I'm still not sure about AI

I’m really not sure what the best approach to AI is, but the more I see ads for AI powered this and that, or wade through AI based articles about “16 cool things to make your business better”, the less interested I feel.

It’s giving me a strong Crypto / NFT feeling, at least in the way the concept has been monetised and promoted across outlets like the Artist formerly Known as Twitter.

I still shake my head in amazement at the people who poured their money into things like a Bored Ape NFT and I’m privately convinced that somewhere out there, there’s a pretty clear link between some of the more outrageous NFT stories and money laundering.

Too Much AI Noise

So, that general noise and fuss around AI, plus the approximately 71 Million News articles that pop up on a quick, old fashioned Google Search just gives me a general “meh” feeling about it.

Oh and please, stop doing the “I asked AI to write my article / sing me a song / murder someone” type thing too, it’s incredibly tedious.

I do have to consider that I’m approaching Grumpy Old Man status and that maybe it’s just a lot of newfangled nonsense, like the motor car and dial up internet, in which case I’m sorry and also, resolutely not sorry.

Chat GPT Potential

I’ve played around with Chat GPT and I’ve even run some of these newsletters through it to get an assessment of the writing style. Then it started to rewrite the pieces for me, and however bad my writing may be, it somehow made things worse. It became even more generic and full of platitudes, and it annoyed me even further.

I know of a radio presenter who gets AI / Chat GPT to write interview questions for them and it’s been pretty effective. I’ve seen some good use of it for search and for building templates, the kind of stuff that you could do yourself on Google but would take much longer.

Some of the applications in medicine look really interesting too, with models able to examine millions of possible gene pairings and identify which ones could case potential hereditary diseases and the application of AI to things like cancer scans could prove to be hugely beneficial in detecting anomalies early.

AI in Audio

When it comes to audio, I’ve been interested in the potential for voice cloning for some time, at the RadioDays back in 2019 in Lausanne I think, I saw an interesting presentation from a company I think was called Lyrebird, who’ve since been acquired by Descript.

The notion of being able to clone voices and play around with audio production was quite appealing, but it was clear back in 2019 that it was still very early days.

In the Daft Doris / Smart 7 Lab we’ve been experimenting with the latest versions of voice cloning which use AI to improve and polish the audio. We also found a tool that tidies up noisy audio by filling in the gaps in say a press conference in a hallway or one of Joe Biden’s Helicopter pad chats.

We’ve created a pretty realistic version of Jamie East, the weirdest thing about it, is that the clone actually takes breaths. That’s a bit of a head melter, but it does make it sound more authentic.

And I know that there are local stations and services out there that are using AI for scripting content and recording it, based on reliable sources.

Hard to Make the Leap

I still can’t quite make the leap though. We’ve piloted an AI voiced Smart 7 but there’s no way we could use AI to script the material, it’s just not sharp enough yet, or good at summarising stories in the way we spend so long each day doing.

I think it’s partly that the art of writing, or crafting something is too important to outsource, for me at least. I’d rather battle my way through the process of finding the right word or phrase and hope that when I’ve finished writing it, I can read it back and see that I’ve captured what I wanted on the page.

I’m feeling the same way about AI for voice, it’s potentially a useful tool, I’ve been thinking about putting some AI voiced links onto the radio station’s overnight shift and playing around to see how convincing it can be. But it doesn’t solve the talent problem that radio has…

We need to find more Talent

We need to be spending more time and effort finding new talent, or trying to recruit some of the exciting talent that is in podcasting back into radio. Just putting on an AI voice and an automated script won’t really help anyone’s talent or creativity.

That’s the real problem with radio at the moment, with a war, a pandemic, a recession and big groups consolidating, there’s less and less space for small, creative and interesting ideas to emerge. There’s less and less places for that interesting, crazy double act to really develop their skill set and become something extraordinary.

I think talent is something that AI can’t replace, it can copy it, or it can imitate it, but the rare, raw talent that emerges from the blood, sweat and tears of overnights, or weekend breakfast, or doing vox pops in the cold is not something that can be simulated. And it’s still that talent that draws audience in, that creates a sense of community and shared enjoyment.

So, I’ll keep trying with AI but I’m not going all in. Not yet.

This has been the 15^th official edition of the RAudio Newsletter

Just to recap, each week I’ll be taking a look at big stories in radio, podcasting and audio.

Any feedback, questions or potential topics are welcome – you can get me on Linkedin https://www.linkedin.com/in/liamthompsonconsulting/ or on Twitter @Maxliam