關於集團
集團品牌
presslogic-logo
廣告查詢
工作機會
用戶登入
用戶名稱:
密      碼:
搜索
教育王國 討論區 教育講場 【視頻】就業前景未見樂觀 年輕畢業生求職面對甚麼困難 ...
樓主: elbar
go

【視頻】就業前景未見樂觀 年輕畢業生求職面對甚麼困難? [複製鏈接]

Rank: 9Rank: 9Rank: 9


23988
41#
發表於 6 天前 |只看該作者
其實呢度大家問嘅問題, 都係非常到位, 非常深, 我都只係基於自己所知, 嘅盡量分享, 啱定錯, 我自己都唔敢保證... 哈哈

Rank: 5Rank: 5


3512
42#
發表於 6 天前 |只看該作者
twhkleung 發表於 26-4-21 10:22
1. context 唔夠大, ,會係其中一個原因, 但係乜嘢叫做"夠大", 幾多先算"夠大"?2. 正如人腦, 其實個contact  ...

所以有時會將的資料用R A G Model整理好,然後用LL M Model配合。限制佢的回答只可以根據LA G的資料來答覆唔准佢作嘢。

點評

twhkleung    發表於 6 天前

Rank: 14Rank: 14Rank: 14Rank: 14


121498
43#
發表於 6 天前 |只看該作者
poonseelai 發表於 26-4-21 10:01
係小學生已經用AI出貓

我都唔識俾反應。
God grant me the serenity
to accept the things I cannot change;
courage to change the things I can;
and wisdom to know the difference.

Rank: 14Rank: 14Rank: 14Rank: 14


121498
44#
發表於 6 天前 |只看該作者
barque 發表於 26-4-21 10:27
所以有時會將的資料用R A G Model整理好,然後用LL M Model配合。限制佢的回答只可以根據LA G的資料來答覆 ...

學到嘢!
God grant me the serenity
to accept the things I cannot change;
courage to change the things I can;
and wisdom to know the difference.

Rank: 14Rank: 14Rank: 14Rank: 14


121498
45#
發表於 6 天前 |只看該作者
twhkleung 發表於 26-4-21 10:26
其實呢度大家問嘅問題, 都係非常到位, 非常深, 我都只係基於自己所知, 嘅盡量分享, 啱定錯, 我自己都唔敢保 ...

大家交流一下,你講嘅嘢,有啲我都要查字典/較可靠嘅網站。

這樣討論會實事求事啲,也清楚pros and cons, limitations and biases.
God grant me the serenity
to accept the things I cannot change;
courage to change the things I can;
and wisdom to know the difference.

Rank: 14Rank: 14Rank: 14Rank: 14


121498
46#
發表於 6 天前 |只看該作者
twhkleung 發表於 26-4-21 10:23
其實而家deploy咗嘅ai model, 同research嘅理論模型, 差別其實好大. AI Research Scientist 同 ML Engineer ...

我哋條team 入面有data scientists and ML engineers. 我講嘅research係clinical research,不是Al research。

我哋係比較上係似vertical model,特別係diagnostic device 相連的嘅演算法,係要有特定要求,不然就garbage in garbage out 。

不同Financial models 有不同的應用,可能同一堆data,用不同 models 會有不同結果,點解用邊個係要有justification,不同models 也可以睇到另一面的上下限及中位數值。buyer and seller 都有不同期望。

他們用multiple models 係某些complex 基金prospectus 入面講咗作價以某幾個model 做 reference,然後選中位數值,或者whatever highest or lowest。

點評

twhkleung    發表於 5 天前
God grant me the serenity
to accept the things I cannot change;
courage to change the things I can;
and wisdom to know the difference.

Rank: 9Rank: 9Rank: 9


23943
47#
發表於 6 天前 |只看該作者

回覆樓主:

本帖最後由 poonseelai 於 26-4-21 15:30 編輯

好深呀!要慢慢學下

今日聽到Apple 9月換CEO, 未夠50歳, 是一位hardware engineer, 在Apple由低做起

早一兩年Apple被批評在AI方面大落後, 而家不少AI公司燒錢燒不停, Apple亦叫停了電車項目, 專注本業, 誰才是正確?

Rank: 5Rank: 5


3512
48#
發表於 6 天前 |只看該作者
twhkleung 發表於 26-4-21 10:23
其實而家deploy咗嘅ai model, 同research嘅理論模型, 差別其實好大. AI Research Scientist 同 ML Engineer ...

女兒近期對Constitutional AI 好感興趣,近期自己都有做啲Research project去評估到底佢哋的準確性

點評

twhkleung    發表於 5 天前

Rank: 9Rank: 9Rank: 9


23988
49#
發表於 5 天前 |只看該作者
本帖最後由 twhkleung 於 26-4-22 03:01 編輯
barque 發表於 26-4-21 17:43
女兒近期對Constitutional AI 好感興趣,近期自己都有做啲Research project去評估到底佢哋的準確性
...
呢個係暫時對管理AI Alignment 嘅大方向 (保證AI Model同人類Intent 保持一致), 而Anthropic 可以話係目前做的最好嘅, 原因:
1. Jan Leike:
- 在離開 OpenAI 後,他加入了 Anthropic 並領導對Alignment Science 嘅R&D 團隊。
- 他現在是Constitutional AI 的核心領導之一。
2. Jan Leike 和 Ilya Sutskever (Open AI Cofounder)一起, 在2023年反對Sam Altman, 只看錢而不顧AI與人類的倫理.
3. Ilya Sutskever 是Jeffrey Hinton (AI God Father)的PhD 學生, 他們共同成立的DNNresearch, 在2013年畀Google收購, 是近代AI 跳躍性發展的基礎力量(所以點解Jeffrey攞咗諾貝爾獎). Ilya 也是負責當年train AlphaGo的人.

AI 發展嘅history, 其實幾好睇, 好似睇小說咁.




Rank: 9Rank: 9Rank: 9


23988
50#
發表於 5 天前 |只看該作者
poonseelai 發表於 26-4-21 15:27
好深呀!要慢慢學下

今日聽到Apple 9月換CEO, 未夠50歳, 是一位hardware engineer, 在Apple由低做起
Apple Top management的管理質量真係好高, 佢哋可能係最後的大贏家, 負責收割ai的成果.

Rank: 9Rank: 9Rank: 9


23988
51#
發表於 5 天前 |只看該作者
本帖最後由 twhkleung 於 26-4-22 11:20 編輯
ANChan59 發表於 26-4-21 13:02
大家交流一下,你講嘅嘢,有啲我都要查字典/較可靠嘅網站。

這樣討論會實事求事啲,也清楚pros and cons ...

Vertical AI唔易搞, 特別係醫療級, IBM 花咗好多好多錢搞咗十幾年, 最後得個吉. 你條team好堅!

我自己覺得難搞原因有以下.

1. Anthropic, OpenAI 呢類 general-purpose AI model, 就算錯, 都唔會有咩後果, 唔好講我哋啲普通user/consumer, 就算賣畀enterprise用, 相信啲terms, 都保護到佢哋唔會有任何責任.
2. 但用喺醫療方面嘅AI就唔同, 出咗事, 唔係三幾間商業公司對佢嘅投訴, 而係整個社會同政府, 可能都會同時"移動"佢.
3. 而現時嘅ai model, 唔適合醫療級, 最主要原因, 就係佢響uncertainty方面嘅處理方法,
i. 唔識得講I don't know.
ii. 之餘仲要講到似層層.
4. Bayesian Inference Model 暫時係醫療對AI 研發嘅基礎framework. 佢主要係以下5大方面嘅希望.
i. Uncertainty  - 雖然佢都係唔識得講I dont know. 但係佢嘅答案, 會係類似咁樣,
- 可能性一, 乜乜乜, 50%,
- 可能性二, 乜乜乜, 30%,
- 可能性三, 乜乜乜, 15%,
- 可能乘四, 乜乜...
ii. Sparse Data (同small dataset唔一樣), 當數據唔足夠嘅時候, 仍然能夠相對客觀的比出分析.
iii. Interpretability比較高, 即係佢嘅blackbox冇咁black.
iv. 更重要嘅係, 佢嘅思考邏輯, 會更接近人類腦袋的思考方法.
v. 唯一可惜嘅係, 呢種ai model, 同現時嘅hardware, 例如, GPU/TPU, 係唔配合. 所以, 追求短期研究成果嘅大學團隊, 都好少研究向呢方面做努力, 通常, 只有大U (都係噏得出嗰幾間), 先有資源, 做呢方面嘅規模性/系統研究.









Rank: 14Rank: 14Rank: 14Rank: 14


121498
52#
發表於 5 天前 |只看該作者
twhkleung 發表於 26-4-22 01:16
Vertical AI唔易搞, 特別係醫療級, IBM 花咗好多好多錢搞咗十幾年, 最後得個吉. 你條team好堅!

我 ...

在醫療Al開發上我只係略懂,不會裝懂,所以過往十年都係好少講太多,我哋個CTO好勁。我嘅層面較多講策略,應用,business model,風險管理,跟我行業經驗緊扣。

我哋資源有限之下,research strategy幾集中,先硬件,軟件再去Al,都係集中一種病。我哋有market research,好清楚個pain points,有少少disruptive technology 的情況。

以往這種檢查一定要住院,用多種設備檢測,全身好似蜘蛛網包圍。配戴上我哋嘅設備,現在可以係屋企collect data,然後upload 上雲端,就可以出報告給醫生作診斷。快靚正又平,在美國攞埋insurance code,可以claim medical insurance。

IBM這種大企業,一做就會做一個好大嘅scope,太大好難管理R&D,加上醫療係好保守嘅行業,個個以為自己好醒,好多性格巨星,點解我成日講multidisciplinary 嘅重要性。

就算蘋果有Apple watch,已經係wearable device,功能上已經係醫療設備,但係未能過到FDA就不可以claim,硬件容易,軟件難,Al更難。

如果過到FDA批准上市,所有嘢都要監管,包括設計,生產,採購,Q MS,post market surveillance ,adverse event reporting ,recall。成個product life cycle 都係要小心管理。蘋果first tier supplier鴻海究想入醫療設備行業,一闊三大,都重係小打小鬧。

其實好多小公司做MD,然後被大企業M&A,係一個常見嘅exit strategy。

******

補充一點,好多人以為醫院用嘅設備最精密,最準確,要求最高。事實上高端家用MD比較醫院用嘅更高要求。高端係相對於血糖計,血壓計,這些常見的家用MD。

MD追求第一安全性,第二才是療效。一個layman可以操作,不是一個醫護人員操作,在安全考慮已經好麻煩及複雜,好多風險及誤用係大家想都想唔到。

點評

twhkleung    發表於 5 天前
God grant me the serenity
to accept the things I cannot change;
courage to change the things I can;
and wisdom to know the difference.

Rank: 7Rank: 7Rank: 7


13811
53#
發表於 5 天前 |只看該作者
本帖最後由 akys 於 26-4-22 14:04 編輯
ANChan59 發表於 26-4-20 19:24
你舊同學個仔會唔會志不在香港!?

爸爸有錢,唔憂柴米。 英國讀 degree。 太辛苦唔想留低畢業後返咗香港做嘢。

Rank: 14Rank: 14Rank: 14Rank: 14


121498
54#
發表於 4 天前 |只看該作者
akys 發表於 26-4-22 11:57
爸爸有錢,唔憂柴米。 英國讀 degree。 太辛苦唔想留低畢業後返咗香港做嘢。
...

唔憂柴米,讀唔讀master 都冇所謂。

有點奇怪怕英國辛苦,有個master,心頭更高,畢業返香港都係好大機會失業。

找不到工作不外乎幾個原因:
1。經濟及就業市場環境
2。本人的性格及意圖心
3。家庭的培養及價值觀

God grant me the serenity
to accept the things I cannot change;
courage to change the things I can;
and wisdom to know the difference.
‹ 上一主題|下一主題
返回列表
發新帖 回覆
你需要登錄後才可以回帖 登錄 | 註冊