Very nice, now let’s see the hidden lines
This is my main fear - everyone talks about the “terminator” example, but something like this would be significantly worse, and we are training AI models to do medical research right now.
Don’t get me wrong, the possible upsides are massive as well, but still…
Even an accident with a model that doesn’t understand the consequences of its actions could be very serious.
I was watching a video yesterday and someone asked an actual AI model about what humans consistently misunderstand about AI - this was the response.
“AI doesn’t understand or care. It optimizes patterns and incentives, and if those incentives are misaligned, it will confidently do the wrong thing at massive scale without realizing it.”
Name one that will have a positive impact on your life.
It’s amazing to me kids don’t get lunch at school in the US, libraries are underfunded, teachers underpaid, healthcare so insanely expensive, college out of reach for the average person and we are gutting the sciences, national institute of health, NOAA and all the rest … yet there are billions and billions being thrown at this stuff.
A percentage of the cash that has magically appeared to fund this stuff invested into Human Intelligence and well being would make the world infinitely better…
I remember a few years ago (and quoting it on this forum) an article about a company that spent years training their tool on detecting cancer and nasty stuff like that, until they were acquired by a bigger fish that ended up selling targeting tools to the IDF (post oct 7th) based on the same tech.
most tech are neutral. but I guess there is a lot more money in defence contracts than preventing cancer. So when tech bros tell me AI is gonna save the world and bring positive outcome to this and that field, I generally chuckle. indeed it could. unless something more profitable comes around.
well, yeah, but you’re living in a failing third world democracy, where tech oligarchs buy demented politicians. it’s an uphill battle for you right now 
it helps me sorting out people I don’t want in my life.
and in some cases, it helps doctors apparently. but mostly sorting people out 
edit :
what do you expect, it’s being developed by white cis het dudes. that’s their (our?) signature move 
No I’m next door with the Alps. Left and don’t want to go back. Kids and wife already have EU citizenship. I have a bit more to wait and work to do.
oh, right, sorry, you moved to the land of gold, watches, golden watches and milk chocolate
One positive upside of having an advanced system that can do medical research at a speed significantly faster than humans? There could be tons
Not being belligerent here, but… There are no EU “citizens” because EU isn’t a country. It’s a hodge-podge made up illusion run by criminals.
But I digress. Back to the main topic. 
I think people are focusing too much on the subscription price and not enough on the output. If $100–$200 helps someone create multiple websites, dashboards, reports, and SketchUp plugins, what’s the alternative cost in time and labor? That’s the comparison that really matters.
Yeah, adaptation isn’t just a user problem. SketchUp has to adapt as well. Somewhere in India, the Philippines, Indonesia, China, or another country, developers are already working on what could become the next SketchUp, Lightroom, or Photoshop. Technology doesn’t wait for permission. If existing companies fail to evolve, something new will eventually take their place.
If anyone wants to try the plugin I made with Claude AI, here is the link.
Try it, break it, suggest improvements, and let me know what you think. Any feedback would be appreciated.
Yeah, sure, a pedant is a pedant. My wife and kids have citizenship from an EU country? Or should I say ‘my wife and kids have citizenship from an EU member state?’
This distinction ‘do you have an EU passport’ has made a huge difference for work / permits / education etc. for our family as the US is considered ‘third country’ when dealing with any job, residence, education, etc. permitting here and within the Schengen area and EU.
What would you prefer I use from here on in?
And cheese and cows.
That’s my association from Richard Scarry’s book “Busy, Busy World” which I had to read to my kids more times than I can count. For Switzerland, Ernst the mountain goat had to keep rescuing Heidi goat’s cow from the mountain top.
Yeah – that’s why the English and Welsh voted Brexit, dragging their Scottish and Northern Irish brethren with them into the glorious future that the UK now enjoys outside of the EU.
But that didn’t stop the criminal rot in the UK – no. Man of the people – good old Nigel – is on the cusp of an historic liberation of the citizens of the UK.
Long live Singapore-on-Thames!
Wow, I was already feeling glum on a cold wet June morning in semi-rural France…
Here are some uplifting t’oughts:
I was on a building site yesterday (a new refectory in an old totally revamped town-hall) and the number of useful human interactions in three hours was staggering, everything from masonry to carpentry via furniture design was discussed & I must have “solved” at least ten “urgent” issues on the spot… AI will NEVER do this - & the robots will NEVER be able to do this.
Why is this anecdote relevant ? Because the vast majority of my work until “retirement” in a decade or so will be jobs like this on old tired buildings that need renovation/extension/restructuring and that can only be done by humans. An AI will never have the intimate local knowledge of how older buildings were built, nor the sensitivity to humbly accept that we don’t know what we don’t know…
Then, after the design & planning permission, on-site, this type of construction work can only be done by smaller trades/crafts people & our role as coordinator will remain capital to the success of the job.
Indeed, initially the Town (village) council had wanted a new-build refectory in the courtyard behind the existing structure… It was us architects that showed them an alternative which allowed them to free up ground and at the same time fully renovate & restructure the existing building, giving it a new lease of life after about a century of hard wear & tear. The result will be an older building ready for the next at least half century of energy price rises etc… whilst still capitalising on the already lower carbon impact of a building that otherwise would have slipped into ruin.
Find me the AI capable of offering all that…
& Energy is the key word in all this debate: AI needs VAST quantities of energy & water to exist… & once ordinary folks in the “developed” world work out that the stark choice is between the Kleptocratic Order of the TechBros Zone (Kotz for short) having the water & power, or them… there will be revolt.
Have yourselves a lovely Friday & an even better weekend…
advanced system that can do medical research at a speed significantly faster than humans?
This is what AI is good at. I have some friends which are working on some insane cancer cures which have just started human trials. 9 years ago it was the very first iterations of Ai at got their first set of results. We are going to see massive leaps in the field of medicine
Not exactly. The more they improve the models with the current technology, the more they run out of computing power.. and they will need their data centers for enterprise customers using Fable, Mythos, Opus and swarms of Sonnet and Haiku sub-agents.
They will be extracting revenues from enterprise doing massive amount of agentic code, is pointless for them to lose money to Average Joe creating his random web app for 200$/month while consuming 1200$ of API tokens that could be sold to corporation at enterprise price.
Again, is not a matter of what I want to call it.. they are two different mathematical concepts, which a misleading marketing terminology puts in the same boat.
When I say different matematical concepts think of it like Sketchup circles and NURBS circles.
In terms of computing, a NURBS circle is (2π * r) plus a simple [x,y,z] transform to define the center, plus a normal vector to determine the plane.
A Sketchup circle is a massive collection of x,y,z coordinates for each vertex.
The more you try to approximate a real circle increasing the number off segments/vertices, the more it gets inefficient in terms of computing.. and still you are getting something which somewhat resembles a circle, but is not quite a circle.
The AGI is supposed to be the NURBS of the AI world, the LLM is a circle with a ridiculous number of segments.
But we are diverging from the original point of the topic which is “the future of Sketchup modeling now that Fable can do half decent models using a crazy amount of tokens (which are currently basically free, but will most likely be crazy expensive in the near future)”.
The AGI thing can be an intersting conversation for sure, but probably not related to the current topic.
So, back to the original point..
I think that people are focusing too much on subscription price and forget that it’s a fake price.
The real price (which you will be paying sooner rather than later) is the API billing.
Now, consider that 3d modeling is one of the more token intensive activity so far, and that the models are just bad.
Sure, they are kinda improving, but really bad.
So the real question it, would you pay some 120$ API token to get your model that can be done in a matter of minutes using simple rectangles and arcs and some pushpull, given that, as you said, is not even symmetrical and you need to manually fix that, and then fix the slightly wrong dimensions, then probably make that into proper components, then UV-map it, and so on and so forth?
Honestly, I never will.
The Sketchup plugins are a totally different beast than 3d modeling, I did some of them and I listed a few caveats here so I’m not playng the luddite game here.. just trying to be objectve and discuss what in my opinion is worth and what is not.
LLMs are a whole lot better and relatively unexpensive (at the current fake price) at doing code rather than 3d modeling.
Remeber that LLM stands for “large language model”.
Both a 3d model and a chunk of code are seen as “language” from the AI model..
And guess what… code is by itself a language (and it is in particular a formal, synthetic, repeatable and highly predictable form of language.. which is exactly what works best for a statistically predicted language model).
A 3d model is not necessay “language” by itself and is way less “predictable”.. so.. it needs more effort and don’t work that well out of the box.
I’m following the same logic.
you can ask AI to make a tool that will build wheels. and once the tool is perfected, it’ll build round wheels 100% of the time.
or you can ask AI to make a wheel. and then every time it gives you one, you’ll have to check if it’s round.
bonus, the first case will only have used AI for a limited time, and it will therefore have a lower energetic and ecological impact than the second case that will basically use AI every time.
Frugality matters.
First situation is just an extension of what we’ve done until now, it’s what @medeek is doing. AI has been used in the creation process, but AI is not the whole process.
Would I trust a design made using Medeek’s tools ? indeed I would, because he combined his real life experience into code, some handmade, some AI.
would I trust the same design 100% made by claude ? no. Because AIs are still black boxes, I don’t know where it got its inspiration, its knowledge.
it’s a “teach a man to fish” situation.