.svg){kind=small}
Mussenstraat 15
1223 RB Hilversum
Selecteer taal
-
- Blogs
- Generative AI is 💩
Generative AI is 💩
until you get it right
OPINION | I know how hard it is to get AI right, and how easy it is to make you scream. 😱 Still, I am excited about all the experiments to build truly meaningful and intelligent features into your learning platform.
FOMO is real
The promise of AI is mindboggling. The examples shared on public social networks are out of this world. Everyone trips over their own feet with predictions on how generative AI will change the future of learning, the future of work, and everything around us.
Analyst after (self-proclaimed) analyst paints a both promising and scary picture of the future of creation.
The fear of missing out is real.
My first sceptic view on gen AI was: long live mediocracy.
It's hard to get AI right
I mean I wander around the fields of AI for 15 years already. I built semantic search, entity recognition and sentiment analysis tools. Introduced knowledge graphs based on tacit experience from employees. Ran smart expert systems. Worked with vector databases before they were a thing. And I even collaborated in research on a robotic cat expressing emotions to older people to fight loneliness.
So I know how hard it is to get AI right, and how easy it is to let it generate poor results. How humans have a nonsense detector proven by the uncanny valley. How close you need to be to passing the Turing test if you want to bring true value.
Practice, practice, practice
Don’t get me wrong. We use AI. And we will increase the use of AI.
But with the same strong design and operating principles we have in place for all product development.
We rather practice a lot, move late and strike with sniper precision than simply jump on the hype train and ruin our beautifully crafted product.
Some examples where AI works for us are Improving the speed and flow translating bespoke content and being a smart power tool when crafting great texts. We also failed countless times with for example smart expertise profiles and related document search capabilities.
I’m wildly optimistic about a wide variety of experiments and ideas we’re cooking to bring meaningful intelligent capabilities to the product that don’t suck.
Our design and operating principles
For those interested:
Our product development design principles are Playful, Personalised and Goal-driven.
And our operating principles:
- Learner experience is priority number one – Even if management of features can become tedious.
- Don’t let perfect stand in the way of good enough – Our users and data will tell us what to improve, what to polish. We may be smart, but the collective behavior of our users will beat us always.
- Adapt or die – no holy houses, everything we created becomes obsolete someday, so adapt before something is obsolete.
- Transparency over secrecy – Release early, be open about the capabilities and shortcomings of what we create, how we want to improve.
Experience our product first hand? Let’s go!
PS. Like a true dinosaur 🐊 zero AI was involved in writing this text. Slow as 💣💀👹💥 but real for sure ;).