A Course Has Two Outputs: What You Can Do, And What You Can Show
Ten courses, ten identical rows, a famous logo beside each one. Rutgers, Michigan, Google, Vanderbilt, IBM, Microsoft, Colorado and three platforms. As a shortlist of places to go and learn something it does the job, and the footer is the right one: course names, providers and availability change often. Check each provider before you enrol.
What the equal row heights hide is that these ten are not comparable commitments.
Two of them are professional certificates — multi-course programmes that take months and are designed to be put on a CV. Others are single short courses: generating images with one specific model, or clustering analysis, each of which is an afternoon or a week. On the card they occupy exactly the same amount of space.
And underneath that, a second split that matters more. A course has two outputs — what you can do afterwards, and what you can show afterwards — and these ten weight them very differently depending on who made them. The two are not correlated, which is precisely why choosing by the biggest logo goes wrong.
Three kinds of provider
| UNIVERSITIES | VENDORS | PUBLISHERS & PLATFORMS |
|---|---|---|
| Rutgers · Michigan · Vanderbilt · Colorado | Google · IBM · Microsoft | Packt · Codio · DeepLearning |
| Concepts, methods and vocabulary | Their own platform, taught properly | One specific skill, taught directly |
| Transfers anywhere. Still true in five years | Transfers if your employer uses that platform. Otherwise not | Transfers as a skill, but nobody checks the certificate |
| Good for: understanding why things work | Good for: a recognised badge and a job filter | Good for: doing one thing by Friday |
All ten, as the card lists them
The first two columns are the card’s own. The rest is ours.
| # | Provider | Course, as named | Kind | Teaches a… | Scale |
|---|---|---|---|---|---|
| 1 | Rutgers | AI Concepts and Strategy | University | Concept | Course |
| 2 | Michigan University | Top 100 Best Selling Products: AI | University | Unclear — the title does not describe an AI skill | Course |
| 3 | Google Cloud Digital Leader Training Professional Certificate | Vendor | Platform | Certificate programme | |
| 4 | Vanderbilt | GPT Vision: Seeing the World Through Generative AI | University | Concept, tied to one model family | Course |
| 5 | IBM | IBM AI Developer Professional Certificate | Vendor | Platform | Certificate programme |
| 6 | Microsoft | Microsoft Azure Machine Learning for Data Scientists | Vendor | Platform | Course |
| 7 | Packt | Advanced Semantic Processing | Publisher | Skill | Course |
| 8 | Codio | Generating Images with DALL-E | Platform | Skill, tied to one named model | Short course |
| 9 | University of Colorado | Clustering Analysis | University | Concept — and not AI-specific | Short course |
| 10 | DeepLearning | Introduction to Generative AI for Software Development | Publisher | Skill | Course |
WATCH THE ORIGINAL — 20 SECONDS
All ten rows with their providers and course titles, and the footer about names and availability changing.
Why a platform course and a concept course are different purchases
Three of these ten teach you a specific cloud or developer platform. They are generally excellent — the companies that built the thing are well placed to teach it, the material is current, and the certificate carries a name that recruiters recognise and filter on.
They are also product training. What you learn is how that platform does machine learning, with that platform’s names for things, inside that platform’s console. If you end up working somewhere that runs it, you are immediately useful. If you end up somewhere that runs a competitor, a meaningful chunk of what you learned is vocabulary you now have to translate.
A concept course has the opposite shape. Clustering analysis, taught at a university, is the same clustering analysis on every platform that has ever existed and every one that will. It is slower to feel useful and it does not expire. Nobody is impressed by the certificate, and nobody needs to be, because what you got was the ability to do the thing.
Neither is better. They are different purchases, and the mistake is buying one while believing you bought the other — enrolling in a vendor certificate hoping for deep understanding, or in a university short course hoping for a badge that opens doors.
Decide which output you need first
| If this is you | Which output you need | Which rows | What to check before enrolling |
|---|---|---|---|
| You need to get past a CV filter | What you can show | The vendor certificates — rows 3 and 5 | Which employers in your market actually ask for it |
| You want to understand what is happening | What you can do | The university concept courses — rows 1, 4, 9 | Whether it assumes maths you do not have yet |
| You need one specific thing by Friday | What you can do | The short skill courses — rows 7, 8, 10 | That the tool or model it names is still current |
| Your employer already runs one platform | Both at once | That vendor’s row, and only that one | Whether your employer will pay for it |
| You are changing field entirely | Show first, then do | One certificate, then concepts underneath it | That the certificate is the recognised one, not a lookalike |
| You make videos about this | What you can do | Any of them — but the project matters more | Nobody watches a certificate. They watch the thing you built |
Four honest caveats
Two rows do not obviously describe an AI course. One is a title about best-selling products, and another is clustering analysis — a classical statistics and data technique that predates the current wave by decades and is genuinely useful without being AI in the sense the headline means. Neither is a bad course. They are a reminder that a list titled “AI courses” has elastic edges.
A certificate is not an accreditation. The professional certificates here are respected industry credentials, which is a different thing from a qualification awarded by an accredited institution. That distinction matters if you are studying for a visa, a professional registration or a formal requirement. For a hiring filter, the industry credential often does more work.
One course names a specific model. Row eight is built around one named image model, and model names in this field change faster than course catalogues do. The skill underneath — writing a description that produces the picture you wanted — transfers. The interface shown in the videos may not. That is the same dating signal we wrote about in You Can Date This Card By One Name On It.
The card does not claim these are free, and mostly they are not. Several of these providers run on monthly subscriptions, with certificates charged separately. Check the price and whether the certificate costs extra before you start a programme that takes months — and remember that the lectures being free is not the same as having a course, which we set out in A University Gives You Five Things. YouTube Gives You One.
Questions people ask
Will a certificate get me a job? It gets you past filters, which is not nothing. What gets you the job is something you built, which no row on this card provides — you have to make that yourself afterwards.
Should I do a vendor certificate if I do not work there? Only if the employers you are targeting run that platform. Otherwise you are learning somebody’s product on your own time for a badge your market does not read.
How do I not drop out in week three? The same way as with anything self-paced: a sequence, a prerequisite check, something that marks your work, and a date. Those four are what a playlist never gives you and what you have to supply yourself.
What if I just want to be good at this? Then do one of the concept rows, build something small with it, and write down what you still cannot explain. That three-step loop — do, test against something you did not write, record the gap — is the method we pulled out of a different BANI card in Twenty Channels, Fifteen Instructions.
Pick the output, then the logo
Ten rows, ten famous names, two outputs and three kinds of provider. The rows look equal and they are not: two are programmes measured in months, several are afternoons, and the thing each one actually hands you at the end is different in kind.
So decide first which output you are short of. If it is the badge, take a vendor certificate from the platform your market actually runs. If it is the understanding, take a concept course and accept that nobody will be impressed by it. If it is one specific skill by Friday, take the short course and do not enrol in a three-month programme to get there.
And whichever you pick, finish it with something built. A certificate says you attended. A thing you made says you can.
Learning a lot does not make anyone good. Doing a lot does.
If you want a deadline rather than an eleventh saved course list, we run a free challenge — there is no barrier to entry, only a commitment fee that is refunded when you finish. It exists because a deadline produces the project, and an enrolment produces an enrolment.
Which output are you actually short of — the doing or the showing? Tell us in the comments under the original reel.
More on building a channel for a global audience at mmoyoutube.com.
