is AI gonna take our jobs?

Gpt-4 is out. Supposedly it's way better than 3.5

Then you find out that 4 is what's been powering Bing's chatbot this whole time -- you know, the hallucinatory psychobot

also, just saw this article which hypothesizes that GPT-4 may have tested on its training data when it aced all the exams they claim it aced (such as the bar) because it can handily solve publicly available coding challenges and other problems from before September 2021 (the cutoff of its training data) but not after: https://aisnakeoil.substack.com/p/gpt-4-and-professional-benchmarks

Midjourney v5 is out, and the results are impressive. It may have improved on its ability to generate hands? One interesting thing to me about AI photos is that they generally look kind of glossy and airbrushed. Probably because so many professionally-produced photos do. Which doesn't limit their commercial applications at all, if anything it's probably a good thing for that. But I haven't seen a single image that looks like a real photo, like what you get when you release the shutter on a real camera.

People are also pointing out the risks of "criti-hype," which is when AI boosters ostensibly criticize AI by predicting its dangerous impact on society, but what they're really doing is hiding a product advertisement inside what looks like an admission against interest, which tricks their critics into parroting their hype.
 
"First they came for the headshot photographers..."

It's the beginning of the end, my friends. $17 for a package of impressively good headshots based on a spread of selfies you upload. "Good enough" meets "Holy crap that's cheap" (and will get cheaper when Google or Walmart horn in on the territory and offer the same thing for $2.99)

And there are other advantages to not doing an actual photoshoot. Executives can save valuable time. Not everyone enjoys posing in front of a camera. Maybe I can get photos of myself clean-shaven without losing my beard.

Not every type of photography will get swept aside by AI so easily. But dare I say that professional headshot photography, the bread and butter of many, may quickly become a niche service.

But will the photos be public domain since they're made by AI? Or maybe the likenesses still have rights protections?

https://petapixel.com/2023/03/21/people-are-paying-17-for-hundreds-of-ai-generated-headshots/

"People are Paying $17 for Hundreds of AI-Generated Headshots"

https://www.tiktok.com/@crystaldbri...21331973,120811592,120810756;null;embed_share
 
Eh, it'll take over the low end snapshots. The high end will then make a point that "This is not AI. This is a real photo and this is a real face".

And then people will begin to ask the same questions. And others will demand to see a certificate that it was shot by a real photographer. In other words, it will become so commoditized, an alternative will be preferred.

Now, the alternative can be sufficiently performed by a mid-range smartphone. That's what should worry the pros.
 
And it should worry the manufacturers too. Not the AI per se, whatever it is, but the manipulation of images from all the various software products. If one can make a smartphone photo appear pretty close to a $5,000 camera, then one would likely skip on a $5,000 camera. One would still need some sort of expertise to work the mobile phone but the gap between the SoNiCan and the iPhone is blurrrrrrring.
 
Eh, it'll take over the low end snapshots. The high end will then make a point that "This is not AI. This is a real photo and this is a real face".

And then people will begin to ask the same questions. And others will demand to see a certificate that it was shot by a real photographer. In other words, it will become so commoditized, an alternative will be preferred.

Now, the alternative can be sufficiently performed by a mid-range smartphone. That's what should worry the pros.

I'm not so convinced because headshots are pretty utilitarian. (Not headshots for actors, which is a totally different story and I think will continue to be taken in real life.) But if you want to put a photo on LinkedIn or on your website, you simply need a representative image of yourself that is appropriately professional and hopefully flattering. There's no emotional attachment to the time and place of the photo, and no bonus points for authenticity. Or do you disagree?

I'm not worried about smartphones. Because if you're going to take a real headshot, the art of the photo is in the directing, location selection, lighting, framing. (Some of that could be fixed in post with increasingly powerful relighting tools, but not the directing part.) That's one of the powerful things about these AI headshot packages-- they give you a smile and a flattering angle and a flattering lighting. The hard parts of real headshot photography.

So, even if you can shoot it with an iPhone, you'll still want a seasoned pro to set up the shot and direct you into flattering poses.
 
Wasn't that my point?

Pros shall remain pros.

But the gap is shrinking.

the gap between camera and smartphone? but not the gap between pro and amateur?

physics favors the camera. and there's only so far that computational photography can take you. but it doesn't matter to me, because I'm principally paid for my skills not my equipment
 
the gap between camera and smartphone? but not the gap between pro and amateur?

physics favors the camera. and there's only so far that computational photography can take you. but it doesn't matter to me, because I'm principally paid for my skills not my equipment

The gap between pros and amateurs is shrinking too but not because of the cameras per se but because of their ubiquity. In other words, nowadays pretty much anyone can practice photography and even "film making" ad nauseam and at very little cost. It may not be on a smartphone but on a $500 used hybrid camera but practice - I am talking about practice, not the game, not the game, but practice - does improve people's skills. Often by a lot.

PS. A friend of mine wants to start his own YouTube channel. He's into motorcycles and cars and ... some other stuff. So, I tell him, "You have iPhone 13. Go watch a bunch of iPhone clips, then look up how to shoot a single person video with an iPhone. Then look up how to light your place. Then how to edit the footage". Now, he may give up at any point but if he does go through all the steps, he'll definitely improve.

A few weeks ago, he sent me a photo of Rock Hudson and Liz Taylor from the "Giant"(1954) and asks, "What do you think of the composition?" Well, both the director (George Stevens) and the cinematographer won an Oscar for the film. The composition may be said to be perfect but it's also an artistic choice in terms of the camera angle and the subject framing. But, IMO, it's a learnable skill, even if you don't aspire to be the next George Stevens.
 
A professor says he's stunned that ChatGPT went from a D grade on his economics test to an A in just 3 months

https://www.yahoo.com/finance/news/professor-says-hes-stunned-chatgpt-090000832.html

PS. Bryan and I exchanged a few emails back in the day. He's a public figure now. I just post on DVX User.

Might be a case of programmers teaching to the test, so to speak. OpenAI knows that one of the easiest benchmarks to promote and communicate is test scores, so that's been a focus for them. (I've read that in some ways its improvement in test-taking has led to a decline in cross-disciplinary performance.) It's not that acing tests is bad, but it's not the same thing as performing real-life tasks of the professionals.

Personally, I'm curious to see if generational improvements continue at the current pace or even accelerate, or if they plateau.

Here's a short film that someone made using AI (a combination of live recording, Midjourney, and Runway). Looks like unwatchable garbage to me.

https://twitter.com/mrjonfinger/status/1637205235465863168?s=20
 
I don't think it's unwatchable. The key here is the amount of money and effort it'd take to make the same film - its imperfections acknowledged - via s standard camera with or without a crew. While it may look like a cheap'ish video game, it's quite possible to tell a story using these tools. And this is just a start.

As to the standardized tests - Deep Blue computer beat Gary Kasparov in chess almost 26 years ago. And it's much harder to beat a great chessplayer than to pass a standardized econ exam. Deep Blue could operate at 11:38 gigaflops. A two year old Samsung Exynos 2100 can do 1,530. And you have it in your Galaxy line of smartphones that came out back in 2021-22.
 
I don't think it's unwatchable. The key here is the amount of money and effort it'd take to make the same film - its imperfections acknowledged - via s standard camera with or without a crew. While it may look like a cheap'ish video game, it's quite possible to tell a story using these tools. And this is just a start.

As to the standardized tests - Deep Blue computer beat Gary Kasparov in chess almost 26 years ago. And it's much harder to beat a great chessplayer than to pass a standardized econ exam. Deep Blue could operate at 11:38 gigaflops. A two year old Samsung Exynos 2100 can do 1,530. And you have it in your Galaxy line of smartphones that came out back in 2021-22.

More computing power is good. It won't lead to AGI though because it's not about how fast you can compute.

To me, that clip was unwatchable. And the thing about mass media is that it's relatively winner take all. Titanic went way over budget, but I'd say that the marginal overage spent perfecting it may have quadrupled its box office gross.

But let's say that generative AI can generate realistic video by making multiple passes and gradually fixing the inconsistencies from frame to frame in a clip. Having a faster computer would help with that process.

I'm guessing that in the near term, high end film/TV production will look similar to its current form. They'll keep shooting key elements, performers etc. But the CGI will be based on AI generations or tweaked by AI. And maybe you don't need to wear a mocap suit anymore because the computer can analyze your movements from a regular recording.

goldman sachs predicts that 300 million jobs worldwide will eventually incorporate AI. Happily, they predict that--

about 7 percent of US workers are in jobs where at least half of their tasks could be done by generative AI and are vulnerable to replacement.
https://m.slashdot.org/story/412476

Who knows if they're right. But this suggests that 93% of jobs will at most only be augmented by AI rather than replaced outright. That would be a better outcome than I had anticipated.

And crucially, goldman expects that lawyers and legal staff are actually the #1 most vulnerable group to replacement by AI. I'm somewhat reassured by this news not because I want lawyers to lose their jobs but because I know that there will be a populous and powerful group adversely impacted by AI, and there will be a political movement to shield them from the fallout of the AI transition. Whereas if it were only photographers and videographers losing their jobs, I know we'd be screwed.

I stumbled across a reddit of coders expressing a similar sentiment-- that they take solace in knowing they'll be in the same boat as their peers, including those with more experience. It's a funny thread. At one point there's an extended analogy about D&D and how they chose the wrong dump stats.

https://www.reddit.com/r/singularit...ais_richard_ngo_i_predict_that_by_the_end_of/
 
That some jobs will be displaced by technology is as inevitable as the electric vehicles not having an internal combustion engine. Which will lead to a lot of car mechanics losing their jobs. And gas stations closing.

As to the various text-to-video apps, they're in such infancy at the moment, no standard can be applied to them. Because there is none. But I would like to see a "film/TV quality" clips - however short - soon. With the BTS on how that clip was made.
 
That some jobs will be displaced by technology is as inevitable as the electric vehicles not having an internal combustion engine. Which will lead to a lot of car mechanics losing their jobs. And gas stations closing.

There are a few big issues here. The speed and breadth of AI deployment will probably dwarf all previous technological transformations. AI can do a lot of different things, and it's really cheap and quick to start using it.

The scale and speed of that job displacement is going to be difficult to deal with.

More to the point, if you don't lose your job then good for you. But what if you just graduated law school with $150k in debt and suddenly your job no longer exists. At a bare minimum, we're looking at years of lost income for anybody who needs to find a new career.

Mechanics will be fine. They'll need to learn how to work with EVs, but their skills will still be in demand. Charging stations will replace gas stations, though I don't know how practicable it is for gas stations to be converted. But there we're talking about a narrow set of industries and a transition that will probably take 15 years minimum.
 
Mechanics will be fine. They'll need to learn how to work with EVs, but their skills will still be in demand. Charging stations will replace gas stations, though I don't know how practicable it is for gas stations to be converted. But there we're talking about a narrow set of industries and a transition that will probably take 15 years minimum.

Hypothetically speaking, a gas station could buy a gas turbine and convert any "fossil fuel" ( though it mostly work on natural gas, which can be delivered in a compressed form) into electric power and a subsequent charging station. The problem for the gas station owners isn't technical but the fact that 80% of all EV owners charge at home. On top of that, the next generation batteries with a much higher density could double the capacity of an EV, thereby making a single charge last up to 600 miles. And then there are roof top solar panels on top of that.
 
Hypothetically speaking, a gas station could buy a gas turbine and convert any "fossil fuel" ( though it mostly work on natural gas, which can be delivered in a compressed form) into electric power and a subsequent charging station. The problem for the gas station owners isn't technical but the fact that 80% of all EV owners charge at home. On top of that, the next generation batteries with a much higher density could double the capacity of an EV, thereby making a single charge last up to 600 miles. And then there are roof top solar panels on top of that.

Possibly true, probably true. The thing is, if you run out of charge somewhere you will need a station close enough that you don't get stranded. And a lot of home chargers charge pretty slowly, I believe, which probably won't change with the shift to higher-density batteries (although some other technology might change it. but high-voltage stuff is pretty dangerous generally, I think.)

And then when you consider that gas stations, like movie theaters, make most of their money off concessions. It's actually not a terrible thing that your customers might be stranded for an hour while they top up their vehicle. Fewer customers, but for longer time.

I think that on the balance, the shift to EVs is certainly bad for gas stations. But not an extinction-level event.
 
Given the latest developments on the battery front, it makes 110% sense for homeowners to install the roof-top solar panels with the inexpensive non-Lithium (the LFP's ain't it) batteries. And connect them to faster chargers, though with the next gen EV's supposedly doubling the range of the current models, a slow charger should be sufficient as well.

As to being stranded, there are companies actually offering a mobile recharge. All one'd need is a truck with a stack of batteries. And a cable.

Which means that a gas station is going the way of the drive-in movie theaters and milkmen. And live boxing on a network TV. And network TV.
 
Back
Top