Hello everyone!
We are having an education day with IT and are trying to figure out what to do with AI and Epicor. Do the users here use ChatGPT with Epicor? If so, what are the concerns about using AI and Epicor? Are we concerned about what we are entering into ChatGPT, do we even care? How are you all using ChatGPT?
I am not, but I do see great potential, and real concerns with it.
Iām not ready to jump down another rabbit hole at this moment.
I saw some neat stuff, and heard some good conversations at Insights 2023.
@josecgomez has played with it, and Eli from Six S (username? @AlwaysFocus ), has played with it a bit.
I didnāt hear anything with AI at Insights. I probably didnāt even think to go to those talks because we arenāt in the position currently to leverage AI just yet. We will in the future though, Iām sure.
MS during MS Buld conf advertised AI for every company with data isolated from each other, like now tenants are isolated in Azure. So it will be secure to use.
Also fun examle of programiing with CoPilot from there. Scott and Mark Learn to Code | WAKEY06 - YouTube
epinoob @Insights
Iāve played with it some:
However I would be extremely hesitant to connect it to your live data / environment thereās a lot of potential for your information to end up in the wrong place.
As @Olga said they are making strides on their privacy side but it is still not good enough IMO. Also I have huge concerns about it providing bad data or āhallucinationsā there isnāt a good way to check for data accuracy yet on how it came up with the responses as some Lawyers are finding out the hard way
https://www.cnn.com/2023/05/27/business/chat-gpt-avianca-mata-lawyers/index.html
Its cute to play with in very limited instances and you manually using it to enhance your productivity Iām all for, but Iād be hard pressed to do anything close to an integration with it (yet) too much unknown.
One big problem, is no matter what it outputs, it does it with the confidence of a 4 year old
in a Batman costume.
Less savvy users will have a hard time distinguishing good data from bad.
Edit: Some savvy users as well 
Right I think they need to have a confidence meter and or a āsourceā link that tells you where / how it found that data and you should be able to also ask it to give you only answers it is 90+% certain about or something like that.
it is statistics, not confidence
I still think we are 10-15 years out on that one.
Of course, it changes so fast, I could be wrong.
I hope Iām not wrong though. The world is changing fast enough as it is.
Fact Checkers for the robot lol?
In programming you can just try to run the program In my case it invented non-existing API and/or suggested the idea that failed on run immediately.
Microsoft Fabric - and not Service Fabric because naming things is hard. Looks very compelling though.
Iām really good at guess and check code. Boom Testing is my specialty!
One thing yāall really need to remember though, as impressive as this all is,
there is something just as powerful, and more sinister running right alongside it:
MARKETING
Chat GPT (as it currently exists) needs to have a fact checkerā¦ look at this screenshot which shows the quesiton and answer, and my calculator showing the ārealā answer:
and when I corrected it, it did give me the correct answer, BUT, the work it showed is still incorrect. I have been asking this same question multiple times over the past few months.
It is being recommended to start setting company policies regarding the use of AI Chats.
it becomes annoying soon - feeling like you are talking with very polite but dumb person
Its funny how āweā went from not trusting the āsinisterā black box algorithms that determine which social media posts we see, to being so eager to employ said method to more serious issuesā¦
Current company I work for isnāt very data centric, so I donāt think a direct application of AI to our ERP system would have much benefit. However, I can see AI making certain kinds of automation much faster and easier, so it could impact the factory floor a lot.
