Literally that Black Mirror episode, where the woman has a brain implant to cure a brain trauma, and they put ads in her speech because she is on the free subscription tier.
Not to defend adding ads to chatgpt, but it's quite different from an implant that turns people into walking ads.
Also, they weren't on the free tier, their subscription was already stretching their finances and then they added ads and a new even more expensive tier to avoid the ads.
Your suppose to create a monopoly THEN shittify your product with ads. OpenAI is like speedrunning being a big tech company before our own eyes, it lines up from what I hear from people working there.
The disconcerting thing is that in LLMs, the ads don't need to be signposted, advertisers can just pay to bias the LLM. You'll have Big Pharma paying to bias the LLM into convincing people they need meds.
Is Sora even that popular versus other AI video generators? They had first mover advantage but its largely consumed off site so they still don't have control of distribution.
I run it on the hardware I have, the data stays with me, offline. I run it on hardware I already have for other purposes, and I even have a portable solar panel (but I use the plug socket).
Qwen3 30b a3b, for example, is brilliant for its size and i can run it on my 8 GB VRAM + 32 GB RAM system at like 20 tokens per second. For lower powered systems, Qwen3 4b + a search tool is also insanely great for its size and can fit in less than 3 GB of RAM or VRAM at Q5 quantization
8B QWEN for example. It's limited, and can't "reason" almost at all, but it does give well-structured answers. You can extract references from it, up to a point.
I am very forgetful, and googling takes forever. So if it does not sound bs, I just accept the answer. If the stakes are higher, I google it's references.
I suspect that for a 5th grade teacher, something written by OpenAI will be one of the best papers they've ever read. Not insightful in any way, but correct spelling, punctuation, paragraphs, vocabulary, etc.
No because kids are smart enough to add "write it like a kinda dumb 5th grader" to the prompt. So its not just slop its intentionally dumb badly written slop that makes you question if a sentient being wrote this.
I do not believe their models are good enough to accurately and organically "prefer" one result over another. Not to say they won't try, just that it sounds futile. How do you even train a model to pick advertisers over non-advertisers? Feels like they're just going to start shipping prepackaged "assistants" with a big, stupid list of company sugar daddies to fluff up.
42 replies
Literally that Black Mirror episode, where the woman has a brain implant to cure a brain trauma, and they put ads in her speech because she is on the free subscription tier.
The common tier wasn't even free x.x
Not to defend adding ads to chatgpt, but it's quite different from an implant that turns people into walking ads.
Also, they weren't on the free tier, their subscription was already stretching their finances and then they added ads and a new even more expensive tier to avoid the ads.
Elons working hard on that one. Don't worry!
Yes, it was the base tier in the episode (not free). You'll forgive me, it's been a some time since I watched it.
It's not a 1:1 comparison, but you'll admit that the resemblance is uncanny.
On a Netflix Original Series...👨🍳💋
Don't remember that one, gotta look it up
It's from the last year's season. S07E01 - Common People.
Your suppose to create a monopoly THEN shittify your product with ads. OpenAI is like speedrunning being a big tech company before our own eyes, it lines up from what I hear from people working there.
They’re burning money. They have no choice.
This. They need some form of actual income, and ad revenue gives them theoretical capital to loan against.
The disconcerting thing is that in LLMs, the ads don't need to be signposted, advertisers can just pay to bias the LLM. You'll have Big Pharma paying to bias the LLM into convincing people they need meds.
I have a feeling that all the OpenAI employees switch to Claude once they leave work
They captured a market though, generated cute video for facbpoks peoples, it is non viable but still a market.
Is Sora even that popular versus other AI video generators? They had first mover advantage but its largely consumed off site so they still don't have control of distribution.
No idea.
This may be what pops the bubble.
Inserting ads into the response would ruin the responses. Making it even more useless
Whats great is they are probably feeding the code to itself. It becomes a giant ad-bot.
But you can subscribe to premium plus, which will only introduce ads next year after they made sure your processes depend on it!
::: spoiler Tap for spoiler Yeah, no idea what kind of processes that should be. Your boss will sure have some great ideas :::
This is going to be hilarious.
Luckily, I have local AI.
And you should too!
With these ram prices?
Rather live without ai
I run it on the hardware I have, the data stays with me, offline. I run it on hardware I already have for other purposes, and I even have a portable solar panel (but I use the plug socket).
doesn't AI need like 96 gigs of ram to be comparable in quality (or lack there of, depending on how you view it) yo the commercial options?
Qwen3 30b a3b, for example, is brilliant for its size and i can run it on my 8 GB VRAM + 32 GB RAM system at like 20 tokens per second. For lower powered systems, Qwen3 4b + a search tool is also insanely great for its size and can fit in less than 3 GB of RAM or VRAM at Q5 quantization
8B QWEN for example. It's limited, and can't "reason" almost at all, but it does give well-structured answers. You can extract references from it, up to a point.
What for? I can't think of a single problem I have in my life where the answer is AI.
I badly classify factures with a local ia. That bad but hey that still things I dont' have to do
I am very forgetful, and googling takes forever. So if it does not sound bs, I just accept the answer. If the stakes are higher, I google it's references.
????
Hey, so I've got this bridge for sale...
Username checks out
Username checks out.
Which model do you run and how? I tried ollama a few months back but it was still quite poor
"This paper brought to you by RAID SHADOW LEGENDS..."
I suspect that for a 5th grade teacher, something written by OpenAI will be one of the best papers they've ever read. Not insightful in any way, but correct spelling, punctuation, paragraphs, vocabulary, etc.
No because kids are smart enough to add "write it like a kinda dumb 5th grader" to the prompt. So its not just slop its intentionally dumb badly written slop that makes you question if a sentient being wrote this.
"Oh this is from War Thunder's russian bias"
Then the next generation of models fed on slop with adverts in and increasingly adopt the writing style of marketing copy.
This was either from April or early December. In either case, I've never heard this repeated anywhere else since.
Edit:
https://www.bleepingcomputer.com/news/artificial-intelligence/leak-confirms-openai-is-preparing-ads-on-chatgpt-for-public-roll-out/
https://futurism.com/artificial-intelligence/openai-chatgpt-sponsored-ads
I do not believe their models are good enough to accurately and organically "prefer" one result over another. Not to say they won't try, just that it sounds futile. How do you even train a model to pick advertisers over non-advertisers? Feels like they're just going to start shipping prepackaged "assistants" with a big, stupid list of company sugar daddies to fluff up.
More cringe would be advertising german bath products.