Because TikTok is free, had no competitors and network effects given that it is a social media platform. ChatGPT already depends on subscription income, has to compete with companies that can offer the same service for free and has no network effects because you're literally talking to a commodified bot
> TikTok [..] had no competitors and network effects
TikTok, or rather ByteDance, acquired Musical.ly as a competitor to absorb the user base and jump start their network. Their also have been a lot of short-form video platforms before (e. G., Vine) and during TikToks growth (Instagram reels, YT Shorts).
Reuters reported that ByteDance (TikTok parent) in Q1 2025 had $48b in revenue.[0] They should surpass $200b for 2025 which would make them bigger than Meta.
In other words, Tiktok has already caught up with Instagram in terms of revenue.
> most of which is from the Chinese market as it continues to face political pressure to divest its U.S. arm.
We're not comparing apples and oranges here. Google and Meta don't operate in China, so there is no giant online ad spend (especially for social) already allocated like there is in the US.
> Like, I've seen no sign of OpenAI building an ads team or product
You just haven't been paying attention. They hired Fidji Simo to lead applications in may, she led monetization/ads at facebook for a decade and have been staffing up aggressively with pros.
Reading between the lines in interview with wired last week[0], they're about to go all in with ads across the board, not just the free version. Start with free, expand everywhere. The monetization opportunities in chatgpt are going to make what google offers with adwords look quaint, and every CMO/performance marketer is going to go in head first. 2% is tiny IMO.
I have indeed being paying attention, thanks. One executive does not an ads product make, though.
I think that ads are definitely a plausible way to make money, but it's legally required that they be clearly marked as such, and inline ads in the responses are at least 1-2 versions away.
The other option is either top ads or bottom ads. It's not clear to me if this will actually work (the precedents in messaging apps are not encouraging) but LLM chat boxes may be perceived differently.
And just because you have a good ad product doesn't mean you'll get loads of budget. You also need targeting options, brand safety, attribution and a massive sales team. It's a lot of work and I still maintain it will take till 2030 at least.
Google should have to make this disclosure as well. I'd guess >50% of their AdX revenue is from click-trick, fake button, scam ads. Across the board I'd expect Google's ad revenue to be at least 10% from scams, if not more.
Source: a decade of running a website monetized with adx and having to hire people to manually monitor and block scam display ads multiple times a day.
It's because when you live rural like this, wood stoves are common, and wood is free.
I live in the northwest, so I can't speak to upstate NY, but downed trees on state and federal land near roads is free to take. Every day there's people posting rounds of wood for free to take.
It's hard work, but it's good exercise and rewarding.
There's some upfront investment: $200 chainsaw, an old maul, and an old pickup truck, but those amortized over a decade is practically speaking $0 heat.
To add to the sibling comment, collecting this wood takes time. I've collected wood the forest service takes down for use in a stove I use but processing all that wood takes time. You bring it home, cut it into small bits, keep it in a dry area to make sure the green wood dries out, and then you meticulously rotate older and newer stock to make sure you use the driest stuff for heating.
If you're living on $432 / month and working 30-40 hours at this cashier job then using your off days to grab and process wood is honestly pretty miserable. There are slums in developing countries with higher standards of living because they can heat their "house" (read: tent or hut) with oil.
I grew up in a 1300sf wood heated house, so I have relevant experience here. It does take time to buck, split, load, unload, and stack the wood. It goes faster if you have a small child (me) to help!
We cut wood for our own use and also sold it, so it didn't require 100% of our time to keep the heat on.
Well, minimum wage in NY state is $15.50/hour. ($432/mo)/($15.50/h) is about 28 hours per month, i.e. 7 hours per week.
https://dol.ny.gov/minimum-wage-0
He also mentions other forms of employment, like raising rare herbs, so maybe he's got a little homegrown operation going that doesn't take much time.
If that's what you do in this situation, why didn't the author write that instead?
There's some upfront investment: $200 chainsaw, an old maul, and an old pickup truck, but those amortized over a decade is practically speaking $0 heat.
I feel like this is really stretching the definition of "$0".
Well water being free also means amortizing the potential maintenance costs of the pump, filters, and testing to make sure you aren’t drinking arsenic or lead.
Wood is free if you scavenge wood from an uncertain source, ignore fuel, equipment, time, oh and labor. Never mind it's green wood, so you need to manage stockpiling to dry it.
I live in a country where for half of the population wood is the default fuel. There's a reason it's a lot of peoples job.
Wood is not free, and you have to consider the time and gas and tools to cut it, also this guy doesn't even have a car not to mention a truck to haul it, and he is living on a lot, not a wooded acreage. 90% of wood people want to get rid of for free is pine or fir which takes 4 times as much to match hardwood heating, and takes more careful stone pipe maintenance to not build up creosote and get a chimney fire, a lot of people exclusively burn hardwood just because of the risks of burning softwood and causing a fire from buildup. Even myself being "lucky" to have Emerald Ash Bore which has killed 95% of ash trees and given me "free" dead hardwood for the last 25 years wouldn't consider it free.
Say you do have a wooded plot, the first year or two it might not be so bad, lots of wood near the edges where you can drive up to to load and move, but what about after that when you have to go deeper into the woods? You need to get in there, it may not be accessible by truck or get swampy where you will get stuck, and now you are considering a tractor or other vehicle, a decent expense to obtain, in order to not have to carry all your wood an armload at a time through the woods longer and longer distances. Chains and gas and oil for cutting it aren't expensive but not free, nor is maintaining a gas chainsaw if you seriously use it for all your heating wood, doubly so if you aren't already mechanically inclined enough to repair engines. And then you still have to split it. There are cheap splitters, but cheap spliters will only split the wood that took little effort to split with an axe, and less than half the wood you cut is going to be that easy to split straight grain wood, so you are either going to need more for a splitter or to be physically fit and capable enough to split a lot of gnarly wood by hand. Some people enjoy it for the exercise, I do, but not everyone is up to it, and it is such a hard physical activity that you need to be in good health to maintain it.
Also splitting mauls are a gimmick, they take far more effort than a long handled axe and are only a good option if you are otherwise incapable of using an axe. Speed applies more kinetic energy than mass, kinetic energy is half of the mass times velocity squared, so doubling the mass you are throwing around is far less effective at applying force into splitting than trying to double your swing speed. And that is the biggest "trick" to a good axe split, swing speed, which is why you want a long handle. Mauls are far slower than an axe, take more energy than an axe to lift and swing, and are far less capable of splitting more gnarly wood as the more aggressive edge angle has a much harder time splitting into and separating the grain as much of the energy merely crushes wood fibers before it bites in and starts wedging. If an axe can't do it, a maul won't do it even more, and then you are getting into a sledge hammer and steel wedges anyways, and a wedge and sledge are easier to set and more maneuverable than a maul with a big ass handle on it.
Burning wood is a decent way to heat a house if someone is always at the house in regular 8 hour intervals or more, but it has a lot of caveats and is not what I would call free. More like subsidizing a portion of the cost of with hard physical labor.
I have no experience with this, and the only time I've gone and tried to get quoted for things like cyber liability, etc, the costs are incredible relative to the value of the business and revenues.
Historically the problem with using LLMs for the super simple conventional NLP stuff is that they were hard to control in terms of output. If you wanted a one-word answer for a classification task, you'd often have to deal with it responding in a paragraph. This obviously hurts precision and accuracy quite a bit. There were tricks you could use (like using few-shot examples or GBNF grammars or training low-rank adapters or even re-asking the model) to constrain output a bit, but they weren't perfect.
Over the last 12-18 months though, the instruction-following capabilities of the models have improved substantially. This new mistral model in particular is fantastic at doing what you ask.
My approach to this personally and professionally is to just benchmark. If I have a classification task, I use a tiny model first, eval both, and see how much improvement I'd get using an LLM. Generally speaking though, the vram costs are so high for the latter that its often not worth it. It really is a case-by-case decision though. Sometimes you want one generic model to do a bunch of tasks rather than train/finetune a dozen small models that you manage in production instead.
Super easy to get started, but lacking for larger datasets where you want to understand a bit more about predictions. You generally lose things like prediction probability (though this can be recovered if you chop the head off and just assign output logits to classes instead of tokens), repeatability across experiments, and the ability to tune the model by changing the data. You can still do fine tuning, though itll be more expensive and painfaul than a BERT model.
Still, you can go from 0 to ~mostly~ clean data in a few prompts and iterations, vs potentially a few hours with a fine tuning pipeline for BERT. They can actually work well in tandem to bootstrap some training data and then use them together to refine your classification.
After prompt optimization with something like DSPy and a good eval set, significantly faster and just about as good. Occasionally higher accuracy on held out data than human labelers given a policy/documentation e.g. customer support cases.
Given the source, I'm skeptical it's not just a troll, but found this explanation [0] plausible as to why those vague spam text exists. If true, this trolling helps the spammers warm those phone numbers up.
Carriers and SMS service providers (like Twillio) obey that, no matter what service is behind.
There are stories of people replying STOP to spam, then never getting a legit SMS because the number was re-used by another service. That's because it's being blocked between the spammer and the phone.
ChatGPT has roughly the same MAU as tiktok. I don't see why their ad business wouldn't meet or exceed what tiktok was able to do in less than 5 years.