惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

博客园_首页
Hacker News: Ask HN
Hacker News: Ask HN
V
Visual Studio Blog
Hugging Face - Blog
Hugging Face - Blog
博客园 - 【当耐特】
I
InfoQ
A
About on SuperTechFans
宝玉的分享
宝玉的分享
Project Zero
Project Zero
C
CXSECURITY Database RSS Feed - CXSecurity.com
量子位
The Register - Security
The Register - Security
大猫的无限游戏
大猫的无限游戏
Blog — PlanetScale
Blog — PlanetScale
Simon Willison's Weblog
Simon Willison's Weblog
C
Cyber Attacks, Cyber Crime and Cyber Security
T
The Exploit Database - CXSecurity.com
Cyberwarzone
Cyberwarzone
L
LINUX DO - 热门话题
博客园 - 聂微东
CTFtime.org: upcoming CTF events
CTFtime.org: upcoming CTF events
P
Privacy International News Feed
T
Tenable Blog
博客园 - 叶小钗
P
Palo Alto Networks Blog
S
Securelist
F
Full Disclosure
Help Net Security
Help Net Security
爱范儿
爱范儿
Cisco Talos Blog
Cisco Talos Blog
F
Fortinet All Blogs
Google DeepMind News
Google DeepMind News
IT之家
IT之家
S
Secure Thoughts
Martin Fowler
Martin Fowler
The Cloudflare Blog
阮一峰的网络日志
阮一峰的网络日志
小众软件
小众软件
G
GRAHAM CLULEY
J
Java Code Geeks
Forbes - Security
Forbes - Security
Application and Cybersecurity Blog
Application and Cybersecurity Blog
A
Arctic Wolf
L
LangChain Blog
Webroot Blog
Webroot Blog
TaoSecurity Blog
TaoSecurity Blog
月光博客
月光博客
T
The Blog of Author Tim Ferriss
Engineering at Meta
Engineering at Meta
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知

Benson's blog

Enjoy life Internship AI on academic research How AI Will Change the Mobile Ecosystem Look ahead Goodbye 2025 Hacker News to Kindle Another project How to imporve english Introduction of Fraud detection PopTranslate Last day in netease Better idea between Copilot-typed and CLI-typed assistant Gemini-cli LLM Post-Training experience Papers I readed recently about LLM application Difference between LLMs and traditional computer technology GRPO Weekly-#26 AI Application Weekly-#25 AI infra and application Weekly-#24 First week as LLM inference engineer Weekly-#23 seeking job Weekly-#22 2025 New Year AutoSwitch Translate Goodbye 2024 Weekly-#20 Breaking of glass Cross Entropy Loss of Triton Weekly-#18 Cross Entropy Loss of Triton Weekly-#17 Triton Puzzles Weekly-#16 AutoBuilder Weekly-#14 Accident in life Weekly-#13 Trying of xiaohongshu Weekly-#12 summary of LLM acceleration Outline of LLM acceleration Weekly-#11 Copilot-type products Weekly-#10 Preparation for next journey Weekly-#9 Startup of YouTube Notes of flash-attention How to learn knowledge in new fields? Weekly-#8 Start Reading Notes of LoRA Acceleration of LLM - Matrix Multiplication Weekly-#8 Summary for two month Weekly-#7 Staying home Weekly-#6 Cost of PopTranslate Weekly-#5 Updating of PopTranslate Validated example of LLM acceleration Weekly-#4 First insight of LLM accelerate Weekly-#3 PopTranslate Weekly-#2 The fail of first product Weekly-#1 First week of indie develop slack迁移discord 雅思备考 2024Q3 中文博客合集 English Diary in May 五一游记 开始休假 离职前的状态 2024-01-01 duckdb 看懂的第一个PR learning english in October learning english in September learning english in August top hack news 收集 大模型调研 自动驾驶的小玩具 旅游 扬州+苏州 small talk of learning english 新年新气象-碎碎念 刷剧 感染新冠 强化学习简介 神经网络解释性 全局的模型无关解释方法合集 社区发现算法概览 图神经网络入门(GNN) 我的第一款 iOS APP AtCoder Beginner Contest 268 人的信息输入方式对比 重叠社区检测 人穷极一生到底在追求什么 重拾生活规划 社区发现算法 - Louvain 《幸福的方法》 读《人类简史》有感 妙峰山骑行 黑客帝国 特征交互 特征工程 累计局部效应图 模型解释性-PDP 模型解释性 Web3 入门科普 总结 2022.4 孪生网络做 query 相似度任务 学习 2022.4 Imagen DeBERTa 读论文 用CNN做query相似度任务
Weekly-#15 Starting of tanble tennis
Benson · 2024-12-08 · via Benson's blog

Product

AutoBuilder

I want to build AutoBuilder which can build simple product totally by AI model, like static websites. Compared copilot or cursor, AutoBuilder and see the result base on the current code after we build and run them.

As a result, AI image model can provide feedback for current code, which can help text model to be better.

First version is too ideal. After I show the static website, I must input sth and click one button to test the result of function, it’s hard to implement this process automatically. So I change to another direction.

I think developer should involve more in the process of programming and building. Developer can be the role of evaluation, he can provide feedback for the result of websites. Meanwhile some of the reasons I do this are I realize the weaknesses of AI model, so, more work of developer are necessary to build a mature product. What’s more, I can update the process if my though works, This is important lesson I learnt from this project.

Think about below points when you design the road of your new product.

  1. Start project from point where ROI is highest, low cost and high benefit.
  2. Make it iterable when there are more features.

Triton

Find a great project for me to learn Triton, which is Triton-Puzzles

Reading

Thought

Table tennis

I enjoy table tennis a lot this week, shows the importance of positive feedback. That’s why I think money also can be important when working.

Evaluation of AI

Evaluation seems became the most important thing in AI period.

Most people talked about the importance of prompt, but the sound is less and less with the development of AI. The existing of prompt engineer show the weaknesses of AI in the initial period of development, reinforcement learning from human feedback can totally solve this kind of problem, but this is relative expensive, Google queries are also great option to meet this kind of problem. When I use ChatGPT, I also realize this’s not big deal. As a result, How to Evaluate the validation of response of AI models becomes the only issue.

Evaluation is a high level task than generations, the straightforward method is the done by human. Many product are produced in this kind of form, like AI coding assistant. Actually, human evaluation and feedback can be the bridge between AI model and mature products. When AI models are not mature enough, human can provide feedback one time again and again, that’s why we called AI as an assistant.

How to check how much worked can be replaced by AI models? Try to find an automatic method of evaluation, if there is, it can be replaced finally. This remind me the work as an engineer in anti-fraud field. The most difficult part of my work is to determine whether if one user or flow a fraud. This is hard to be decide by AI model. Consequently, we can implement AI models in fields where there is solid evaluation method on. Coding is definitely first field.

Competitive programming is also filed where I want to build a AI tool. It’s easy to get a solid bad case when wrong answer is returned for submitted code. I will try it after AutoBuilder.

History news

How about to read news happened in the past, like one year ago or three years ago. Like HackerNews one years ago content.

We can know whether the thing are important or not, we know more how to evaluate whether one thing it’s important or not.

What kinds of documents of news can be read

  1. The economist
  2. Hacker News
  3. v2ex
  4. News magazine

Internet is a great tool to realize that.

Keyboard for iPad

Today, I bought a new keyboard for my iPad, which is useful when I just brought a iPad outside. Actually, I don’t have a laptop since I was in the third years in my university. Laptop is always not a good choice whether in any environment, it more likely an intermediate product. When people work in a stable place, like working environment or personal house, he usually need a huge monitor and comfortable keyboard, the original screen and keyboard of laptop is unused. In contrast, laptop is a little heavy to carry outside. For me, I don’t like carry such a huge thing outside. Of course, more important reason is that I don’t have a laptop right now and just have a iPad, which also can support my demand of outside temporarily.

I don’t know why I have more patience to write something when I write towards a small screen, it’s an interesting phenomenon. Maybe it’s easy to concentrate for me when towarding a relative small screen.

Summary

Table tennis seems make my life better. Enjoy more.

Spent less time on Phone and short videos, which is a great trend.

This week

  • Work visa: hold
  • Summer internship: 15 application
  • Triton: Not start
  • Youtube: 1
  • Reading and Writing: read less but write more
  • Exercise: 4
  • AutoBuilder: new project, 60%

Next week

  • Summer internship
  • AutoBuilder
  • Triton
  • Youtube
  • RW: Read history news
  • Exercise

This post is licensed under CC BY 4.0 by the author.