视频+艺术字 酷

在一篇最新的衡量AI生产力的论文里，有一些非常有趣的数据。
- 质量和速度的全面提升，使用人工智能的顾问平均完成了12.2%更多的任务，并且产生的结果质量比没有使用人工智能的顾问高40%。
- 拉平技能，得分最低的顾问在使用人工智能后，他们的表现有了最大的提升，达到了43%，而顶级的顾问只提升了 17% 。
论文地址：
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4573321

Ssrn

Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity…

The public release of Large Language Models (LLMs) has sparked tremendous interest in how humans will use Artificial Intelligence (AI) to accomplish a variety o

OpenMoE
仅使用 30% 的成本达到 LLaMa-7B 水平。 https://github.com/XueFuzhao/OpenMoE/tree/main

Open Interpreter 开源版本的 ChatGPT code Interpreter
It’s an open-source version of OpenAI’s Code Interpreter https://vxtwitter.com/mckaywrigley/status/1699409632396198393

vxTwitter / fixvx

Mckay Wrigley (@mckaywrigley)

Open Interpreter is amazing! It’s an open-source version of OpenAI’s Code Interpreter by @hellokillian. Best implementation I’ve seen yet - it just works. Tools like these give us a peek at what AI-pow… ***💖*** 1.33K ***🔁*** 223

Meta's Segment Anything Model on the CPU https://vxtwitter.com/ggerganov/status/1699092329607450880

vxTwitter / fixvx

Georgi Gerganov (@ggerganov)

sam.cpp ***👀*** Inference of Meta's Segment Anything Model on the CPU Project by @YavorGI - powered by http://ggml.ai https://t.co/u7vLhdsUzE ***💖*** 1.99K ***🔁*** 283

code llama 的基座太好了…
ft 也很猛…

基于开源模型的代码能力超过 GPT4。
一个标志性的事件。
https://www.phind.com/blog/code-llama-beats-gpt4

Introducing SeamlessM4T, the first all-in-one, multilingual multimodal translation model.

This single model can perform tasks across speech-to-text, speech-to-speech, text-to-text translation & speech recognition for up to 100 languages depending on the task. https://vxtwitter.com/MetaAI/status/1694020437532151820?s=20

vxTwitter / fixvx

AI at Meta (@AIatMeta)

Introducing SeamlessM4T, the first all-in-one, multilingual multimodal translation model. This single model can perform tasks across speech-to-text, speech-to-speech, text-to-text translation & speech recognition for up to 100 languages depending on the…

GPT-3.5 API 支持 Fine-tuning 了。
通过写一系列的问答对来进行FT，可以训练自己指定风格的角色或具有特定功能的工具。
还可以把Prompt压缩到模型里，优化最多 90% 的prompt占用。
对应的 API 价格也水涨船高，FT 后的价格会变为原来的9倍，仍然比4便宜。（这次定价非常狠）
https://openai.com/blog/gpt-3-5-turbo-fine-tuning-and-api-updates

Openai

GPT-3.5 Turbo fine-tuning and API updates

Developers can now bring their own data to customize GPT-3.5 Turbo for their use cases.

为了理解AI-Native，我们做了几款AI应用https://mp.weixin.qq.com/s/QCpGXsqYRzfHIPBiHAlsRw?utm_source=substack&utm_medium=email

Weixin Official Accounts Platform

拾象实践：为了理解AI-Native，我们做了几款AI应用

快速了解一个行业的最佳的方式就是参与其中。