A podcast is brought to life from editing and the tools to perform these tasks, help instrumentally. Podcasts are not only unedited clips of people talking to a mic about a topic and very few are but editing helps introduce and explain the topic fluidly while keeping the tempo steady. One tool that could help me substantially is the application called “Descript” (introduced by my teachers to me), a collaborative audio/video editor that works like a doc. 

Including a wide variety of tools, it is the AI that catches my eye. Having the ability to make a transcript that can reflect the changes made on the document to the recording. Doing the most tedious task of editing sheds light on the possibilities of this software. Most impressively you can delete, add or edit words that sound like the original audio, though it does cost a lot for this feature. 

To my current understanding, there are 2 main ways to train your AI, have information put into the neural network to learn or creating another AI to try and compete against your first one. Mimicking the human brain, neural networks are a series of algorithms that tries to recognize underlying relationships in a set of data

The benefit of having 2 AI’s is that for small startups it allows lots of training in a short time with both getting better as they improve. It is almost reflecting real life, how one creature will evolve to survive better and the other creature will evolve to remove those advantages.

One instance of this are superbugs and antibiotic resistance:

Antibiotics act as selective pressure. They kill sensitive bacteria and leave behind the resistant bacteria. It then creates a population that is immune to that antibiotic making it ineffective 

An interesting video that I encourage everyone to watch, that goes over this AI learning in a fun game of “Hide and Seek” with moveable obstacles. The AI have a competition to overcome obstacles that the other team proposes in which they have to come up with new ideas and master techniques.

AI needs a lot of training, perimeters and information to succeed. If Descript was only done by one AI then it probably trained with hundreds of thousands of Audio clips to be able to mimic, detect and delete specific words.

The free version only allows so much when using the software and limits the use of essential tool. I have tested out the free trail of pro with my current podcast I am working on and it helped dramatically in my editing process. Removing filler words and shortening pauses in a short period of time I have probably saved four hours of work if I were to do it in GarageBand