Pinned
00:00
Jay Hack
5,434 posts
Jay Hack
@mathemagic1an
- Parrots are clearly intelligent enough to understand video UIs They also apparently prefer watching videos of other parrots 🤔 This implies an opportunity for a "parrot streaming" platform. Looking for a team who is as excited about this opportunity as Iam
00:00 - Got called a Boomer at the office for using Cursor Everybody needs to take a deep breath
- Introducing text-to-figma: build and edit @figma designs with natural language! Join the waitlist here: text-to-figma.com 1/n
00:00 - Sounds like rumors were accurate: GPT-4 will come in multimodal flavors (including video!) GPT-4 release next week heise.de/news/GPT-4-is-…
- 🤔 What comes after Copilot? My take: a conversation with your codebase! Introducing Tensai, your repo-level code assistant ❔ Ask complex questions ✅ Automatically generate PRs for complex tasks TensaiCode.com More👇
00:00 - My thoughts on Toolformer IMO the most important paper in the past few weeks. arxiv.org/abs/2302.04761 Teach an LLM to use tools, like a calculator or search engine, in a *self-supervised manner* Interesting hack to resolve many blind spots of current LLMs Here's how 👇
- Google announces Bard, their ChatGPT competitor. 👏👏👏 blog.google/technology/ai/…
00:00 - An intriguing trend in AI 🤖: “Models all the way down” (aka "stacking") Have models invoke other models, then watch as emergent intelligence develops ✨ Here’s a discussion of what, how, and why this is important to watch 👇
- DeepSeek v3 is an order of magnitude cheaper because it likely trained on frontier model outputs, in obvious violation of ToS ToS laundering by training on DeepSeek outputs is impossible to prevent. Does not bode well for economics of training frontier models
- Been waiting for something like this for a while: Etched.ai: printing specific model architectures on a chip. Claims 100x speedup over GPUs. Not hard to imagine. What happens when you can run a GPT forward pass at the speed of electricity, no clock needed?
- ChatGPT is capable of de-minifying JS, including adding descriptive variable names. Nice.
- The 'data engine' idea of defensibility in AI may not be as defensible as we thought: In SELF-INSTRUCT, authors get GPT-3 to generate it's *own* dataset for instruction tuning, outperforming vanilla GPT-3 and comparable to InstructGPT. arxiv.org/pdf/2212.10560… Here's how 👇
- What if you could fit an *entire codebase* in an LLM? 🤔 "Efficiently Scaling Transformer Inference" (11/2022) arxiv.org/pdf/2211.05102… Jeff Dean + co break out all the hacks to scale PALM-540B's context length to 43,000 tokens! Here's how 👇
