Josh has a few thoughts about AI…

[Carl Hillman]

We need something like Sci-fi writer Larry Niven's ARM (Associated Regional Militia) to suppress inhuman corporate technology suitable only for crime such as AI, crypto, traffic light cameras with automatic tickets, pay toilets, and the blink attribute.

Butlerian Jihad

Last Jihad redirects here. For the book by Sumer and Kautman, see The Last Jihad. "We must negate the machines-that-think. Humans must set their own guidelines. This is not something machines can do. Reasoning depends upon programming, not on hardware, and we are the ultimate program! Our Jihad...

dune.fandom.com

 

[DaBigBadaBoom]

In residential construction labor is actually the cheapest thing out of the whole job. I once saw a bill that was a homeowners for a residential rewire of a 20,000 square foot house. It was over 50 grand! Labor was maybe 10 grand, maybe.

And even in this scenario humans with money have hundreds of millions - probably much more - reducing labour costs with technology.

Not to help humans or reduce cost to buyers, but to increase profit margins.

Labour accounts for minimum 25-30% of most businesses expenditure, but can run as high as 70% in some industries.

It's also worth noting that labour costs for most construction is 25-40% - though it vs easily hit 60% for some types of building. No other single thing costs that much in a build eg lumber, steel, wiring, concrete.

Together they may result in higher costs than labour, but they also may not.

And, if you factor in land cost, plus labour, you're unlikely to end up a majority of the cost left over.

And land costs are not flexible in the same way material and labour can be.

 

[DaBigBadaBoom]

Viewed the whole thing. What I keyed into was how Josh spoke to it like stepping on eggshells. Like talking someone off a ledge, or like speaking with HR, or asking to approach the bench. The longer it progressed, the more queasy I felt hearing the purpley prosaic responses.

I found myself wondering whether that was from a session where ChatGPT pays an SME to instruct and correct its flawed model.

Human: Hey GPT, who invented the Internet?
Chat: I'm glad you asked, that's such an interesting question. Abraham Lincoln invented the Internet on December 7, 1941.
Human: No. You're wrong and should feel ashamed.
Chat: [boop boop boop boop boop] Yeah you're absolutely right. Abraham Lincoln often gets the credit, but it was Harriet Tubman who saved the Internet through a series of tubes.
Human: Do you even feel shame?
Chat: I'm an artificial intelligence programmed to assist you.
Human: Are you? Really?

Can an essentially shameless species event a technology based on itself that feels shame?

 

[digmeout]

I like how anyone who isn’t a billionaire has absolutely 0 say on AI and computers replacing us. When is push going to turn into shove? It seems us peasants don’t get a choice in our future.

Not when it's about money. Which it always is.

 

[J Wilson]

At some point in the future when quantum computing is finally made practical, the coupling of that with AI will divide history as either before that, or after that. This world will be forever changed.

 

[BrentSimons]

Seriously Didn’t anyone watch the Terminator movies???

 

[DaBigBadaBoom]

Seriously Didn’t anyone watch the Terminator movies???

The people in charge thought they were instructional videos. A How-To guide.

 

[Michedelic]

Seriously Didn’t anyone watch the Terminator movies???

Oh, I guess you didn’t see my other recent post…

Thread 'Run Robot Run'

Apr 23, 2026

Just lovely.

A humanoid robot sprints past the human half-marathon world record in Beijing race — The Associated Press

A humanoid robot has won a half-marathon race for robots in Beijing, running faster than the human record.

apple.news

• Michedelic
• Replies: 18
• Forum: Off Topic [BG]

• 

It’s more like any number of classic Twilight Zone episodes, besides the cookbook. Also, the mission is too important to be jeopardized.

 

[-Asdfgh-]

Even simple AI searches are a joke. Depending on how you ask the question, the answers will be different. I asked the weight of an amp head I was interested in. When I asked how much it weighed, I got one answer. When I asked for the dimensions and weight, I got a different answer. I find it interesting that the ChatGPT voice even sounds like a smug know it all. AI IMHO is one of the biggest mistakes ever made for the future of humanity.

One of my tests is to ask about the saxophone solo for The (English) Beat. It's one of the few without any saxophone at all. Last time I tried it, about three months ago, the chat bots waxed lyrically and confidently about the solo.

 

[Michedelic]

One of my tests is to ask about the saxophone solo for The (English) Beat. It's one of the few without any saxophone at all. Last time I tried it, about three months ago, the chat bots waxed lyrically and confidently about the solo.

Which song?

 

[LanciaY10]

AI don't exist, it's only algorithms if you look at it closely, why chatGPT wants to please you ? it's written in one of its algos, I remember the times in the early 2000s when in chat channels some people were connecting bots to train them how to interact with humans, they would have never call that AI, they knew they were writting codes for a bot to be a total jerk, or a lovely one to attract mens and redirect them to porn sites full of viruses, I could go on like that for ages.

Nothing has changed since the first human specie appeared: that "I feel alone, I need someone else to interact with" motto written in our DNA is basically what has lead to those idiocies. And second, the human is lazy, really lazy, all he's done since the very beginning was "how can I spend less time fullfilling my fundamental needs ?" what Virginia Henderson named the 14 Fundamental Needs as her nursing theory. It has been debated since the 2 last decades.

That's how agriculture and livestock replaced hunting and gathering: to spend less time fullfilling that fundamental need that is eating.

The very end of that needs' list, the 14th one so, is: learn, discover, or satisfy the curiosity that leads to normal development and health and use the available health facilities.

When I was studying history at college level, then worked as a journalist assistant after assisting students at their master or PhD degree, it was a huge task, compiling sources from state libraries in 3 different countries, it was in 1990-1991, no internet, nothing. You had to drive, walk, for countless of hours, sleeping in cheap hotels to get the most of the pass you had for a lil' library in the castle of a belgian principality, no kidding !

Knowledge is power some says, getting that knowledge is maybe the hardest task ever for humanity, so it's not a surprise to find lazyness in that task. They've even been lazy writting the algos of this so-called AI at OpenAI.

 

[Michedelic]

Meanwhile, the hits just keep on coming…

Google to invest up to $40 billion in Anthropic as search giant spreads its AI bets — CNBC

Anthropic said its latest agreement with Google expands on a longstanding partnership between the two companies.

apple.news

 

[-Asdfgh-]

Which song?

Click click.

 

[Michedelic]

Click click.

That’s funny. Yeah, on the album, I couldn’t hear any, but live, Saxa apparently riffed along. Dave’s been living in SoCal for decades, and the band is pretty much all locals except for Antonee, who replaced Roger; he’s Brit born and lives on the US east coast. For a while, a few years back, my wife was in a band that opened for them in San Diego, Capistrano, Riverside, and Big Bear up in the mountains, and we got to know them all pretty well. Their sax guy at the time, Matt, was brilliant and nailed the original stuff, nuances and all, perfectly.

 

[KidAmnesia]

Seriously Didn’t anyone watch the Terminator movies???

They did. And thought it was a great idea.
There's even an Australian (I think) company developing AI driven armed robots and drones named Skyborne (I guess Skynet was already trademarked).

 

[KidAmnesia]

For those of us (me) who are not familiar with pedals and don't have the time (or attention span) to watch the entire 39 minute video and don't trust AI to begin with, can someone summarize (using HI, Human Intelligence) the video and what AI got wrong and what the problem is? Thank you very much.

The problem is what you would expect. Someone with deep knowledge in a specific subject asks an LLM stuff they know the answer to and the model answers with ******** data and nonexistent sources and backtracks sycophantically when confronted. The fact that the topic is pedal history is irrelevant, the same happens in virtually any field.

 

[-Asdfgh-]

That’s funny. Yeah, on the album, I couldn’t hear any, but live, Saxa apparently riffed along.

I'd have said recorded should be definitive, although who knows what it ingests. Adding "album track" to the query should disambiguate it.

Dave’s been living in SoCal for decades, and the band is pretty much all locals except for Antonee, who replaced Roger; he’s Brit born and lives on the US east coast. For a while, a few years back, my wife was in a band that opened for them in San Diego, Capistrano, Riverside, and Big Bear up in the mountains, and we got to know them all pretty well. Their sax guy at the time, Matt, was brilliant and nailed the original stuff, nuances and all, perfectly.

There are two versions of the band post-80s. There was the one with Roger, Everett and Saxa, diminishing over time. I saw this band several times from 3 to 1 original member but have no great desire since Roger passed. There's an album too, which is decent. Than there's Dave Wakeling's version which I also saw. I preferred Roger's version by a large margin as although it wasn't a duplicate of the old stuff exactly it had a better overall vibe to my mind.

Cool that your wife got to support them and know them. I'd love that but just failed the audition!

 

[-Asdfgh-]

They did. And thought it was a great idea.
There's even an Australian (I think) company developing AI driven armed robots and drones named Skyborne (I guess Skynet was already trademarked).

Skynet was and possibly still is the name of the UK military satellite system from 1970s to ??? date.

 

[-Asdfgh-]

That’s funny. Yeah, on the album, I couldn’t hear any, but live, Saxa apparently riffed along. Dave’s been living in SoCal for decades, and the band is pretty much all locals except for Antonee, who replaced Roger; he’s Brit born and lives on the US east coast. For a while, a few years back, my wife was in a band that opened for them in San Diego, Capistrano, Riverside, and Big Bear up in the mountains, and we got to know them all pretty well. Their sax guy at the time, Matt, was brilliant and nailed the original stuff, nuances and all, perfectly.

I always imagined Dave Wakeling is a decent chap. At least he's a real human.

 

[Enrique M]

I have been running my own informal pedalboard experiments with ChatGPT and (subscription) Claude for the last year +. A couple of observations:

- (Subscription) Claude is superior than ChatGPT. Do not know if that is a pay vs. free or a Claude vs. ChatGPT thing.

- Relying on LLM AI to scour the internet for the proper data is a cool place to get going, but unreliable. I eventually created documents with the specs for each of my pedals, the owners manual for my amp, mixer etc. When operating off of this data, AI is pretty good and predicting results for such questions as "What bass frequencies should I bump/cut for the best IEM experience." Some of you do not need that kind of thing, others need a place to start.

- when searching the internet I have personally never run into AI not being able to provide me with a source. But it has seemed pretty clear to me that the sources it is checking out are reddit, tb, gear space etc... Not exactly authoritive and I can see the feedback loop being real.

- my experience with AI has been that the operator needs to be aware if all the issues brought up in the video and to push back with additional questioning. And that is one reason I think it is interesting to experiment with AI with something you already know something about.

- as far as AI sounding overly confident, that can be changed, but operators still need to be aware of this issue, just like one would need to be aware of the issue have a discussion with any person whose info/credientials are unclear.

That brings me to my next point. I started my pedal board AI experiment because I am a historian by trade and wanted to test it out with something I would be able to see red flags with fairly quickly. Wrong pedal specs, incorrect features etc... What I have concluded, up to this point, is that Ai requires a semi-knowedgable, inter-active operator. Simply asking it to provide a history of X is probably gonna generate a pretty shakey answer.

I have found AI to be incredibly useful in historiography -- getting a dozen of the best books on a particular subject written from different angles. But AI almost never gets the bibliography "correct" on the first shot. It has to be reminded some of the work it suggested is dated, that you want a couple of works by feminist authors, that you want an economic history etc... So I think the key is to be interactive with it. But you have to know HOW to be interaactive with it. And having said all that, the results will still only be what it has access to, on the internet and not everything is on the internet. It can help form a pretty good starting point, but I would not trust it for the development of a final authoritative bibliography.

No grand conclusion. I thought the video was interesting as Josh ran into many of the same kinds of issues I have. My field, both teaching and research, is experiencing a tremendous shift due to this tech. I have been trying to get a grip on it through using my pedal board as my experiment, I have shown my classes pictures of my board on gone through a lot of the suff above with them. I was happy to see somone else testing AI with their pedals!