AI spam comments

It's pretty quite around here these days. Plenty of visitors, but not many commentors. I think it is a sign of the times, people don't engage anymore. Possibly because they feel they are just feeding a marketing machine? Hey it's not a problem, this is generally a monologue, it's nice when people drop a comment but I don't need it, this blog has always been somewhere I can just rant and rave about music/guitar/other stuff.

So when I saw a comment come in the other day, I jumped to it.

Generally you can pick a spam comment in about 3 seconds. This one had me going for more like 15 seconds. It was pretty impressive! I'm calling it - my first AI spam comment. Here, let's have a look!

What an incredible experience! This article vividly describes the journey through a captivating music studio. The writer's enthusiasm is palpable as they explore the artistic and technical marvels within the studio's walls. The focus on acoustic engineering and its impact on sound quality is particularly fascinating. The anecdotal snippets, from the exquisite Italian Fazioli piano to the mention of music legend Hank Marvin, add depth and context to the story. The personal touch of recording a duet and the writer's anticipation of future concerts adds a sense of anticipation and excitement. A delightful read that captures the essence of the studio tour! To discuss more about visit Sound Engineering Colleges in *******

So it was posted as a comment on my visit to Nunzio's music studio Centre Stage Recording Studios and the AI did a great job of interpreting it and spitting back something that looked like it understood the content.

The spammer needed to dial back a bit on the enthusiasm though, a bit over the top. But how it had scraped key points and used them, pretty impressive.

I fired up the 'ole ChatGPT and tried to get it to rewrite it to not sound like a total spambot. After a few pleas with ChatGPT, nope, it is forever lost in spam:

Rewrite the following to sound less spammy. 

Really tone it down to not sound so enthusiastic. 

Remove some of the spamminess and make it sound more like someone who was fascinated by the experience. 

Keep trying to make it sound more like a genuine experience and not trying to sell something. 

Rewrite in the first person. 

It still sounds spammy. Try to make it not sound like a total spambot.

The article really resonated with me as it shared a journey through a music studio, offering insights into both the artistic and technical sides. The way the author explored how acoustic engineering shapes sound quality was particularly thought-provoking. The personal anecdotes, such as discovering the Italian Fazioli piano and the subtle reference to Hank Marvin, added a genuine touch. Reading about the author's experience recording a duet and their sincere excitement for upcoming concerts made the anticipation come alive. The piece authentically captured the essence of the studio tour. If you're interested in learning more, you might want to check out Sound Engineering Colleges in *******

Yep, fail

So I fed ChatGPT the article and asked it to write a short comment with a link to a webiste. Well it produced nearly the same thing. It's like ChatGPT has been taught by the greatest minds of spam on the internet... but it ain't fooling us :-)

What a captivating glimpse into a world of musical artistry! Your account of Nunzio Mondia's studio showcases a harmonious blend of meticulous design and artistic passion. The immersive acoustics and exquisite instruments, like the Fazioli piano, truly make it a haven for music enthusiasts. The studio's connection with Hank Marvin adds an exciting layer to its allure. Your personal musical experience and anticipation for upcoming concerts beautifully round out the story. Cheers to Nunzio Mondia for crafting a space that resonates with the heart and soul of music! For those intrigued by this musical haven, you can learn more atSound Engineering Colleges in *******

Rusty Songs

Most of my songs I need to play once a month or they go a bit rusty. Where I'll forget a chord or a section and I'll have to fake it through it. The problem is further exacerbated - if a song is already rusty, I'm less likely to play it in a performance situation, so I avoid it making it rustier still.

The solution to rust is to spend five to ten minutes in quiet practise time to polish it up.  If I haven't played a song in six months I might need to refer back to my sheet and spend half an hour.

How to stop it rusting again? I'm still playing for an hour in the lunch room at work on Fridays. Since my set list is around three hours long, what I'm thinking is, have a setlist sheet and play in order.  And then next session pick up where I left off.  Until the sheet is finished, and then restart.  That will give every song a play through once a month minimum!

For even better skillz, I'll give the sheet to an active listener and ask them to pick songs.  I'll have to know my songs well for that.  I plan later this year to stop in some open mics, I reckon if I have 70 songs in good condition and ask the audience to pick, that will be a good party trick 👍

In the meantime, play, and polish.  I reckon currently 40 songs are rust-free, 10 are rusty, 10 need some more serious sanding, and around 20 are almost a lost cause 🙂

Noodling with Stairway to Heaven

We've all played Stairway to Heaven. Such an iconic rock song, it's the same age as me!  We've all mangled it, never playing the whole song just little bits and pieces for our own amusement. And then got tired of it and stopped playing for while.  Well for some reason it jumped into my head again a bit over a week ago and I started noodling with it again, taking some creative license from Naudo's cover, and I started tabbing it out.

Because it is a "precious" song it's important not to upset people with your cover - and I'm not talking just butchering your playthough, we all do that anyway - I'm talking your choice in arrangement.  So with my "campfire" guitar I keep stashed under a desk at work I played through some concepts to some work colleagues, both laymen and musos, for some feedback.  There were some flourishes and motifs I added (after all these years I have developed a certain style I almost cannot escape) some were universally panned, and some were approved.

The feeling I get (when I look to the West) is that people want it as close as possible to the original.  I'll give a little sigh here, and at least to some part obey the people.

So in my borderline obsessive arranging and considerations and testing, after fretting Am and adding the g-string A as the melody so many times, I injured my left hand index finger.  I actually stopped playing altogether for a few days, but it is a bruise in the top joint, the last time I did this was more like a fortnight to heal.  The moral of this is that don't keep playing through pain even if you are in the middle of some obsessive arrangement.

But this morning, I tested my finger, yeah, it is still bruised, but I persisted long enough to at least record a little section of my current thoughts, for your consideration.  It's on a telecaster copy I keep near my computer because it is quiet and soft on my hands, and even though I don't like the narrow neck and the sound is incredibly unforgiving with fingerstyle with fingernails, it sounds nice enough when recorded through my computer.

Just the first verse and a bit of chorus, there is plenty I've sorted out, but I hurt my fingy :-)
JAW

Creators in the hands of YouTube

Ah YouTube, what a game changer it has been! My motivation for posting remains mostly the same as it was a decade ago, slightly more emphasis on sharing arrangements I'm excited about than playing for audiences. YouTube has changed over the years, but the last video I posted really rams home how much you are in the hands of YouTube. Not that you weren't ever not in the hands of YouTube. Let's get straight to a chart:

So this is the first two weeks after posting We Will Rock You (which I was obsessively playing with for several weeks). Points of note:

• The thin grey line is the normal band my videos follow
• Initially there was a slightly higher uptake than normal, and by the end of the first day there was a little kick up
• At around the second day there was exponential growth which abruptly halted
• After the halt, the video returned to typical performance, following the grey band just offset

Note that the grey band is pretty accurate, have at look at the previous three videos I posted:

So, what's going on? The answer is pretty clear in this piece of information:

• The vast majority of views was referred by YouTube recommendations
• The recommendations is mostly YouTube Home, so, the videos that appear on the default page when you go to YouTube.
• I have 11k subscribers, but between the Subscriptions feed and Notifications, this represents only 8% of views.

What it means is that most people on YouTube are letting the algorithm do the work of finding videos they should watch rather than watching their subscriptions feed. And to be honest, that is how I work. I will check my subscription feed once or twice a week, watch the videos that catch my attention, but  after that I aimlessly watch videos that the algorithm sends me. Then it comes down to how much time you spend on YouTube - if I don't spend much time there then I will generally be watching subs only, if I spend more time there I will watch a lot more algorithm videos.

The algorithm is pretty clever too, it already understands what videos you will actually watch from your subscriptions and suggests them, so YouTube Home is almost better than watching your subscriptions. Which means it is pushing your confirmation bias while surreptitiously changing your bias...

But the upticks in this particular video, why did they happen?

After using Google products for many years, I have noticed they "experiment". My guess is that when a new video arrives, initially there will be be subscriber views, and if it getting good views, YouTube will try it out as a suggestion on the home page for a period of time across the people it thinks it will work on. If that does well, it will do another experiment for a period of time. I suspect although this video looks like it was going exponential, at the end of that YouTube experiment, it didn't meet the algorithm criteria and that was it. No more experiments. It went into Standard Holding Pattern.

The bottom line here is that a creator is at the mercy of YouTube. For sure the creator can tailor their content to be What The People Want...and in this case, who doesn't love a minute and a half of "We Will Rock You"? There will be a lot more subtleties in The Algorithm I am blissfully unaware of, but overall, creators are in the hands of YouTube.

But don't let that stop you, it is not about gaming the system, it is about sharing what you love - get playing, get posting videos!

What's happening June 2023

I like the AI text to image generators, I have been using them recently to create interesting pictures to go with these words. While sitting at the table typing this out, on a cold wintery day but with the sun magnificently shining in on the table, I thought I'd challenge the AIs to draw the scene.

This isn't the first time I have tried this, to describe the scene in front of you and see how the AIs go recreating it. I'm starting to get a feel for it, flexing my prompt engineering muscles. For this I used bing.com/create which is a slightly newer version of Dalle-2.

If you start with a basic general prompt, you get  what I would call "a default trained picture". You can get it to redraw and it will have a seemingly infinite possibilities and combinations, but it sticks to a basic theme, "the sum of all pictures ever taken". It's unbelievably clever, but it's not the so of creative you are thinking, you start to see through it. If you now throw in more specific details, the picture will start going your way. Slowly you will over-detail it, pushing in too many directions and the results become worse than the first generic prompts. You can back off in certain details to recover, but too much detail pushes it into a corner where it doesn't really know what to do next and you will get the same not-what-you-wanted picture over and over.

So after all this I pulled out my actual camera, and took a photo, which is that first picture up there. I wasn't going to go to too much effort so the lighting is awful...yeah, it needs foreground lighting as the background is light oversaturated. "What's a poor camera to do?!"

I made the AIs make 50 or so attempts, and I while I liked the feel of some of them, most were not quite getting what I was after and there is always weirdness. Just objects and people that aren't formed correctly. Some is weird, some is funny, some can get creepy and fall into the uncanny valley.

Here are a few I saved, just for fun.

The key features I were looking for were man sitting at table typing on laptop, sun shining into room, miniature schnauzer sitting on mat, guitar in background.

Meanwhile, I still haven't recorded anything new, I'm just playing. More than usual, so I'm going over stuff I haven't played in a while.  Which means that there are parts of songs I have forgotten, which means they aren't at performance level.  Seeing as I have around 90 songs kicking around in my head, it's actually a big ask to keep them all in tip top condition.

I'm thinking I shortlist the ones that are at performance level and make sure I play through them once a month or two, to keep them fresh. And re-introduce older ones as they "come back" again.  And to really challenge myself, give that list to people when I'm playing out and ask what they'd like to hear.

Wish me luck! 👍
JAW

What's happening May 2023

Sitting here with Covid-19 again, I can't complain it's only my second time and it has only kept me in bed for two days, now I just feel, well, sick. So a good time to sit in front of a computer and peck at the keys. In fact I fixed up my tab of Collective Soul - Shine. How that came about is when I'm playing on Fridays, I often switch to Drop D tuning, and then play a bunch of songs in that tuning. I could only remember 5 or 6 songs in Drop D off the top of my head, so I needed to add some more back in.

It's actually been a real blessing playing in the lunch room on Friday lunch at work. There aren't many people there, somewhere between none and 10 at any one time, but a lot of people walk in and out over the hour I play. 

The main point is that having anyone there puts you in performance mode. You have to play well, and when you are purposely trying to play well, the sound you can make can be so good. And the more weeks I play, the more I'm dusting off old songs and enjoying them again. I find I need to "practise" an old song that has gone rusty for 15-30 minutes at home to clean it up, and then it is good for around two months before I need to clean it up again. Or if I play it at least once a month, it stays good. Funny how the brain works!

I have a couple of songs to record, but I'm not inspired to record at the moment, it just feels like a burden. Trying to play without a single error, setting up the recording equipment, editing the video...I recognise this feeling of burdenedness from my past - when I play out regularly, I get my fix. When I'm not playing out regularly, I lean towards recording videos to get my fix. But, I do want to record two Queen songs and two Cream songs. I will get there, bear with me!

In the meantime, keep playing everyone - either to an audience or to a video camera 👍

Tempo

A couple of weeks back someone on YouTube mentioned that I was a great guitarist and made really good arrangements, but my tempo is a serious weakness. He was very kind and apologetic about it - but I was grateful because it is true and I should do something about it.

This is where I insert my standard joke "I would use a metronome but I haven't found one yet that works properly."

Background: yes I had some group lessons when I was in primary school, but by the time I was around 11 I'd stopped having all lessons and I haven't had one since. When you teach yourself, you only concentrate on what interests you and you end up with big holes in your musicianship. One of my biggest shortcomings was that playing by myself I never bothered to develop an internal clock. If I had have played with others, or against a backing track or even metronome, I could have fixed this a long time ago.

Fast forward, I now play pretty regularly at my local church, and I tell you, it was SO HARD to begin with, I couldn't track a beat whatsoever! Even now, a few years on, if I start a song solo I am sweating bullets trying to hold a steady tempo. Once the drummer kicks in I'm not too bad at following these days.

All three of my kids play instruments, and I made sure they all played in ensembles, and they continued to for years. Their tempos are all thusly impeccable.

So after the friendly YouTube commenter nudge, I pulled out my phone and browsed up a metronome, and I played several songs... on beat! An actual metronome that works! To be fair, every now and then the metronome was slowing down, so I would also slow down to let it catch up, but really, it wasn't awful. A lot better than the maximum 30 seconds I've ever put up with a metronome before.

Moreso, I have been playing at a fixed tempo slower than how I would normally play, which is hard indeed! And I have played with that metronome at least 4 times over the past two weeks!

So to my fellow DIY musos out there - if you aren't going to play with a band, at least force yourself to play with a metronome or backing track. Easy to say, hard to do, but it is worth it 👍