??$2000 a month LLMs??AI scale by 2030 ??4,500 dev years of work

??$2000 a month LLMs??AI scale by 2030 ??4,500 dev years of work

OpenAI is in talks to raise a new round from Thrive Capital, Apple, Microsoft, and Nvidia to a valuation of $100 billion. While they consider changing their corporate structure there are also rumors of $2000 a month subscriptions for their new LLMs.?

They’re not the only one raising prices – Canva just announced a 300% annual price increase based on value from generative AI features.?

While companies work to make generative AI profitable, Elon Musk launched xAI’s Colossus – a cluster of 100,000 Nvidia H100 GPUs. The scale of AI continues to grow – Meta has reached 400 million monthly active users and their Llama models have been downloaded 350 million times.

This report from Epoch AI goes into deep analysis of the scalability of AI through 2030.

Four of the main constraints on AI scale:

  • Power supply
  • Chip manufacturing capacity
  • Data scarcity
  • Latency wall

While there are new competitors in the AI chip market, the first MLPerf benchmarks for Nvidia Blackwell, AMD, Google, and Untether GPUs show that the H100s still have a strong lead for inference.


While market competition for AI exists in billions of dollars, AI chips, power supply, and data center scale, all companies are trying to build AI products that actually make or save money.

Klarna is a company that replaced 700 full-time human customer support agents with AI. While there are endless arguments about whether or not that’s a good thing, customer support agents are one of the only proven business cases for AI. Anthropic just released a Quickstart on Github so companies can build their own CS agents faster with Claude.?

And it looks like some huge work efficiencies in software development are actually possible. Amazon CEO Andy Jassy also just announced they saved 4,500 developer years worth of work using Amazon Q to upgrade production apps to Java 17.

As always, here are the most useful links in AI and software development that we’ve found:

Useful Links & News

Google AI that can hear signs of sickness

Magic Dev announces 100m token context window

California legislature passes AI kill switch bill?

Midjourney reaches into hardware

Google Gemini gets new image features

Water consumption on data center alley

Anthropic Artifacts available for all users

Lessons from big tech on building resilient payment systems

How Amazon AWS Lambda works


AI Research Papers & Technical Links

The effects of generative AI on high skilled work

Anthropic releases Claude system prompts

GPT-4 judged more human than humans

Efficient LLM scheduling by learning to rank

Text2SQL is Not Enough: unifying AI and databases with TAG

Reward modeling as next-token prediction


PS

We hope these newsletters help you stay informed. We want to know what you think is useful or hear about anything you're excited about. Tell us in comments below!


要查看或添加评论,请登录

Codingscape的更多文章

社区洞察

其他会员也浏览了