bitL

Sorry, the message was probably obfuscated in my original post. I didn't mean that generative AI would replace art producers, just individual tools and it didn't occur to me at the time of typing it that most people would associate it with replacing art producers (as the tooling aspect was "obvious" to me).

I presume that the technician had at least one PhD. I had a colleague who worked at Los Alamos with a top PhD and he was basically driving around placing and collecting data from some sensors all day long.

You could be the former CEO of Adobe with a PhD in ML concentrating in image generation and a wing dedicated to your artwork at the Met and still be too out of touch with current image manipulation tooling to make useful predictions about it.

This was a good one, I love it!

My SWengineering background tells me that many of the tools Adobe has are tedious to replace but doable by many algorithmically gifted folks. Some need some investment to get to pro level like color calibration and correction. Some need camera manufacturer support like initial sensor RAW processing. The ones that were causing awe like their famous patented content aware fill based on complicated differential equations was out of reach for most. So were their precise selection/masking tools and a few more. Now we suddenly can select objects/background/hair fairly easily and reliably using ClipSeg or Segment Anything, removing the masking obstacle. To fill/replace content, we simply select the area and let stable diffusion hallucinate options until we are happy. To simulate puppet tool, we can use ControlNet with stable diffusion though implementing ARAP is also fairly easy. So a dedicated company that wants to get to the parity with Adobe in their most advanced tooling suddenly has a clear road ahead. If they improve UX by e.g. voice or gesture control (plenty of places where Adobe tools are difficult to use for no reason) and do some decent image format compatibility, they can really make a dent in Adobe's market share.

This might sound self-aggrandizing, but given infinite time and energy I alone could replicate most of the CS6 functionality of PS at the same or better quality (and I did create some powerful tools for one of their competitors) and know a few folks capable of the same.

Thanks for ad hominem. I never said 'Adobe became "irrelevant" "years ago"'. I also didn't say anyone can replace Adobe easily, but outlined a path how a motivated company can replace their offering. If somebody offers better UX (and there are many options to do better than PS), lowering the switching costs in required training, then it can happen fairly quickly. Beside being a former pro photo/cinematographer, I also worked for one Adobe competitor and invented a bunch of new geometric and image processing algorithms, then went on to study Deep Learning at Stanford, so maybe I know what I am talking about, or maybe I am just a fallacious idiot who has no clue.

My business background tells me Adobe Photoshop is going to be commoditized because their main advantage in retouching, content filling and masking can be now done by any 10-year old with a beefier GPU at the same or better quality. Even ridiculously bad GIMP can now get the same state-of-art tools Adobe has so inevitably its value proposition will be only appealing to existing "legacy" customers and dropping everywhere else.

You are missing the point. It's fairly easy to replace Photoshop's layering, color correction, filters etc. by a regular software engineering. Where Adobe had extra edge was their retouching, masking and content filling ability. That is now going to be possible to incorporate anywhere by stable diffusion et al. Now Adobe will still have foothold in "legacy" projects with proprietary formats but all the new entrants will have no need to use it. Suddenly folks in Affinity/Serif can add those missing features and continue carving out more from Adobe's market share the way Japan went from crappy manufacturing in the 60s to bleeding edge tech in 80s.

Photoshop will be eaten by generative AI in the next few years. Lightroom is still the best, Illustrator is in a duopoly with CorelDRAW. I used to be a pro photo/cinematographer and I can replace most of my Photoshop needs with stable diffusion + ControlNet + ClipSeg. DaVinci with plugins made After Effect irrelevant for my needs as well. Certainly Adobe still has momentum but their future after they switched from engineering/art company to rentier monopoly company doesn't look that rosy.

From Wikipedia about ARM:

"The company was founded in November 1990 as Advanced RISC Machines Ltd and structured as a joint venture between Acorn Computers, Apple, and VLSI Technology."

All latest node tech from TSMC was co-financed by Apple who in turn had the first-user right.

That's in theory; in practice countries often look the other way or delay actions when they see fit. I guess Ireland is running out of money they planned to get and are now trigger happy on Meta.

There are plugins for DaVinci that can obliterate the need to ever touch AE though it's not out of the box.

You can kinda view ARM and TSMC as their babies. They founded ARM and financed all the node tech in TSMC that made TSMC competitive.

As somebody who went through the MBA education for defensive reasons (i.e. acquiring skills to protect me from managerial BS), I would say you can extend it to most business jobs excluding accountants and quants.

DaVinci Resolve is eating their lunch in cinematography, nobody cares about Premiere or AfterEffects anymore.

I still own the CS6 suite, the last one before Adobe went mad with subscriptions. Even back then they tried to switch CS6 users to CC users by accepting some ridiculous license agreement that would lead to losing access to CS6. Adobe became truly horrible when the current CEO took the reins.

How are they going to inject ads there though? Moreover, the cost per ad will be massive given how expensive the inference is. And a self-hosted Vicuna will likely behave similarly, potentially rendering the whole search experience pointless in the future. Anyone can incorporate a bunch of DDG results to a self-hosted good-enough LLM as well and make a basic desktop app for it, bypassing Google completely.

I mean that's precisely how Ireland raised their GDP, they took all US companies and offered them low tax deals, making the rest of EU upset. Now they probably burned through all that money and are trying to outsmart those corps to get more money from them. Tax offices always project the same or higher income for the next year and when they get hit, they need to squeeze it out from somewhere.

Ireland wants to lose money US corps pay them in order to bypass higher EU taxes I guess. I am wondering if they did some game theory scenario about what fine Meta can absorb before it's easier for them to relocate elsewhere with laxer standards that would love some more tax income.

Former colleagues told me about a guy who went fishing on a lake and got an idea how to solve some distributed system problem there, then came back, wrote a system using water level, bubbles and fish names and called it an aquarium or something. It worked well for the next decade.

Some schools are doing stackable degrees. You take a block of a few classes (e.g. 3-4 classes), get a graduate certificate for that area and once you get a few of them, you can convert them to a graduate degree. It might be trickier for undergrad as that's where the grit is being created.

We'll see I guess. All that GPT models showed us so far is that we humans aren't as intelligent as we thought. We might be forced to redefine what makes us human as intelligence might not be our most distinguishing trait fairly soon.

100x is an assumption. LLMs can hit a ceiling (already prohibitively expensive to train), GPUs can hit a ceiling (we are not far from silicon semiconductor limits) and we might end up with a decade of "almost there, but not yet".

Recently so many unrelated sites pop up a tiny Google login window somewhere on the page, obstructing what I want to see, like as if Google was super desperate for more tracking and engagement. Like why would I like to login to Google while viewing a forum about skyscrapers if I just wanted to read about the progress of some construction? Passive-aggressive at its core.

Taking a break from work and just going for a walk/bike ride in a forest or visiting some interesting place while traveling and working remotely often gave me really great ideas that helped me to solve problems at hand much faster than sitting in an office, fighting with a problem. Like what would take me a week to solve was often done in a day when changing scenery.

How is Stability AI going to make money? What they provide is super cool and VCs are into it but at some point they need to start giving them returns. Is the plan to get acquired?

I think they should first stop treating depression as a single illness. There are likely thousands different reasons for brain to end up with similar symptoms. Many forms of nutritional deficiencies likely turn into something that can be symptomatically classified as depression and sometimes be fully reversible by just restocking that missing nutrient or reducing its intake. Yet we clinically classify such conditions the same as ones caused by some brain injury or mentally horrible experiences that rewire brain circuits in weird ways.

Two PhDs should be the minimal qualification to enter!

That risk is eliminated with foreign contractors. It's import/export of services.

You don't have to. Your foreign workers can be independent consultants receiving money from you, paying their own taxes. You just need their W-8BEN every year.

Power grab and preservation. This will be quickly approved to preserve existing players and power status quo (lawyers, managers and doctors don't want to be automated away).

Low wages, highly educated workforce which keeps grinding, large Apple investments to the latest nodetech, Intel's execution failure.