Work AIn progress

When prosthetics is transcribed as prostitutes, you know we are still early and have a long journey ahead.

Earlier this week, I was in a call with a charity that refurbishes used prosthetic legs and sends them to a country in Africa to help people there who have lost limbs unlock mobility. They do impressive work, and we decided to direct part of our fundraising to their organisation. You can see a video of what they do in the Poésie de la vie section below.

What was a bit distracting during the call, was the transcription provided by the video call tool. This tool uses AI to convert what is being said into text. A very useful tool, if it works well. This one clearly has a way to go to become reliably useful, as some of the conversions were still very, well, wrong. Some were comical, and others were troubling. When prosthetics is consistently transcribed as prostitutes, we can conclude that this tool is proving it can be useful, but needs to improve its accuracy.

AI has proved how useful it can be, but we still need to adapt. Whether it are misinterpretations in transcriptions, hallucinations, or overvaluations of misinformation, AI still needs us very much to verify whether what has been produced is actually usable. This, I believe, underscores the need for learning methodology of not only how AI works and is trained, but also of how to do what you want the AI to do. And a thorough subject matter knowledge to spot the hallucinations.

We are on our way, but not there yet.

What's the funniest hallucination AI has presented you?

Poésie de la vie

Watch the story of Lamin, a welder who could continue his work thanks to a donated prosthetic leg.

Play Video

Arjan Tupan is writing poetry, podcasting and incubating empathy
Entrepreneur, dad, European, storyteller, poet. Curious, too. I tell stories on different platforms and using a variety of tools. Sometimes I explain how I do stuff. Most of my content is free of charge.