Five Steps to Create a New AI Model

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.

[skffingtonai]

This guy is a great teacher, thank you IBM, here's your feedback!

[Parsley]

This is an awesome description. More please!

[RajendraBichu]

Really a very good Teacher, explains in simple lucid model - fantastic

[AutobotikaSocial]

Amazing video! very clear and easy to follow thru!

[harmeetsingh]

Wow, great explanation with simple breakdown.

[MrThurobrand]

Love your brewing videos. These are pretty good as well. ;)

[DilshanBoange]

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.

[ericgcastile]

Great info and how you made this video is amazing editing!

[JohnJamalSingh]

Good teacher...clearly covered in a short time...

[ktqdtnh]

Another useful video ! Thank you IBM !

[SiegelBantuBear]

Martin you are saving lives🤣. Thank you Sir🙏🏼👍🏼

[ShaunDavid-vjpl]

sorry to ask but a burning question are you writing on that board backwards like this ledom so we can see it like model ?

[KumR]

Are u referring to large models like LLM or smaller models for a specific purpose or both?

[rezakopaee]

What is the name of the presenter? Engaging, Clever, Concise. Awesome work.

[anthonyferrentino]

Are steps 1-3 here taken care of through the foundational model?

[dzulihsan]

Cool presentation. Wonder how you write backwards on the screen. Mind to share with the viewers what application you are using?

[JiffyDealer]

lol, had to rewind 3 times because it sounded like “Jupiter data” when saying “duplicate data”

[rajdeepnaha]

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

[petermaree]

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.

[agutinleonid]

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?