Five Steps to Create a New AI Model
Vložit
- čas přidán 12. 06. 2024
- Learn more about watsonx: ibm.biz/BdvDnr
AI promises to touch every aspect of work and life, but how do they get made?
In this video Martin keen walks through a five step framework for how to build and deploy AI models.
This guy is a great teacher, thank you IBM, here's your feedback!
This is an awesome description. More please!
Really a very good Teacher, explains in simple lucid model - fantastic
Amazing video! very clear and easy to follow thru!
Wow, great explanation with simple breakdown.
Good teacher...clearly covered in a short time...
Love your brewing videos. These are pretty good as well. ;)
Great video, thanks!
Another useful video ! Thank you IBM !
What an awesome video! Great insights delivered in a simple layman friendly way. Martin's style and tone makes this otherwise very high tech topic actually graspable at a very basic level for people like me who do not come from a tech background. Very commendable. Truly laudable. Please keep up the good work. Cheers to IBM, Martin Keen & team.
Martin you are saving lives🤣. Thank you Sir🙏🏼👍🏼
Great info and how you made this video is amazing editing!
Love your brew channel 👍
Thank you!
Are u referring to large models like LLM or smaller models for a specific purpose or both?
00:06 Deep learning enables building specialized models with enough labeled data.
01:04 Fine tuning and adapting base foundation models speeds up AI model development
01:58 Filtering data for language, content, and duplicates.
02:40 The process of training an AI model involves two stages: data versioning and tagging, and model selection.
03:29 Choose a foundational model based on use case and match data pile
04:25 Model Card can be created to showcase the trained model and its benchmark scores.
05:19 Deployment of the model has options for cloud service or edge application
06:20 Watsonx.data is a modern data lakehouse that establishes connections with data repositories and Watsonx.governance manages data and model cards
Thank you
Are steps 1-3 here taken care of through the foundational model?
Hi there! Great explanation, thank you, but it is not clear what exactly contains the “fine tune” step. What the difference between “fine tune” and “prepare data”+”train” steps? essential just in amount of data or the “fine tune” process designed differently?
It sounds like the fine tune step is what optimizes the bridge between the trained model and all it's potential outputs. I've used models that appear to have been trained on all the right data but are a major pain when it comes to prompting and producing accurate outputs. I'm no expert but I believe that's what it entails
Hi there, I am a Technical Graphic Designer (FA) working in a Pre-press environment. I work in Adobe Creative suite and Prenergy, surrently. I am interested in this course. But need assurance on how I can combine skills from AI or ML to automate repetitive tasks and standardise output quality from pre-flighting, proofing all the way to image processing stage. What do you advise / recommend?
Evening. Love your screen and layout. Was this done and then flipped in the video? Did you use a screen and ultra violet light glas board or do you have na app with smart pen? Really curious.
You are right! He's using his left hand, which suggests that the video was flipped for the viewers.
Cool presentation. Wonder how you write backwards on the screen. Mind to share with the viewers what application you are using?
I was honestly looking for this comment
There are a couple of ways of doing light board presentations. One is to reverse the video display after creating another is to use a mirror while video-ing.
How much time each phase take?
Great video thank u ❤🖤🦾🍏
The brew meister !
What is the name of the presenter? Engaging, Clever, Concise. Awesome work.
Could anyone explain clearer the stage 4 which is about tuning model with additional local data, what does it mean ?
Currently, I find that LLM can read vector database to get knowledge and answer the questions, is this the stage 4 ?
It basically means training it again, but for a specific use. Models that are fine tuned can be specialized in, for ex, coding, but maybe arent great at generating images or speech of essays, etc.
lol, had to rewind 3 times because it sounded like “Jupiter data” when saying “duplicate data”
How do you make presentations like this, i mean how do you write on screen?
Hello teacher, I am a viewer from China and I think your video content is very good. Also, I would like to know how you achieved this video effect? Really interested.
It's a light board and there are a couple of ways of using it.
hi. please help me. how to create custom model from many pdfs in Persian language? tank you.
Better than my coding teacher
can someone who has never learned coding make an AI tool in 3 weeks?
@@maryabouissa after spending 8-weeks and hundreds$ I made to Amazon for a RTX 4060. Probably another 2 weeks to get eGPU/laptop running
Does red hat and open shift will perform all the 5 stages without any human intervention
how to edit an video like this
AI
damn, there is a white pixel, bottom left
Pls teach us all ML ALGORITHMS with Mathematics
But it was supposed to be easily trained to use machine data ...not text ....the machine data contains user information and actions so ....why they ignore it ...i don't want application i want better performing machine based on my actions real and responsive
Permission to embed this video on our website 😅
Are we not gonna talk about how he writes backwards?
New AI Models and Robotics-
Today's founders are faced with scale and labour issues simultaneously, requiring more demands on artifical intelligence and robotics. With robotic AI models, similar to 3d printing, the business development of tomorrow will be easily supported by IBMs product lines. Robotic foundation models to move pallets, for welding, and to cap bottles already exist, propelling the future of AI to a contemporary bar with home brews and reciepes for a larger scale. Larger scale creates larger profits, more creative projects, and yes more data.
- IBM and Kawasaki Robotics preparation of data ( cough...you have to pay to read the rest).😢
there are no actual steps here, there is no download link. another nothingburger. no training AI or anything
this video is completely useless.
You lost me at “hate speech”, as such a thing does not actually exist, and any intelligent person worth learning from would have already worked that logic out.