  • NVIDIA chip-maker competitor Cerebras files for an IPO

🔥 On X..

The rise of AI voice agents

AI voice agents are on 🔥

We're moving from the innovator -> early adopter part of the curve, with new startups sprouting up weekly to serve different verticals.

What @illscience and I are seeing @a16z, and why we're excited 👇

— Olivia Moore (@omooretweets)
3:50 PM • Oct 1, 2024

Canvas is trained on fully-synthetic data

The future is synthetic indeed!

We see more synthetic data use cases in pre-training, post-training, and evals each day.

(Great work by @karinanguyen_ and @OpenAI)

How OpenAI's New Canvas Model Utilizes Synthetic Data in the three use cases above:

1. Accelerated Model… x.com/i/web/status/1…

— kerimkaya (@kerimrocks)
10:07 PM • Oct 3, 2024

Lars on AI agents being “fake”

all “Agent” projects today are fake

that is, they sell you glorified bots that are larping as Agents

some of them use AI, some of them don’t. but bots using AI is not new. and it is not what make them Agents

a write-up of how True Agents will change the way you make money 🧵

— Lars 🛠️🪙 (@0xLars_)
9:52 AM • Oct 1, 2024

a16z’s podcast on decentralised AI featuring Nous Research

Could the next big open source model be built by a global network of independent builders?

@NousResearch’s DisTrO is showing it’s possible—training powerful AI models using the public internet, without relying on big tech or massive data centers.

In this episode of the AI +… x.com/i/web/status/1…

— a16z (@a16z)
11:03 PM • Sep 27, 2024

DMD reviews the yield calculations on Wayfinder’s PROMPT token

x.com/i/article/1840…

— DMD (@Daveeemor)
5:52 PM • Sep 30, 2024

Menlo Ventures discusses how AI agents are driving the next wave of enterprise automation

Beyond Bots: How AI Agents Are Driving the Next Wave of Enterprise Automation - Menlo Ventures

Unlike previous generations of software, new cognitive architectures enable AI agents to dynamically automate end-to-end processes.

menlovc.com/perspective/beyond-bots-how-ai-agents-are-driving-the-next-wave-of-enterprise-automation

That’s it for this week! If you have specific feedback or anything interesting you’d like to share, please just reply to this email. We read everything.

Cheers,

Teng Yan & Joshua

This newsletter is intended solely for educational purposes and does not constitute financial advice. It is not an endorsement to buy or sell assets or make financial decisions. Always conduct your own research and exercise caution when making investment choices.

Others You May Like