SG Talk

S’pore not likely to use AI in sentencing in foreseeable future: Chief Justice ..

https://www.singaporelawwatch.sg/Headlin...ef-Justice

Thank you for the news article

i did a research on using AI but to triage medical texts.

People said AI can help lighten the load of the lawyers or even made them redundant but it's wishful thinking.

The lawyers will never ever let computers take away any of their work, so that they will forever be relevant to society.

(01-11-2022, 06:54 PM)Blasterlord2 Wrote: [ -> ]People said AI can help lighten the load of the lawyers or even made them redundant but it's wishful thinking.

The lawyers will never ever let computers take away any of their work, so that they will forever be relevant to society.

haha because Lawyers , no offence but isn't so?

The Skill of " flip prata " , can turn white to black and black to white. Or else why people hire you fight their case right?

it's straightforward to train a decision tree to do the job, but too many exceptions which are difficult to handle, like medallists, low iq, mental and depression cases, those with a bright future and had contributed to society and etc. so it's better to leave it as is

(01-11-2022, 07:26 PM)WhatDoYouThink! Wrote: [ -> ]it's straightforward to train a decision tree to do the job, but too many exceptions which are difficult to handle, like medallists, low iq, mental and depression cases, those with a bright future and had contributed to society and etc. so it's better to leave it as is

not with decision trees. you have your use NLP and identify all the entities first (see below) you will need to vectorise the entities, and discover the relationship among the entities, for that  i use association rules mining. i prefer to use neural network to do the classification but if you triage/classify by history or would you want to refer to latest criteria to decide the latest classification. so, i did a comprehensive list of medical consideration critieria, then perform cosine similarity to find the nearest match. but this can be improved if we can build a model of the convict where we place both vectorised legal entities of defence and accuse. Sorry, didnt invest much time to think about this unless I get a commission to do so. I did want to do it for AU's migration law where AAT's appeal cases are available on AustLii website.

(01-11-2022, 08:20 PM)fishbuff Wrote: [ -> ]not with decision trees. you have your use NLP and identify all the entities first (see below) you will need to vectorise the entities, and discover the relationship among the entities, for that  i use association rules mining. i prefer to use neural network to do the classification but if you triage/classify by history or would you want to refer to latest criteria to decide the latest classification. so, i did a comprehensive list of medical consideration critieria, then perform cosine similarity to find the nearest match. but this can be improved if we can build a model of the convict where we place both vectorised legal entities of defence and accuse. Sorry, didnt invest much time to think about this unless I get a commission to do so. I did want to do it for AU's migration law where AAT's appeal cases are available on AustLii website.

medical diagnosis much easier because there aren't too many abovementioned exceptions to take into consideration

(01-11-2022, 08:34 PM)WhatDoYouThink! Wrote: [ -> ]medical diagnosis much easier because there aren't too many abovementioned exceptions to take into consideration

lol. medical is easier?. havent look at all the different medical cases, each with patients' demographics, medications, dosage, conditions, symptom-treatment, etc.. but I wont stay that it is final coz even my paper on medical triaging was published with a tier 1 medical AI journal, yet there are so much I need to do in order to spruce it further, especially building a patent medical journey model.

i want to try on legal NLP though. but no such opportunity arise yet

(01-11-2022, 09:03 PM)fishbuff Wrote: [ -> ]lol. medical is easier?. havent look at all the different medical cases, each with patients' demographics, medications, dosage, conditions, symptom-treatment, etc.. but I wont stay that it is final coz even my paper on medical triaging was published with a tier 1 medical AI journal, yet there are so much I need to do in order to spruce it further, especially building a patent medical journey model.

i want to try on legal NLP though. but no such opportunity arise yet

Singapore now wants AI talents. You can come back here.

(01-11-2022, 09:07 PM)Blasterlord2 Wrote: [ -> ]Singapore now wants AI talents. You can come back here.

no thanks. 
AU govt paid for my phd tuition fees, all $120k. PAP did nothing for me... 

I will continue to support AU's organisations.

machine learning and deep learning are already widely used for disease diagnosis, treatments, drug discovery, patient risk identification and many more.

if you hv written any journal paper, shd hv come across some of them

(01-11-2022, 09:11 PM)WhatDoYouThink! Wrote: [ -> ]machine learning and deep learning are already widely used for disease diagnosis, treatments, drug discovery, patient risk identification and many more.

if you hv written any journal paper, shd hv come across some of them

Duh.. I got a phd in AI

(01-11-2022, 09:29 PM)fishbuff Wrote: [ -> ]Duh.. I got a phd in AI

hv you developed, tested, developed any working system in use, or your ideas remain on paper?

(01-11-2022, 09:47 PM)WhatDoYouThink! Wrote: [ -> ]hv you developed, tested, developed any working system in use, or your ideas remain on paper?

No need to argue.. state your qualification first.  Alof of unqualified people out there that think studying some bootcamps can automatically be called data scientist" without even publish any papers at high tiered journal. Pushing some compiled model to flask on kubenetes doesn't qualify you to be a real scientist at all.

I'm working as a senior data scientist with the private sector working with reinforcement learning applications for simulation modelling optimization in Australia

but i am happy for you that you are in this line. we need more sg people to be in DS field

(02-11-2022, 05:18 AM)fishbuff Wrote: [ -> ]No need to argue.. state your qualification first.  Alof of unqualified people out there that think studying some bootcamps can automatically be called data scientist" without even publish any papers at high tiered journal. Pushing some compiled model to flask on kubenetes doesn't qualify you to be a real scientist at all.

I'm working as a senior data scientist with the private sector working with reinforcement learning applications for simulation modelling optimization in Australia

but i am happy for you that you are in this line. we need more sg people to be in DS field

nice to hear about what you're doing, really 为国争光

not lah, I'm not qualified to argue at all. you must be new dont know here got all sorts of talent, experts, gurus, masters, professors, maybe a few AI professors and many more. Im humbly learning from them all along

so other than doing some simulations and publishing some paper ideas, you haven't developed any functional AI systems as yet?

(02-11-2022, 07:43 AM)WhatDoYouThink! Wrote: [ -> ]nice to hear about what you're doing, really 为国争光

not lah, I'm not qualified to argue at all. you must be new dont know here got all sorts of talent, experts, gurus, masters, professors, maybe a few AI professors and many more. Im humbly learning from them all along

so other than doing some simulations and publishing some paper ideas, you haven't developed any functional AI systems as yet?

walao.. quite siong liao. I studied part-time phd for nearly 7 years and work full-time as a data engineer at my current age of 55. managed to pull and published nearly 14 publications as 1st author. my hair all white now. how many sinkie ah lao can do that? i busted my balls to pull through this. i should retire by now but i won't give up as I didnt get the chance to do this in SG so I will push the envelope here as far as I can.

The discrete event simulation is not done by me but by the US team, it is integrating with dynamic system modelling which emulate the electric mining machineries. my TL is developing the wrapper and I take care of the RL's heuristic optimization

have you heard of alpha-go zero?

(01-11-2022, 06:22 PM)Bigiron Wrote: [ -> ]S’pore not likely to use AI in sentencing in foreseeable future: Chief Justice ..
https://www.singaporelawwatch.sg/Headlin...ef-Justice

So now CJ scare his job kenna replaced by AI...

(02-11-2022, 09:04 AM)fishbuff Wrote: [ -> ]walao.. quite siong liao. I studied part-time phd for nearly 7 years and work full-time as a data engineer at my current age of 55. managed to pull and published nearly 14 publications as 1st author. my hair all white now. how many sinkie ah lao can do that? i busted my balls to pull through this. i should retire by now but i won't give up as I didnt get the chance to do this in SG so I will push the envelope here as far as I can.

The discrete event simulation is not done by me but by the US team, it is integrating with dynamic system modelling which emulate the electric mining machineries. my TL is developing the wrapper and I take care of the RL's heuristic optimization

have you heard of alpha-go zero?

walao... at 48 still do 7 yrs of phd? that's amazingly courageous. most ppl just tangping and bailan liao at this age already

I only heard of covid zero only, duno simi alpha go zero, very unlettered, illiterate and uninitiated. but still hv a full head of thick and fluffy dark hair. guess you duno how to take care of yr head?

(02-11-2022, 09:26 AM)WhatDoYouThink! Wrote: [ -> ]walao... at 48 still do 7 yrs of phd? that's amazingly courageous. most ppl just tangping and bailan liao at this age already

I only heard of covid zero only, duno simi alpha go zero, very unlettered, illiterate and uninitiated. but still hv a full head of thick and fluffy dark hair. guess you duno how to take care of yr head?

of course i still have hair but mostly white/grey now.  most of my peers are bald around 55 or have receding hair line. I'm happy i still got white hair.

alphago-zero


alphastar

talking about journal papers, every yrs there are tons of them being published by china, usa and many others. mostly are just toying with ideas, but ofc there are aso useful ones which are patented and even spinned off as new start-ups.

in these regards australia is quite a small player, lagging behind s korea and even taiwan province

depend on many factors, first is the ranking of the conference or journal, then narrow down to the field of interest
http://www.conferenceranks.com/ or
https://www.scimagojr.com/journalrank.php
papers like KDD, ICML, are of very high quality and they are cited by many researchers, both academic and commercial.

some research will not surface until there is awareness. e.g. Prof Hinton's paper on neural network was published in the 70s and it remained in the academic world until the late 80s or 90s when there is an explosion of AI needs. There are many, many papers out there. I'm not sure if you work on time series but there had been a lot of research since the 18th century. each research is built on top of other research. The saying goes; we stand on the shoulders of giants.
https://en.wikipedia.org/wiki/Standing_o..._of_giants

not going to dick comparison in term of countries when it comes to publication output but... but for a country of 24million, it isn't too shabby.