How is this built? What'd be the approach if I'd like to achieve similar results against proprietary data.
References article speak of RAG and RIG - but I wonder if they factor into fine-tuning the models. AFAIK, RAG doesn't play nicely with structured data.
99% of "software engineering" is about as much "engineering" as laying out and installing the plumbing for a house or smallish apartment building. To the extent that it's more challenging, it's mostly because we're running around trying to hammer nails with screwdrivers and cut copper pipes with butter knives and constantly changing standard pipe & fitting diameters for no good reason.
Canada is in a strange moment where most professional engineering bodies are taking the position that any use of the term 'engineer' by a non-professional engineer (someone with a P.Eng degree) is illegal but US tech companies, which have a lot of influence into Canada's market and aren't affected by the same laws, don't really care. This means it is often used in Canada despite the position taken by the engineering bodies.
The law is sort of vague on it. It says that you aren't allowed to mislead people into thinking you are a professional engineer or use terms that would lead people to think you are a professional engineer if you are not. Whether or not 'software engineer' does this can depend on your viewpoint and whether you're building software for web stacks or for airplane control. But the engineering bodies have tried to make examples of companies or individuals that don't have big enough legal resources to want to go to court over it.
IMO I understand the professional bodies wanting to ensure nobody is misrepresenting themselves as licensed P.Eng holders but claiming a monopoly on the engineering mindset is a bit of gatekeeping and nothing else.
P.Eng is not a degree. You get it by writing a pass/fail ethics exam for the local regulatory board after completing an actual Engineering degree (BASc/BEng), and paying yearly dues.
P.Eng is not required for most engineering activities. One notable exception being civil engineers, since there's so much signing off on blueprints for infrastructure.
Yah, by 'p.eng degree' I meant 'one of the degrees that qualify for a p.eng license'. Thanks for clarifying.
The engineering bodies categorically take the position that you can't use the word 'engineer' anywhere in your title, say that you do engineering, or do engineering under their definition of it without their license. See this recent letter for example [0] or their FAQ on it [1]
I guess you need to look at the current TAM for visual production in general.
That's the baseline (so includes visualization studios, agencies, game studios etc). Generally this potentially can help in many labour intensive parts of a creative visual process.
Is that a "large" organization market or not depends on your metric and what the market positioning of the offering is. I would see applications in both specialist content creation tools as well as "stock photos and merch".
In terms of finding stock photos, if you add a better text api that is easier to control this probably can compete with static stock photos in the sense that people can tune their images as much as they like. For example with their corporate merch (Imagine producing a slideset at Acme co. "Please give me an elephant and walrus wearing acme caps".
Ad agencies already love that they can train a model to quickly iterate product shot ideas extremely rapidly.
Then we have "the usual" effect automation has on market demand - automation increases the productivity of a task requiring labour, hence allowing to reduce the cost of a unit of production, which generally increases the demand. I.e. creative stuff will be cheaper to do, you won't replace artists, but suddenly the dude or dudette who spent hours just tweaking stuff has their own art studio at finger tips to command. They can get so much more done much faster.
The tech is not 100% bullet proof yet but at this pace it will be good enough soon (or probably is for several applications if there was just an UX sugaring targeting specific domain workflow).
Does it actually make anything in the corporate world better to use generated images in slides? When coworkers use stock photos which were presumably made by humans operating actual cameras, I don't think it's clear that their presentation is actually more valuable as a result.
I suspect those applications will come from specializing the model. For example, there's people that have avatar generators or automated ad creatives. A cool application I've been toying with is generating icons.
Every soccer league is on a different provider. I'm in Canada. The Premier League is on DAZN. MLS is on Apple TV. La Liga is on fuboTV (or Amazon Prime). It's all the same sport. Multiply that by X if another family member is interested another sport. I don't know what the future holds, but the sports entertainment fragmentation might be hurting more than helping. Perhaps Apple is on their way to consolidate.
Over here in europe you need multiple providers for the same competition since the rights are split. Champions League football (as in soccer) requires a Sky, DAZN and Amazon Prime subscription if you want to see all matches (of your favorite club). Same thing with most domestic leagues.
> I ended up spending more time on tooling setup than actual business logic.
This is exactly why, in 2004, I moved from Java (Struts/Spring/Websphere/XML/...) to Rails. History repeats but we never learn. Kudos to Rails for remaining relevant.
References article speak of RAG and RIG - but I wonder if they factor into fine-tuning the models. AFAIK, RAG doesn't play nicely with structured data.