#ai-learning

2023-12-01
leechiuhui 00:04:46
@leechiuhui has joined the channel
2023-12-05
kiang 14:22:50
http://cchsu.info:921/photos/

可以上傳圖片或影片自動判斷是否經過深層偽造
kiang 14:22:50
http://cchsu.info:921/photos/

可以上傳圖片或影片自動判斷是否經過深層偽造
kiang 14:25:46
聽起來很多學校的研究室都很難自己建置運算池，必須依賴國網中心資源，有個分享者提到最多只能分配 8 張 a100
kiang 14:25:46
聽起來很多學校的研究室都很難自己建置運算池，必須依賴國網中心資源，有個分享者提到最多只能分配 8 張 a100
2023-12-06
kiang 11:17:22
rt from @pofeng - llm 運作過程的視覺化 https://bbycroft.net/llm
bbycroft.net
LLM Visualization
A 3D animated visualization of an LLM with a walkthrough.
Teemo 2023-12-06 12:08:50
好潮
paulpengtw 2023-12-06 14:54:04
炫...
kiang 11:17:22
rt from @pofeng - llm 運作過程的視覺化 https://bbycroft.net/llm
Teemo 2023-12-06 12:08:50
好潮
paulpengtw 2023-12-06 14:54:04
炫...
Teemo 12:08:50
好潮
paulpengtw 14:54:04
炫...
2023-12-07
kiang 02:30:17
不知道實際試用時有沒有展示的這麼順暢
https://www.businessyee.com/article/2822-google-gemini-llm-ai-model
kiang 02:30:17
不知道實際試用時有沒有展示的這麼順暢
https://www.businessyee.com/article/2822-google-gemini-llm-ai-model
商益
【快訊】Google閃電發表最新AI模型！看得見、能交談的多模態生成式 AI「Gemini」 - 商益
Google 稍早（美國時間 6 日）閃電推出最新生成式 AI 模型 Gemini。 Gemini 的特別之處在於，它從零開始時、就被設計為一個多模態學習的模型。以影片為例，一名女子一邊準備蔬菜煎蛋捲的材料，一邊與 Gemini 口語對話討論食譜，依照 Gemini 的指示一步步完成料理。
2023-12-08
kiang 09:53:19
新的聚焦程式碼的 LLM
Magicoder : Source Code Is All You Need.

Demo: https://67cc8c194b67d37b94.gradio.live/
paper:https://arxiv.org/abs/2312.02120
github: https://github.com/ise-uiuc/magicoder
kiang 2023-12-08 10:12:06
Ollama 可以直接拉下來玩
https://ollama.ai/library/magicoder
kiang 09:53:19
新的聚焦程式碼的 LLM
Magicoder : Source Code Is All You Need.

Demo: https://67cc8c194b67d37b94.gradio.live/
paper:https://arxiv.org/abs/2312.02120
github: https://github.com/ise-uiuc/magicoder
arXiv.org
Magicoder: Source Code Is All You Need
We introduce Magicoder, a series of fully open-source (code, weights, and data) Large Language Models (LLMs) for code that significantly closes the gap with top code models while having no more than 7B parameters. Magicoder models are trained on 75K synthetic instruction data using OSS-Instruct, a novel approach to enlightening LLMs with open-source code snippets to generate high-quality instruction data for code. Our main motivation is to mitigate the inherent bias of the synthetic data generated by LLMs by empowering them with a wealth of open-source references for the production of more diverse, realistic, and controllable data. The orthogonality of OSS-Instruct and other data generation methods like Evol-Instruct further enables us to build an enhanced MagicoderS. Both Magicoder and MagicoderS substantially outperform state-of-the-art code models with similar or even larger sizes on a wide range of coding benchmarks, including Python text-to-code generation, multilingual coding, and data-science program completion. Notably, MagicoderS-CL-7B based on CodeLlama even surpasses the prominent ChatGPT on HumanEval+ (66.5 vs. 65.9 in pass@1). Overall, OSS-Instruct opens a new direction for low-bias and high-quality instruction tuning using abundant open-source references.
ise-uiuc/magicoder
Magicoder: Source Code Is All You Need
kiang 2023-12-08 10:12:06
Ollama 可以直接拉下來玩
https://ollama.ai/library/magicoder
kiang 10:12:06
Ollama 可以直接拉下來玩
https://ollama.ai/library/magicoder
2023-12-09
isabelhou 11:40:07
今天工作坊共筆 https://hackmd.io/k0hMoHegRYOCNjCIWdJ7tA @sdfghj1001
hackmd.io
HackMD - Collaborative Markdown Knowledge Base
  • 2
Peter 2023-12-09 11:41:50
@tomy7912348
isabelhou 11:40:07
今天工作坊共筆 https://hackmd.io/k0hMoHegRYOCNjCIWdJ7tA @sdfghj1001
Peter 2023-12-09 11:41:50
@tomy7912348
Peter 11:41:50
@tomy7912348
2023-12-12
kiang 09:02:19
https://g0v-tw.slack.com/archives/C02G2SXKX/p1702342860206059
kiang 09:02:19
https://g0v-tw.slack.com/archives/C02G2SXKX/p1702342860206059
YouTube
正體中文開源資料集 社群課程：大語言模型的民主化
#正體中文開源資料集社群課程【大語言模型的民主化】線上直播時間：2023年12月12日，晚上8:00-8:30課程簡報 <https://www.canva.com/design/DAFzN3te2_Y/lQS9_mNO5nRzY83JCXFImA/viewg0v> tw hackath56n — 中場短講 — 從 ...
  • 1
  • ❤️2
2023-12-13
Peter 21:50:02
感覺 @teemocogs 會有興趣的題目，但是我不知道為何他會出現在我的推播：https://www.youtube.com/watch?v=PB7HacNKsRM
YouTube
實證臨床決策支援系統中的生成式AI
  • 3
Teemo 2023-12-13 22:29:54
感謝 Tag 我

講者 14:00-15:00 在交大演講
Peter 21:50:02
感覺 @teemocogs 會有興趣的題目，但是我不知道為何他會出現在我的推播：https://www.youtube.com/watch?v=PB7HacNKsRM
Teemo 2023-12-13 22:29:54
感謝 Tag 我

講者 14:00-15:00 在交大演講
Teemo 22:29:54
感謝 Tag 我

講者 14:00-15:00 在交大演講
2023-12-14
Peter 20:59:36
https://www.cnbc.com/2023/12/13/openai-and-axel-springer-strike-unprecedented-deal-[…]egment_id=152568&te=1&user_id=a8d646d000eb1cddb46d7af415e0a45b
CNBC
OpenAI and Axel Springer strike unprecedented deal to offer news in ChatGPT
OpenAI and Axel Springer, the global news publisher, have struck an unprecedented deal allowing ChatGPT to summarize news stories.
Peter 20:59:36
https://www.cnbc.com/2023/12/13/openai-and-axel-springer-strike-unprecedented-deal-[…]egment_id=152568&te=1&user_id=a8d646d000eb1cddb46d7af415e0a45b
2023-12-15
Peter 09:35:55
https://future.mozilla.org/blog/introducing-llamafile/?s=04&fbclid=IwAR2ADJJDTscBfnINEsMynLZADqHIU7S-sBPH27Z2Rj4_OHOZg8DHds8Alq0
Peter 09:35:55
https://future.mozilla.org/blog/introducing-llamafile/?s=04&fbclid=IwAR2ADJJDTscBfnINEsMynLZADqHIU7S-sBPH27Z2Rj4_OHOZg8DHds8Alq0
Mozilla Innovations
llamafile: bringing LLMs to the people, and to your own computer
  • 1
Peter 16:05:30
https://www.accupass.com/event/2312120710021774788931?utm_source=facebook&utm_medium=Social&utm_campaign=accu_231213cu7et7w0
accupass.com
《時代》百大AI影響力人物－李飛飛博士　唯一新書演講｜Accupass 活動通
李飛飛博士，是全球極具影響力的AI領域學者，被李開復譽為「AI領域的良心和先驅者」。天下文化特別邀請到李飛飛本人，以及台大資工系副教授陳縕儂，一起在線上和大家談談新書《AI科學家李飛飛的視界之旅》。邀請各階段的家長、教育者，或關心AI如何影響未來的讀者，把握機會，加入李飛飛博士唯一新書線上演講。
  • 1
Peter 16:05:30
https://www.accupass.com/event/2312120710021774788931?utm_source=facebook&utm_medium=Social&utm_campaign=accu_231213cu7et7w0
2023-12-17
Peter 10:52:25
https://g0v.hackmd.io/2AwsZ3TGSfSudD0tEjcAuw
kiang 2023-12-17 11:06:21
看到新聞的時候也是擔心 AI 幻覺問題，畢竟要從頭訓練會需要花大量的心力做資料清理與標註
Peter 2023-12-17 11:15:48
沒錯（你是不是在現場XD
kiang 2023-12-17 11:21:07
我只是曾經想玩資料，結果看到一團混亂就默默把電腦關上 XD
Peter 10:52:25
https://g0v.hackmd.io/2AwsZ3TGSfSudD0tEjcAuw
g0v.hackmd.io
司改會xSMC 20231217 講座紀錄 - HackMD
  • 1
kiang 2023-12-17 11:06:21
看到新聞的時候也是擔心 AI 幻覺問題，畢竟要從頭訓練會需要花大量的心力做資料清理與標註
Peter 2023-12-17 11:15:48
沒錯（你是不是在現場XD
kiang 2023-12-17 11:21:07
我只是曾經想玩資料，結果看到一團混亂就默默把電腦關上 XD
kiang 11:06:21
看到新聞的時候也是擔心 AI 幻覺問題，畢竟要從頭訓練會需要花大量的心力做資料清理與標註
Peter 11:15:48
沒錯（你是不是在現場XD
kiang 11:21:07
我只是曾經想玩資料，結果看到一團混亂就默默把電腦關上 XD
2023-12-18
Kt 17:08:41
@katy881115 has joined the channel
2023-12-22
paulpengtw 15:35:11
剛剛才知道原來先前 Lessig 教授 "AI and Democracy" 演講和對談影片已經上傳了，想看的朋友歡迎～
https://www.youtube.com/watch?v=W3M-H6pYHWg 含 isabelhou clkao 與談完整版影片
https://vimeo.com/867832751 僅 lessig 教授演講影片
  • 👍5
  • 1
paulpengtw 15:35:11
剛剛才知道原來先前 Lessig 教授 "AI and Democracy" 演講和對談影片已經上傳了，想看的朋友歡迎～
https://www.youtube.com/watch?v=W3M-H6pYHWg 含 isabelhou clkao 與談完整版影片
https://vimeo.com/867832751 僅 lessig 教授演講影片
2023-12-24
Teemo 20:51:34
https://youtu.be/VePvTLH7ZmQ
Teemo 20:51:34
https://youtu.be/VePvTLH7ZmQ
YouTube
AI and Clinical Practice—Can AI Accelerate Medical Education?
  • 👍1
2023-12-25
Teemo 21:36:58
https://www.angle.com.tw/book.asp?BKID=15580
Teemo 21:36:58
https://www.angle.com.tw/book.asp?BKID=15580

AI/ChatGPT v.智慧財產權──美國生成式AI案例評析-元照出版
2023-12-27
kiang 01:02:06
https://g0v-tw.slack.com/archives/C067VNP6P88/p1703609331048629
kiang 01:02:06
https://g0v-tw.slack.com/archives/C067VNP6P88/p1703609331048629
kiang
民視有透過 ChatGPT 去摘要三個候選人的發言內容
https://www.facebook.com/FTVNews53/posts/pfbid0F6pp3sTZuQKv8Cz9B3bKhJgpQvfUigZQAc2x5qdghacTQBAeKMLf9Xb6bSGp4zCyl
singing 14:05:52
@singing has joined the channel
2023-12-30
mengting 00:41:43
@mengting has joined the channel