Generative AI ChatGPT នឹង​មាន​នៅ​គ្រប់​ទីកន្លែង​នៅ​ពេល​ដែល​វិបផតថល API ត្រូវ​បាន​បើក​ក្នុង​ពេល​ឆាប់​ៗ​នេះ ធ្វើ​ឱ្យ​ខូច​ក្រមសីលធម៌ AI និង​ច្បាប់ AI

Release the Kraken!

You are undoubtedly familiar with that famous catchphrase as especially uttered by actor Liam Neeson in ការប៉ះទង្គិចនៃទីតានិច when he commands that the legendary sea monster be released, aiming to wreak immense havoc and outsized destruction. The line has been endlessly repeated and spawned all manner of memes. Despite the various parodies, most people still at least viscerally sense that the remark foretells something of a shadowy and dangerous emergence is about to be unleashed.

Perhaps the same sentiment can be applied these days to Artificial Intelligence (AI).

អនុញ្ញាតឱ្យខ្ញុំបកស្រាយ។

A recent announcement indicated that a now resoundingly famous AI app called ChatGPT made by the organization OpenAI is soon going to be made available for access by other programs. This is big news. I say this even though little of the regular media has picked up on the pronouncement. Other than fleeting mentions, the full impact of this upcoming access is going to be pretty darned significant.

In today’s column, I’ll be explaining why this is the case. You can prepare yourself accordingly.

Some adamantly believe that this will be akin to letting loose the Kraken, namely that all kinds of bad things are going to arise. Others see this as making available a crucial resource that can boost tons of other apps by leveraging the grand capabilities of ChatGPT. It is either the worst of times or the best of times. We will herein consider both sides of the debate and you can decide for yourself which camp you land in.

សរុបមកនេះ មានការពិចារណាអំពីក្រមសីលធម៌ AI និងច្បាប់ AI ។

សូម​ដឹង​ថា​មាន​កិច្ច​ខិត​ខំ​ប្រឹង​ប្រែង​ជា​បន្ត​ដើម្បី​ដាក់​បញ្ចូល​គោលការណ៍ Ethical AI ក្នុង​ការ​អភិវឌ្ឍ និង​ការ​ដាក់​បញ្ចូល​កម្មវិធី AI។ ការកើនឡើងនៃការព្រួយបារម្ភ និងខណៈពេលដែលអ្នកក្រមសីលធម៌ AI កំពុងព្យាយាមធានាថា កិច្ចខិតខំប្រឹងប្រែងដើម្បីបង្កើត និងអនុម័ត AI គិតគូរពីទិដ្ឋភាពនៃការធ្វើ អាយអេសដើម្បីភាពល្អ និងការបដិសេធ អាយអេសសម្រាប់អាក្រក់. ដូចគ្នានេះដែរ មានច្បាប់ AI ថ្មីដែលត្រូវបានស្នើឡើង ដែលកំពុងត្រូវបានរុំព័ទ្ធជាដំណោះស្រាយដ៏មានសក្ដានុពល ដើម្បីរក្សាការខិតខំប្រឹងប្រែងរបស់ AI មិនឱ្យប៉ះពាល់ដល់សិទ្ធិមនុស្ស និងផ្សេងៗទៀត។ សម្រាប់ការគ្របដណ្តប់ដែលកំពុងបន្ត និងទូលំទូលាយនៃក្រមសីលធម៌ AI និងច្បាប់ AI សូមមើល តំណភ្ជាប់នៅទីនេះ និង តំណភ្ជាប់នៅទីនេះ, គ្រាន់តែឈ្មោះមួយចំនួន។

There have been growing qualms that ChatGPT and other similar AI apps have an ugly underbelly that maybe we aren’t ready to handle. For example, you might have heard that students in schools are potentially able to cheat when it comes to writing assigned essays via using ChatGPT. The AI does all the writing for them. Meanwhile, the student is able to seemingly scot-free turn in the essay as though they did the writing from their own noggin. Not what we presumably want AI to do for humankind.

A few key essentials might be helpful to set the stage for what this is all about.

ChatGPT is a type of AI commonly referred to as AI បង្កើត. These trending generative-based AI apps allow you to enter a brief prompt and have the app generate outputs for you. In the case of ChatGPT, the output is text. Thus, you enter a text prompt and the ChatGPT app produces text for you. I tend to describe this as a particular subtype of Generative AI that is honed to generate text-to-essay outputs (there are other subtypes such as text-to-images, text-to-video, and so on).

The AI maker of ChatGPT has indicated that soon an API (Application Programming Interface) will be made available for the AI app. In short, an API is a means of allowing other programs to go ahead and use a program that makes available a portal into the given application. This means that just about any other program on this planet can potentially leverage the use of ChatGPT (well, as licensed and upon approval by the AI maker of ChatGPT, as will be further discussed momentarily herein).

The upshot is that the use and uses of ChatGPT could potentially shoot through the roof.

Whereas today there is an impressive number of signups entailing people who on an បុគ្គល មូលដ្ឋាន can use ChatGPT, capped by the AI maker at a million users, this is actually going to likely be a drop in the bucket of what is about to come.

Realize that those existing million signups consist of some portion that used ChatGPT on a one-time frolic and then after the thrill wore off, they haven’t used it since. Many were presumably attracted to the AI app as a social media viral reactive response. In short, if everyone else was wanting to use it, they wanted to do so too. Upon some initial experimentation with the generative-based AI, they felt satisfied that they had averted their FOMO (fear of missing out).

To make it stridently clear, I am not suggesting that people aren’t using ChatGPT. They are. Those that signed up are increasingly finding that the AI app is overloaded. Lots and lots of people are using the app. You get a few cleverly composed sorrowful indications from time to time that the system is busy and you should try back later on. Word on the street is that the existing infrastructure for ChatGPT has been straining to cope with the avid fans using the AI app.

And though having a million potential users is nothing to sneeze at, the number is likely going to readily be eclipsed multifold once the API is made available. Developers of other programs that today have nothing to do with generative AI are going to want to tap into the generative AI bandwagon. They are going to want to connect their program with ChatGPT. Their heart of hearts hopes is that this will boost their existing program into the stratosphere of popularity.

Think of it this way. Assume that all manner of software companies that make programs today that reach many millions of users, often reaching into the tens and hundreds of millions of users all told, opt to pair up their respective programs with ChatGPT. This suggests that the volume of users that are using ChatGPT could go sky-high.

The Kraken is released.

Why would various software companies want to pair up with ChatGPT, you might be wondering?

A straightforward answer is that they might as well exploit the amazing tailwinds that are pushing ChatGPT onward and upward. Some will do so for sensible and aboveboard reasons. Others will do so merely to try and gain their own semblance of fifteen minutes of fame.

I like to stratify the pairings to ChatGPT as consisting of two major intentions:

• Genuine Pairing With ChatGPT
• Fakery Pairing With ChatGPT

In the first case, the notion is that there is a bona fide basis for pairing up with ChatGPT. The makers of a given program are able to well articulate the tangible and functional benefits that will arise due to a pairing of their program with ChatGPT. We can all in a reasonable frame of mind see that the pairing is a match made in heaven.

For the other case, consisting of what I denote as fakery, some will seek to pair up with ChatGPT on a flighty or shaky basis. The business case does not consist of anything especially substantive. The pairing is a desperate attempt to ride the tails of ChatGPT. Any reasonable inspection would reveal that the pairing is marginal in value. Now then, whether you think that this is a proper or improper form of pairing is somewhat hanging in the air. One could try to argue that a particular pairing with ChatGPT, even if the pairing doesn’t accomplish anything other than boost usage and has no other functional additive value, presumably is pairing still worthy of undertaking.

A bit of a downside will be those that falsely portray the pairing and lead people to believe that something notable is occurring when it really is not. We can certainly expect that some will try this. Those in AI Ethics are stridently concerned about snake oil uses that are going to come out of the woodwork. There is a chance too that if this gets out of hand, we might see new AI-related laws that will be spurred into being drafted and enacted.

Let’s take a closer exploration of what constitutes genuine pairings and what also constitutes fakery pairings.

First, we ought to make sure that we are all on the same page about what Generative AI consists of and also what ChatGPT is all about. Once we cover that foundational facet, we can perform a cogent assessment of how the API into ChatGPT is going to radically change things.

ប្រសិនបើអ្នកធ្លាប់ស្គាល់ច្រើនជាមួយ Generative AI និង ChatGPT រួចហើយ អ្នកប្រហែលជាអាចរំលងផ្នែកបន្ទាប់ ហើយបន្តជាមួយផ្នែកដែលធ្វើតាមវា។ ខ្ញុំជឿថាអ្នកផ្សេងទៀតនឹងរកឃើញការណែនាំអំពីព័ត៌មានលម្អិតសំខាន់ៗអំពីបញ្ហាទាំងនេះដោយអានផ្នែកនេះឱ្យបានដិតដល់ និងបង្កើនល្បឿន។

បឋមរហ័សអំពី Generative AI និង ChatGPT

ChatGPT គឺជាប្រព័ន្ធតម្រង់ទិសសន្ទនាអន្តរកម្ម AI គោលបំណងសំខាន់ជា chatbot ទូទៅដែលហាក់ដូចជាគ្មានកំហុស ទោះបីជាយ៉ាងណាក៏ដោយ វាត្រូវបានប្រើប្រាស់យ៉ាងសកម្ម និងដោយមនុស្សយ៉ាងសកម្មក្នុងវិធីដែលទាក់ទាញមនុស្សជាច្រើនដែលមិនប្រុងប្រយ័ត្ន ដូចដែលខ្ញុំនឹងរៀបរាប់លម្អិតក្នុងពេលឆាប់ៗនេះ។ កម្មវិធី AI នេះប្រើប្រាស់បច្ចេកទេស និងបច្ចេកវិទ្យានៅក្នុងអាណាចក្រ AI ដែលជារឿយៗត្រូវបានគេហៅថា AI បង្កើត. AI បង្កើតលទ្ធផលដូចជាអត្ថបទ ដែលជាអ្វីដែល ChatGPT ធ្វើ។ កម្មវិធី AI ដែលមានមូលដ្ឋានលើជំនាន់ផ្សេងទៀតបង្កើតរូបភាពដូចជារូបភាព ឬស្នាដៃសិល្បៈ ខណៈដែលកម្មវិធីផ្សេងទៀតបង្កើតឯកសារអូឌីយ៉ូ ឬវីដេអូ។

ខ្ញុំនឹងផ្តោតលើកម្មវិធី AI បង្កើតមូលដ្ឋានលើអត្ថបទនៅក្នុងការពិភាក្សានេះ ចាប់តាំងពីនោះជាអ្វីដែល ChatGPT ធ្វើ។

កម្មវិធី AI ជំនាន់ថ្មីគឺងាយស្រួលប្រើណាស់។

អ្វីដែលអ្នកត្រូវធ្វើគឺបញ្ចូលប្រអប់បញ្ចូល ហើយកម្មវិធី AI នឹងបង្កើតអត្ថបទមួយសម្រាប់អ្នកដែលព្យាយាមឆ្លើយតបទៅនឹងការជម្រុញរបស់អ្នក។ អត្ថបទ​ដែល​តែង​នឹង​ហាក់​ដូច​ជា​អត្ថបទ​ត្រូវ​បាន​សរសេរ​ដោយ​ដៃ​និង​ចិត្ត​របស់​មនុស្ស។ ប្រសិនបើអ្នកត្រូវបញ្ចូលប្រអប់បញ្ចូលដែលនិយាយថា "ប្រាប់ខ្ញុំអំពី Abraham Lincoln" AI ជំនាន់ថ្មីនឹងផ្តល់ឱ្យអ្នកនូវអត្ថបទអំពី Lincoln ។ នេះត្រូវបានចាត់ថ្នាក់ជាទូទៅថាជា AI ជំនាន់ដែលដំណើរការ អត្ថបទទៅអត្ថបទ ឬអ្នកខ្លះចូលចិត្តហៅវា។ អត្ថបទទៅអត្ថបទ ទិន្នផល។ ដូចដែលបានរៀបរាប់រួច មានរបៀបបង្កើត AI ផ្សេងទៀត ដូចជាអត្ថបទទៅសិល្បៈ និងអត្ថបទទៅវីដេអូ។

គំនិតដំបូងរបស់អ្នកប្រហែលជាថាសមត្ថភាពបង្កើតនេះហាក់ដូចជាមិនមែនជារឿងធំបែបនេះទេ ទាក់ទងនឹងការផលិតអត្ថបទ។ អ្នកអាចធ្វើការស្វែងរកតាមអ៊ិនធរណេតបានយ៉ាងងាយស្រួល ហើយស្វែងរកអត្ថបទជាច្រើនអំពីប្រធានាធិបតី Lincoln យ៉ាងងាយស្រួល។ ចំនុចសំខាន់នៅក្នុងករណីនៃ AI ជំនាន់នេះគឺថាអត្ថបទដែលបានបង្កើតគឺមានលក្ខណៈប្លែកពីគេ ហើយផ្តល់នូវសមាសភាពដើមជាជាង copycat ។ ប្រសិនបើអ្នកព្យាយាមស្វែងរកអត្ថបទដែលផលិតដោយ AI តាមអ៊ីនធឺណិតនៅកន្លែងណាមួយ អ្នកទំនងជាមិនអាចរកឃើញវាបានទេ។

Generative AI ត្រូវបានបណ្តុះបណ្តាលជាមុន និងប្រើប្រាស់រូបមន្តគណិតវិទ្យា និងការគណនាដ៏ស្មុគស្មាញមួយដែលត្រូវបានបង្កើតឡើងដោយការពិនិត្យមើលគំរូនៅក្នុងពាក្យ និងរឿងដែលបានសរសេរនៅលើបណ្តាញ។ ជាលទ្ធផលនៃការពិនិត្យមើលអត្ថបទដែលបានសរសេររាប់ពាន់លាន AI អាចបញ្ចេញអត្ថបទ និងរឿងថ្មីៗ ដែលជាការខុសឆ្គងនៃអ្វីដែលបានរកឃើញ។ ដោយការបន្ថែមមុខងារប្រូបាប៊ីលីតេផ្សេងៗ អត្ថបទលទ្ធផលគឺមានលក្ខណៈប្លែកពីគេក្នុងការប្រៀបធៀបទៅនឹងអ្វីដែលបានប្រើនៅក្នុងឈុតបណ្តុះបណ្តាល។

ដូច្នេះហើយ​ទើប​មាន​ការ​ចលាចល​អំពី​សិស្ស​អាច​បន្លំ​ពេល​សរសេរ​អត្ថបទ​នៅ​ក្រៅ​ថ្នាក់។ គ្រូម្នាក់មិនអាចគ្រាន់តែយកអត្ថបទដែលសិស្សបោកបញ្ឆោតអះអាងថាជាការសរសេររបស់ពួកគេផ្ទាល់ ហើយស្វែងរកថាតើវាត្រូវបានចម្លងពីប្រភពអ៊ីនធឺណិតផ្សេងទៀតដែរឬទេ។ សរុបមក វា​នឹង​មិន​មាន​អត្ថបទ​ដែល​មាន​ស្រាប់​តាម​អ៊ីនធឺណិត​ដែល​ត្រូវ​នឹង​អត្ថបទ​ដែល​បង្កើត​ដោយ AI ទេ។ អ្វី​ដែល​បាន​ប្រាប់ គ្រូ​នឹង​ត្រូវ​ទទួល​យក​ដោយ​ក្រៀមក្រំ​ថា សិស្ស​សរសេរ​អត្ថបទ​នោះ​ជា​ស្នាដៃ​ដើម។

មានការព្រួយបារម្ភបន្ថែមអំពី AI ជំនាន់។

គុណវិបត្តិដ៏សំខាន់មួយគឺថា អត្ថបទដែលផលិតដោយកម្មវិធី AI ដែលមានមូលដ្ឋានលើជំនាន់អាចមានភាពមិនពិតជាច្រើនដែលបានបង្កប់ រួមទាំងការពិតដែលមិនពិតតាមប៉ាតង់ ការពិតដែលបង្ហាញឱ្យយល់ខុស និងការពិតជាក់ស្តែងដែលត្រូវបានប្រឌិតទាំងស្រុង។ ទិដ្ឋភាព​ប្រឌិត​ទាំងនោះ​ត្រូវ​បាន​គេ​សំដៅ​ជា​ញឹកញាប់​ថា​ជា​ទម្រង់​នៃ​ AI ភាពច្របូកច្របល់ជាឃ្លាដែលខ្ញុំមិនពេញចិត្ត ប៉ុន្តែគួរឱ្យសោកស្ដាយ ហាក់ដូចជាកំពុងទទួលបានប្រជាប្រិយភាពយ៉ាងណាក៏ដោយ (សម្រាប់ការពន្យល់លម្អិតរបស់ខ្ញុំអំពីមូលហេតុដែលវាក្យសព្ទនេះអាក្រក់ និងមិនសមរម្យ សូមមើលការគ្របដណ្តប់របស់ខ្ញុំនៅ តំណភ្ជាប់នៅទីនេះ).

ខ្ញុំ​ចង់​បញ្ជាក់​ពី​ទិដ្ឋភាព​សំខាន់​មួយ​មុន​ពេល​យើង​ចូល​ទៅ​ក្នុង​រឿង​ដ៏ក្រាស់​លើ​ប្រធានបទ​នេះ។

មាន​ការ​អះអាង​លើស​ពី​ zany មួយ​ចំនួន​នៅ​លើ​ប្រព័ន្ធ​ផ្សព្វផ្សាយ​សង្គម​អំពី AI បង្កើត ការ​អះអាង​ថា​កំណែ​ចុងក្រោយ​របស់ AI នេះ​គឺ​ជា​ការ​ពិត AI អារម្មណ៍ (អត់ទេ គេខុស!) អ្នកនៅក្នុងក្រមសីលធម៌ AI និងច្បាប់ AI មានការព្រួយបារម្ភយ៉ាងខ្លាំងអំពីនិន្នាការដែលកំពុងកើនឡើងនៃការទាមទារដែលលាតសន្ធឹង។ អ្នក​ប្រហែល​ជា​និយាយ​ដោយ​គួរសម​ថា​មនុស្ស​មួយ​ចំនួន​កំពុង​និយាយ​ហួស​ពី​អ្វី​ដែល AI សព្វថ្ងៃ​អាច​ធ្វើ​បាន។ ពួកគេសន្មត់ថា AI មានសមត្ថភាពដែលយើងមិនទាន់អាចសម្រេចបាន។ នោះ​ហើយ​ជាអកុសល។ អាក្រក់ជាងនេះទៅទៀត ពួកគេអាចអនុញ្ញាតឱ្យខ្លួន និងអ្នកដ៏ទៃធ្លាក់ក្នុងស្ថានភាពធ្ងន់ធ្ងរ ដោយសារតែការសន្មត់ថា AI នឹងមានភាពស្រើបស្រាល ឬដូចមនុស្សក្នុងការធ្វើសកម្មភាព។

កុំធ្វើនិយតកម្ម AI ។

ការធ្វើដូច្នេះនឹងធ្វើឱ្យអ្នកជាប់ក្នុងអន្ទាក់ការពឹងផ្អែកដ៏ស្អិតរមួតនៃការរំពឹងថា AI ដើម្បីធ្វើរឿងដែលវាមិនអាចអនុវត្តបាន។ ជាមួយនឹងអ្វីដែលត្រូវបានគេនិយាយនោះ AI ជំនាន់ចុងក្រោយបំផុតគឺគួរឱ្យចាប់អារម្មណ៍សម្រាប់អ្វីដែលវាអាចធ្វើបាន។ សូម​ដឹង​ថា​មាន​កម្រិត​សំខាន់​ដែល​អ្នក​គួរ​តែ​ចងចាំ​ជា​បន្តបន្ទាប់​នៅ​ពេល​ដែល​ប្រើ​កម្មវិធី AI ជំនាន់​ណា​មួយ។

ប្រសិនបើអ្នកចាប់អារម្មណ៍លើភាពចលាចលដែលកំពុងរីកចម្រើនយ៉ាងឆាប់រហ័សអំពី ChatGPT និង Generative AI ដែលត្រូវបានប្រាប់ទាំងអស់នោះ ខ្ញុំបាននិងកំពុងធ្វើស៊េរីផ្តោតអារម្មណ៍នៅក្នុងជួរឈររបស់ខ្ញុំ ដែលអ្នកអាចស្វែងរកព័ត៌មាន។ នេះ​ជា​ការ​ក្រឡេក​មើល​ក្នុង​ករណី​ណា​មួយ​នៃ​ប្រធាន​បទ​ទាំង​នេះ​ចាប់​អារម្មណ៍​របស់​អ្នក៖

• 1) ការទស្សន៍ទាយនៃ AI ជំនាន់មុននឹងមកដល់។ ប្រសិនបើអ្នកចង់ដឹងពីអ្វីដែលទំនងជានឹងលាតត្រដាងអំពី AI ពេញមួយឆ្នាំ 2023 និងលើសពីនេះ រួមទាំងការជឿនលឿននាពេលខាងមុខនៃ AI និង ChatGPT នោះ អ្នកនឹងចង់អានបញ្ជីទូលំទូលាយនៃការព្យាករណ៍ឆ្នាំ 2023 របស់ខ្ញុំនៅ តំណភ្ជាប់នៅទីនេះ.
• 2) Generative AI និងការណែនាំអំពីសុខភាពផ្លូវចិត្ត។ ខ្ញុំបានជ្រើសរើសពិនិត្យមើលពីរបៀបដែល AI និង ChatGPT ដែលកំពុងត្រូវបានប្រើប្រាស់សម្រាប់ការណែនាំអំពីសុខភាពផ្លូវចិត្ត ដែលជានិន្នាការមានបញ្ហា ទៅតាមការវិភាគដែលផ្តោតលើរបស់ខ្ញុំនៅ តំណភ្ជាប់នៅទីនេះ.
• 3) មូលដ្ឋានគ្រឹះនៃ Generative AI និង ChatGPT ។ បំណែកនេះស្វែងយល់ពីធាតុសំខាន់ៗនៃរបៀបដែល AI ដំណើរការ និងជាពិសេសចូលទៅក្នុងកម្មវិធី ChatGPT រួមទាំងការវិភាគនៃការភ្ញាក់ផ្អើល និងអ្នកគាំទ្រនៅ តំណភ្ជាប់នៅទីនេះ.
• 4) Tension Between Teachers And Students Over Generative AI And ChatGPT. Here are the ways that students will deviously use generative AI and ChatGPT. In addition, here are ways for teachers to contend with this tidal wave. See តំណភ្ជាប់នៅទីនេះ.
• 5) បរិបទនិងការប្រើប្រាស់ AI ជំនាន់។ ខ្ញុំក៏បានធ្វើការពិនិត្យអណ្តាតតាមរដូវកាលអំពីបរិបទដែលទាក់ទងនឹងសាន់តាដែលពាក់ព័ន្ធនឹង ChatGPT និងការបង្កើត AI នៅ តំណភ្ជាប់នៅទីនេះ.
• 6) អ្នកបោកប្រាស់ដោយប្រើ AI ជំនាន់. ជាការកត់សម្គាល់ អ្នកបោកប្រាស់មួយចំនួនបានស្វែងយល់ពីរបៀបប្រើប្រាស់ AI និង ChatGPT ទូទៅដើម្បីធ្វើអំពើខុសឆ្គង រួមទាំងការបង្កើតអ៊ីមែលបោកប្រាស់ និងថែមទាំងផលិតកូដកម្មវិធីសម្រាប់មេរោគ សូមមើលការវិភាគរបស់ខ្ញុំនៅ តំណភ្ជាប់នៅទីនេះ.
• 7) កំហុសរបស់ Rookie ដោយប្រើ AI ជំនាន់. មនុស្សជាច្រើនកំពុងមើលរំលង និងគួរឱ្យភ្ញាក់ផ្អើលនូវអ្វីដែល AI និង ChatGPT ជំនាន់ថ្មីអាចធ្វើបាន ដូច្នេះខ្ញុំបានមើលជាពិសេសទៅលើការសន្និដ្ឋានដែល AI rookies មានទំនោរបង្កើត សូមមើលការពិភាក្សានៅ តំណភ្ជាប់នៅទីនេះ.
• 8) ទប់ទល់នឹងការជម្រុញ AI ជំនាន់មុន និង AI Hallucinations. ខ្ញុំពណ៌នាអំពីវិធីសាស្រ្តឈានមុខគេក្នុងការប្រើកម្មវិធីបន្ថែម AI ដើម្បីដោះស្រាយបញ្ហាផ្សេងៗដែលទាក់ទងនឹងការព្យាយាមបញ្ចូលការជម្រុញសមស្របទៅក្នុង AI ជំនាន់ថ្មី បូករួមទាំងមានកម្មវិធីបន្ថែម AI បន្ថែមសម្រាប់ការរកឃើញនូវអ្វីដែលហៅថា AI លទ្ធផលដែលបំភាន់ និងភាពមិនពិត ដូចជា គ្របដណ្តប់នៅ តំណភ្ជាប់នៅទីនេះ.
• 9) Debunking Bonehead ទាមទារអំពីការរកឃើញអត្ថបទដែលផលិតដោយ AI. មានការប្រញាប់ប្រញាល់មាសដែលបំភាន់នៃកម្មវិធី AI ដែលប្រកាសថាអាចបញ្ជាក់បានថាតើអត្ថបទដែលបានផ្តល់ឱ្យណាមួយត្រូវបានបង្កើតឡើងដោយមនុស្សធៀបនឹង AI បង្កើត។ សរុបមក នេះគឺជាការយល់ច្រឡំ ហើយក្នុងករណីខ្លះ ការទាមទារដែលមានក្បាលឆ្អឹង និងមិនអាចទទួលយកបាន សូមមើលការគ្របដណ្តប់របស់ខ្ញុំនៅ តំណភ្ជាប់នៅទីនេះ.
• 10) ការលេងតួនាទីតាមរយៈ AI ជំនាន់មុនអាចបង្ហាញពីគុណវិបត្តិនៃសុខភាពផ្លូវចិត្ត. អ្នកខ្លះកំពុងប្រើ AI ជំនាន់មុនដូចជា ChatGPT ដើម្បីដើរតួជាតួអង្គ ដែលកម្មវិធី AI ឆ្លើយតបទៅនឹងមនុស្សដូចជាមាននៅក្នុងពិភពរវើរវាយ ឬការកំណត់ផ្សេងៗ។ វាអាចមានផលប៉ះពាល់ដល់សុខភាពផ្លូវចិត្ត សូមមើល តំណភ្ជាប់នៅទីនេះ.
• 11) ការលាតត្រដាងជួរនៃកំហុសដែលបានចេញនិងភាពមិនពិត។ បញ្ជីដែលប្រមូលបានជាច្រើនកំពុងត្រូវបានដាក់បញ្ចូលគ្នា ដើម្បីសាកល្បង និងបង្ហាញពីលក្ខណៈនៃកំហុស និងភាពមិនពិតដែលផលិតដោយ ChatGPT។ អ្នក​ខ្លះ​ជឿ​ថា​នេះ​ជា​ការ​ចាំបាច់ ខណៈ​អ្នក​ខ្លះ​ទៀត​និយាយ​ថា​លំហាត់​នេះ​គ្មាន​ប្រយោជន៍ សូម​មើល​ការ​វិភាគ​របស់​ខ្ញុំ​នៅ តំណភ្ជាប់នៅទីនេះ.
• 12) សាលារៀនដែលហាមឃាត់ការបង្កើត AI ChatGPT កំពុងបាត់ទូក។ អ្នកប្រហែលជាដឹងថាសាលាផ្សេងៗដូចជានាយកដ្ឋានអប់រំនៅទីក្រុងញូវយ៉ក (NYC) បានប្រកាសហាមឃាត់ការប្រើប្រាស់ ChatGPT នៅលើបណ្តាញរបស់ពួកគេ និងឧបករណ៍ដែលពាក់ព័ន្ធ។ ថ្វីត្បិតតែវាហាក់ដូចជាការប្រុងប្រយ័ត្នដ៏មានប្រយោជន៍ក៏ដោយ ប៉ុន្តែវានឹងមិនរើម្ជុលឡើយ ហើយការនឹកទូកយ៉ាងសោកសៅ សូមមើលការគ្របដណ្តប់របស់ខ្ញុំនៅ តំណភ្ជាប់នៅទីនេះ.

អ្នកប្រហែលជាចាប់អារម្មណ៍ថា ChatGPT គឺផ្អែកលើកំណែកម្មវិធី AI ជំនាន់មុនដែលគេស្គាល់ថា GPT-3។ ChatGPT ត្រូវបានចាត់ទុកថាជាជំហានបន្ទាប់បន្តិច ដែលហៅថា GPT-3.5។ វាត្រូវបានគេរំពឹងទុកថា GPT-4 ទំនងជានឹងត្រូវបានចេញផ្សាយនៅនិទាឃរដូវឆ្នាំ 2023។ សន្មត់ថា GPT-4 នឹងជាជំហានដ៏គួរឱ្យចាប់អារម្មណ៍មួយឆ្ពោះទៅមុខក្នុងលក្ខខណ្ឌនៃការអាចផលិតអត្ថបទដែលហាក់ដូចជាកាន់តែស្ទាត់ជំនាញ កាន់តែស៊ីជម្រៅ និងគួរឱ្យស្ញប់ស្ញែង។ - បំផុសគំនិតអស្ចារ្យចំពោះសមាសភាពដែលវាអាចផលិតបាន។

អ្នកអាចរំពឹងថានឹងឃើញជុំថ្មីនៃការងឿងឆ្ងល់នៅពេលដែលរដូវផ្ការីកមកដល់ ហើយ AI ជំនាន់ចុងក្រោយបំផុតត្រូវបានចេញផ្សាយ។

ខ្ញុំលើកយកចំណុចនេះឡើង ពីព្រោះមានមុំមួយទៀតដែលត្រូវចងចាំ រួមមានកែងជើង Achilles ដ៏មានសក្ដានុពលចំពោះកម្មវិធី AI ជំនាន់កាន់តែល្អ និងធំជាងនេះ។ ប្រសិនបើអ្នកលក់ AI ណាមួយធ្វើឱ្យមានកម្មវិធី AI ជំនាន់ថ្មីដែលបញ្ចេញនូវភាពល្ងង់ខ្លៅ វាអាចបំផ្លាញក្តីសង្ឃឹមរបស់អ្នកបង្កើត AI ទាំងនោះ។ ការរីករាលដាលនៃសង្គមអាចបណ្តាលឱ្យ AI ជំនាន់ទាំងអស់ទទួលបានភ្នែកខ្មៅធ្ងន់ធ្ងរ។ មនុស្សច្បាស់ជានឹងមានការខកចិត្តយ៉ាងខ្លាំងចំពោះលទ្ធផលមិនល្អ ដែលបានកើតឡើងជាច្រើនដងរួចមកហើយ និងនាំឱ្យមានការថ្កោលទោសសង្គមដ៏ក្ដៅគគុកចំពោះ AI ។

ការព្រមានចុងក្រោយមួយសម្រាប់ពេលនេះ។

អ្វីដែលអ្នកឃើញ ឬអាននៅក្នុងការឆ្លើយតប AI ជំនាន់នោះ។ ហាក់ដូចជា ដើម្បីបង្ហាញការពិតសុទ្ធសាធ (កាលបរិច្ឆេទ ទីកន្លែង មនុស្ស។

បាទ/ចាស កាលបរិច្ឆេទអាចត្រូវបានផ្សំ ទីកន្លែងអាចត្រូវបានបង្កើតឡើង ហើយធាតុដែលយើងរំពឹងថានឹងលើសពីការប្រមាថគឺ ទាំងអស់ ប្រធានបទនៃការសង្ស័យ។ កុំជឿអ្វីដែលអ្នកបានអាន ហើយរក្សាភ្នែកដែលមានការសង្ស័យនៅពេលពិនិត្យមើលអត្ថបទ ឬលទ្ធផលនៃ AI ណាមួយ។ ប្រសិនបើកម្មវិធី AI ជំនាន់ថ្មីប្រាប់អ្នកថា Abraham Lincoln បានហោះហើរជុំវិញប្រទេសក្នុងយន្តហោះឯកជនរបស់គាត់ នោះអ្នកនឹងដឹងច្បាស់ថានេះគឺជាជំងឺគ្រុនចាញ់។ ជាអកុសល មនុស្សមួយចំនួនប្រហែលជាមិនបានដឹងថា យន្តហោះចម្បាំងមិនមាននៅក្នុងសម័យរបស់គាត់ទេ ឬពួកគេអាចដឹង ប៉ុន្តែមិនបានកត់សំគាល់ថា អត្ថបទនេះធ្វើឱ្យមានការអះអាងខុសឆ្គង និងហួសចិត្ត។

កម្រិតនៃការសង្ស័យដែលមានសុខភាពល្អ និងផ្នត់គំនិតមិនជឿជាប់លាប់នឹងជាទ្រព្យសម្បត្តិដ៏ល្អបំផុតរបស់អ្នកនៅពេលប្រើ AI ជំនាន់។

យើងត្រៀមខ្លួនរួចជាស្រេចដើម្បីឈានទៅដំណាក់កាលបន្ទាប់នៃការបំភ្លឺនេះ។

ការដោះលែងសត្វសាហាវ

Now that we’ve got the fundamentals established, we can dive into the business-oriented and societal repercussions due to the ChatGPT API aspects.

An announcement was recently made by Microsoft in conjunction with OpenAI about the upcoming availability of ChatGPT on the Azure cloud platform of Microsoft (per online posting entitled “General Availability of Azure OpenAI Service Expands Access to Large, Advanced AI Models with Added Enterprise Benefits”, January 16, 2023):

• “Large language models are quickly becoming an essential platform for people to innovate, apply AI to solve big problems, and imagine what’s possible. Today, we are excited to announce the general availability of Azure OpenAI Service as part of Microsoft’s continued commitment to democratizing AI, and ongoing partnership with OpenAI. With Azure OpenAI Service now generally available, more businesses can apply for access to the most advanced AI models in the world—including GPT-3.5, Codex, and DALL•E 2—backed by the trusted enterprise-grade capabilities and AI-optimized infrastructure of Microsoft Azure, to create cutting-edge applications. Customers will also be able to access ChatGPT—a fine-tuned version of GPT-3.5 that has been trained and runs inference on Azure AI infrastructure—through Azure OpenAI Service soon.”

You might have noticed in that statement that other various AI apps that have been devised by OpenAI will also be available. Indeed, some of those AI apps have already been accessible for quite a while, as mentioned further in the recent above pronouncement: “We debuted Azure OpenAI Service in November 2021 to enable customers to tap into the power of large-scale generative AI models with the enterprise promises customers have come to expect from our Azure cloud and computing infrastructure—security, reliability, compliance, data privacy, and built-in Responsible AI capabilities” (ibid).

I earlier mentioned that both AI Ethics and AI Law are trying to balance the អាយអេសដើម្បីភាពល្អ aspirations with the potential អាយអេសសម្រាប់អាក្រក់ that can at times arise. Within the AI realm, there is a movement afoot toward having ទទួលខុសត្រូវ AI or sometimes coined as Trustworthy AI or Human-Centered AI, see my coverage at តំណភ្ជាប់នៅទីនេះ. All AI makers are urged to devise and field their AI toward អាយអេសដើម្បីភាពល្អ and seek overtly to curtail or mitigate any អាយអេសសម្រាប់អាក្រក់ that might emerge.

This is a tall order.

In any case, the aforementioned pronouncement did address the Responsible AI considerations:

• “As an industry leader, we recognize that any innovation in AI must be done responsibly. This becomes even more important with powerful, new technologies like generative models. We have taken an iterative approach to large models, working closely with our partner OpenAI and our customers to carefully assess use cases, learn, and address potential risks. Additionally, we’ve implemented our own guardrails for Azure OpenAI Service that align with our Responsible AI principles. As part of our Limited Access Framework, developers are required to apply for access, describing their intended use case or application before they are given access to the service. Content filters uniquely designed to catch abusive, hateful, and offensive content constantly monitor the input provided to the service as well as the generated content. In the event of a confirmed policy violation, we may ask the developer to take immediate action to prevent further abuse” (ibid).

The crux of that Responsible AI perspective is that by requiring a formal request to access ChatGPT on a program API basis, there is a chance of weeding out the unsavory submissions. If there is suitable due diligence in choosing which other firms and their programs can access ChatGPT, perhaps there is a fighting chance of preventing the full wrath of a released Kraken.

Maybe yes, maybe not.

Some pundits are wringing their hands that the money-making possibilities of allowing the ChatGPT API to be put into use will strain the counterbalancing notion of wanting to keep the beast reasonably and safely contained. Will the scrutiny really be sufficiently careful upfront? Might we see instead that a loosey-goosey-wobbly approval process occurs as the volume of requests gets out of hand? Some are fearful that only once the cat is out of the bag might a belated fuller scrutiny truly occur, though by then the damage will already have been done.

Well, you can at least give due credit that there is a vetting process involved. There are some generative AI apps that either lack a coherent vetting process or that are of a sketchy cursory nature. In addition, there are open-source versions of generative AI that generally can be used by nearly anyone that wants to do so, albeit some modicum of licensing restrictions are supposed to be followed (trying to enforce this is harder than it might seem).

Let’s take a quick look at the existing rules regarding limiting access to the Azure OpenAI service to see what other software makers will need to do to potentially connect up with ChatGPT. Per the online posted Microsoft Policies (latest posting indicated as December 14, 2022):

• “As part of Microsoft’s commitment to responsible AI, we are designing and releasing Azure OpenAI Service with the intention of protecting the rights of individuals and society and fostering transparent human-computer interaction. For this reason, we currently limit the access and use of Azure OpenAI, including limiting access to the ability to modify content filters and modify abuse monitoring. Azure OpenAI requires registration and is currently only available to managed customers and partners working with Microsoft account teams. Customers who wish to use Azure OpenAI are required to submit a registration form both for initial access for experimentation and for approval to move from experimentation to production.”
• “For experimentation, customers attest to using the service only for the intended uses submitted at the time of registration and commit to incorporating human oversight, strong technical limits on inputs and outputs, feedback channels, and thorough testing. For production, customers explain how these have been implemented to mitigate risk. Customers who wish to modify content filters and modify abuse monitoring after they have onboarded to the service are subject to additional scenario restrictions and are required to register here.”
• “Access to the Azure OpenAI Service is subject to Microsoft’s sole discretion based on eligibility criteria and a vetting process and customers must acknowledge that they have reviewed and agree to the Azure terms of service for Azure OpenAI Service. Microsoft may require customers to re-verify this information. Azure OpenAI Service is made available to customers under the terms governing their subscription to Microsoft Azure Services, including the Azure OpenAI section of the Microsoft Product Terms. Please review these terms carefully as they contain important conditions and obligations governing your use of Azure OpenAI Service.”

That is on the Microsoft side of things.

OpenAI also has its usage policies associated with its API:

• “We want everyone to be able to use our API safely and responsibly. To that end, we’ve created use-case and content policies. By following them, you’ll help us make sure that our technology is used for good. If we discover that your product doesn’t follow these policies, we’ll ask you to make necessary changes. If you don’t comply, we may take further action, including terminating your account.”
• “We prohibit building products that target the following use-cases:”
• “— Illegal or harmful industries”
• “— Misuse of personal data”
• “— Promoting dishonesty”
• “— Deceiving or manipulating users”
• “— Trying to influence politics”
• “The following set of use cases carry a greater risk of potential harm: criminal justice, law enforcement, legal, government and civil services, healthcare, therapy, wellness, coaching, finance, news. For these use-cases, you must:”
• “1) Thoroughly test our models for accuracy in your use case and be transparent with your users about limitations”
• “2) Ensure your team has domain expertise and understands/follows relevant laws”
• “We also don’t allow you or end-users of your application to generate the following types of content:”
• “— Hate”
• “— Harassment”
• “— Violence”
• “— Self-harm”
• “— Sexual”
• “— Political”
• “— Spam”
• “— Deception”
• “— Malware”

A big question will be whether these ideals can be observed if there is a fervent rush of requests to connect with ChatGPT. Perhaps there will be an overwhelming tsunami of requests. The human labor to examine and carefully vet each one could be costly and difficult to manage. Will the desire to be suitably restrictive get watered down, inadvertently in the face of the immense demand for access?

As the famed witticism goes, the best of plans can sometimes be set asunder upon first contact with abundant forces.

There is also a lot of leeway in how to interpret the stated rules. As we have seen in general about the rise of disinformation and misinformation, trying to separate the wheat from the chaff can be quite challenging. How is one to determine whether generated content abides by or violates the provisions of not being hateful, political, deceptive, and the like?

A looming difficulty could be that if the ChatGPT API is made available to a software maker that is pairing up their program with ChatGPT, and the resulting output unambiguously violates the stated precepts, will the horse already be out of the barn? Some suggest that there is a strong possibility of reputational harm that can be incurred to all parties involved. Whether this can be overcome by simply disengaging the API to that particular offender is unclear. The damage, in a sense, might linger and spoil the barrel all told. Plenty of blame will be spirited to all comers.

Stratifying The API Pairings

I noted earlier that the pairings to ChatGPT can be conveniently grouped into two major intentions:

• Genuine Pairing With ChatGPT
• Fakery Pairing With ChatGPT

Let’s examine first the genuine or bona fide pairings.

As background, the way that this occurs is somewhat straightforward. The ChatGPT app allows other programs to invoke the app. Typically, this would consist of say a program we’ll call Widget that passes to ChatGPT a prompt in text format, and then after ChatGPT does its thing, an essay or text is returned to the program Widget. This is almost like a person doing the same thing, though we will have a program do those actions in lieu of a person doing so.

For example, suppose someone devises a program that does searches on the web. The program asks a user what they want to search for. The program then provides the user a listing of various search hits or finds that hopefully showcase relevant websites based on the user query.

Imagine that the firm that makes this web searching program wants to spruce up its app.

They request access to the ChatGPT API. Assume they do all the proper paperwork and ultimately get approved. Their program that does web searches would then need to be modified to include a call-out to the ChatGPT app via the API. Assume they opt to make those mods.

Here’s how that might then work altogether. When a user enters a query for a web search into the core program, this program not only does a conventional search of the web, but it also passes the query over to ChatGPT via the API. ChatGPT then processes the text and returns a resultant essay to the core program. The web search core program now presents to the user two facets, namely the web search results and the additional outputted essay from ChatGPT.

A person using this core program will not necessarily know that ChatGPT was being used on the backend. It can happen within the confines of the core program and the user is blissfully unaware that ChatGPT was involved. On the other hand, the core program could be devised to inform the user that ChatGPT is being used. This usually hinges on whether the makers of the core program want to reveal that the other app, in this case, ChatGPT, was being used. In some arrangements, the maker of the being-invoked program insists that the API-invoking program must let users know that the other program is being utilized. It all depends on preferences and licensing particulars.

For genuine pairings, here are the customary approaches:

• 1) Straight pass-thru to ChatGPT
• 2) Add-on to augment ChatGPT
• 3) Allied app that coincides with ChatGPT
• 4) Fully integrative immersion with ChatGPT

Briefly, in the first listed approach, the idea is that I might devise a program that is merely a frontend for ChatGPT, and as such, all that my program does is passes the text to ChatGPT and gets the text back from ChatGPT. I make available my program for anyone that wants to use ChatGPT and who otherwise hadn’t signed up to use it. This is one approach.

Second, I might devise a program that serves as an add-on to ChatGPT. For example, when ChatGPT generates an essay, it might contain falsehoods. Suppose I craft a program that examines the ChatGPT output and tries to screen for falsehoods. I make my program available such that people enter a text prompt into my program, which then sends the prompt to ChatGPT. ChatGPT produces an essay that comes back to my program. Before my program shows you the essay, it prescreens the essay and attempts to flag or remove falsehoods. You then see the resulting essay after my program has done the screening.

The third approach consists of having an allied app that in sense coincides with ChatGPT. Suppose I develop a program that aids people in doing creative writing. My program provides canned tips and suggestions on how to write creatively. The program merely prods or spurs the user to do so. Meanwhile, what I’d like to be able to do is show the user what creative writing consists of. As such, I establish an API with ChatGPT. My program then takes a prompt from the user and invokes ChatGPT to provide a blub of an essay that might demonstrate creative writing. This could be done iteratively and invoke ChatGPT multiple times in the process.

In the case of the fourth listed approach, ChatGPT is fully integrated into some other program or set of programs. For example, if I had a word-processing app and a spreadsheet app, I might want to integrate ChatGPT into those apps. They would in a manner of speaking function hand-in-hand with each other. I’ll be covering in an upcoming column posting the floated possibility that Microsoft could opt to infuse ChatGPT into their office productivity suite, so be on the look for that coming analysis.

Those then are the key ways that a genuine pairing might take place.

Let’s next consider some of the fakery pairings.

Here are some overall fakery pairings that you ought to be watchful of:

• Gimmickry Pairing With ChatGPT – marginally utilizes ChatGPT, mainly done for show and to garner publicity with no added value
• Alluring Promises About ChatGPT Pairing – software vendor claims they are in the midst of pairing to ChatGPT, seeking to be in the shining spotlight, when the reality is that they aren’t going to do so and are doing a classic head fake and rendering a false pledge
• Knockoffs Proclaiming To Be ChatGPT-like – rather than pairing with ChatGPT, some software vendors will use something else, which is fine, but they will attempt to imply it is ChatGPT when it is not, hoping to get some of the afterglows of ChatGPT
• ផ្សេងទៀត – lots of additional outstretched and conniving schemes are conceivable

There will undoubtedly be a lot of shenanigans going on about all of this. It will be part and parcel of the release of the Kraken.

សន្និដ្ឋាន

Allow me to toss some wrenches and other obstructions into this matter.

What about the cost?

Currently, those that signed up to use ChatGPT are doing so for free. I have previously remarked that at some point there will need to be outright monetization involved. This might entail being charged a per transaction fee or maybe paying for a subscription. Another possibility is that ads might be used to bring in the dough, whereby each time you use ChatGPT an ad will appear. Etc.

Those that opt to establish a pairing with ChatGPT via the API should seriously be mulling over the potential costs involved. There is likely a cost pertaining to the use of the Microsoft Azure cloud for the running of the ChatGPT app. There is bound to be a cost from OpenAI to use the ChatGPT API and invoke the AI app. A software vendor will incur their own internal costs too, such as modifying their existing programs to use the API or developing programs anew around the pairing with ChatGPT. Envision both a getting started cost and an ongoing upkeep set of costs too.

The gist is that this layering of costs is going to moderate to some degree the gold rush toward leveraging the ChatGPT API. Software vendors presumably should do a prudent ROI (return on investment) calculation. Will whatever they can make via augmenting their program by using ChatGPT bring in sufficient added monies to offset the costs?

Not everyone is necessarily going to be quite so cost-conscious. If you have deep pockets and believe that your use of ChatGPT will propel your program into the most known or highly recognized realm of apps, you might decide that the cost right now is worth it. Build a name for your app by riding on the coattails of ChatGPT. Down the road, once your program is popular or otherwise making money, you either make up for the earlier lack of profit or write it off as the cost required to get into the big-time.

A small startup backed by a Venture Capital (VC) firm might be willing to fork over a chunk of its seed investment to get paired up with the ChatGPT API. Fame might instantly arise. Fortune might be a long way down the road, but that’s something to be dealt with later on. Grab the limelight when the getting is good, as they say.

One supposes that there might be non-profits and social enterprises that will decide to also kick the tires on this. Suppose a non-profit firm identifies a beneficial use of invoking the ChatGPT API that will seemingly support their altruistic or societally beneficial goals. Maybe they raise funds for this via an online funding campaign. Perhaps they try to cut a deal with the vendors so that they pay a nominal amount or get the use for free.

ពេលវេលានឹងប្រាប់។

The last pointer that I’ll leave you with is the risk factor.

I don’t want to seem unduly downbeat, but I mentioned earlier that the outputs from ChatGPT can contain falsehoods and have other potential downsides. The thing is, if a software vendor making an app gets mixed into using the ChatGPT API, they run the risk of having sour and dour outputs, which they ought to be anticipating. Do not put your head in the sand.

The troubles arise that those outputs could taint the app that opts to utilize them. In that sense, your app that is at first riding the glory of ChatGPT could also end up plowing into a brick wall if the outputs provided via ChatGPT are relied upon. Perhaps the outputs are presented to users and this causes a horrendous ruckus. They take out their angst on you and your app.

In turn, you try to point a finger at ChatGPT. Will that get you out of the conundrum? Once there is a stink, it permeates widely and few are spared. Along those lines, it could be that via a widened use of ChatGPT, the awareness of the foul outputs gets more commonly known. Thus, oddly enough, or ironically, the expanded use of ChatGPT due to the API could shoot their own foot.

I don’t want to conclude my remarks with a sad face so let’s try to shift into a happy face.

If all of the aspirational limits and constraints are mindfully and judiciously followed and adhered to rigorously, tapping into ChatGPT via the API can be a good thing. The chances are too that this will further spur other generative AI apps into action. The rising tide could rise all boats.

That seems upbeat.

You might know that Zeus was said to be in control of the Kraken. Sophocles, the ancient Greek playwright, said this about Zeus: “The dice of Zeus always fall luckily.”

Maybe the same will be said of how generative AI will inevitably land, let’s at least hope so.