I do wonder how some of these companies intend to actually make money with it.
Anthropic, fair enough, they’ve convinced companies to spend huge amount of money making sure that programmers can’t remember anything or know what that mass of spaghetti does any more.
And MS, presumably it’s all companies spending money on Office 365 Copilot so emails get put in automated holding patterns, and quarterly reports get flashier graphs.
Nvidia, they sell shovels.
But Meta? Their entire business model is scraping data from real people, packaging and selling it to far right politicians and scammers. I’m struggling to figure how AI does anything other than fuck that up.
I read recently (ish, couple months now?) that Meta was reshaping its engineering workforce toward having them train an AI to do their work. Big push to record all user activity with their machines and engineers re-tasked away from engineering to doing the crucial, all-important labeling of this recorded data, so it can be used to train a model explicitly intended to replace them.
So yeah I don’t think stuff is going great over there (thank fuck).
Meta is an advertising company. AI is the tool that they can use both to read every post and identify the categories each user will be most receptive to, and also to create customised advertising tailored to that user.
For example, the AI notices that your attention lingers a little longer on user posts about travel, and then inserts an advert into your stream which looks indistinguishable from those posts and which encourages you to take that overseas holiday you’ve been thinking about.
From there, it becomes even more incidious. Advertising is about promotion and persuasion, and word of mouth is a very effective persuasion tool, so imagine the AI becomes your ‘friend’, or at least one of the AIs many personalities does, because it can post from millions of accounts, each of which can slightly alter your worldview, biases, morals, and influence your needs, your desires, and so on. Now it’s not just inserting an advert into your post stream, it’s actively participating in the content. Each post is unique to you - while you see posts from your ‘friends’ who are thinking about about travel to Antarctica, someone else sees stories about replacing tiles on their roof.
AI can collate a sequence of posts about climate change and then need to move away from fossil fuels, in order to persuade someone who it already knows feels strongly about that topic, in order to encourage them to buy a new car, and not just any car, but a particular brand. It might even subtly offer a discount that the advertiser has authorised it to do, for example, mentioning that they saw a dealer just down the road from you has one car left in stock at a huge discount, but it will surely be gone soon.
This is why Google (for example) is pushing gemini hard - it wants it to become everyone’s friend and trusted source of information, because once they’ve achieved that, they can use it to subtly drop hints about products in its responses - nothing obvious, that would trigger the skip button, but just a careful comment here, another one there, a post about how a “user” had problems getting a particular brand of car repaired, and so on and so on.
It’s all about ad revenue, and you’re the product.
Gemini is a dark corner of the internet tbh. You can dig and dig and dig thinking youre getting accurate info. All it takes is one second to step back and be like, is this info accurate? Then you find it doesnt even remotely hold up. Their LLM is trash.
It’s been a couple years, but I used Gemini for coding and I found it to be at least better than ChatGPT at the time, but for everything else it was so much worse than it. But yes, even when coding and having it supposeldy dig through docs, it had quite the error rate.
The main product they’re pushing is Muse. It’s a glorified Bonzi Buddy. A chat bot with customizable avatars. That’s it. That’s the selling point. They’re screwed unless they have an enterprise LLM product. I don’t know if they do.
Facebook remade the world. It was a one in a generation thing. Zuck isn’t escaping the allegations that he was in the right place at the right time for that. He’s not going to catch lightning in a bottle twice.
I do wonder how some of these companies intend to actually make money with it.
Anthropic, fair enough, they’ve convinced companies to spend huge amount of money making sure that programmers can’t remember anything or know what that mass of spaghetti does any more.
And MS, presumably it’s all companies spending money on Office 365 Copilot so emails get put in automated holding patterns, and quarterly reports get flashier graphs.
Nvidia, they sell shovels.
But Meta? Their entire business model is scraping data from real people, packaging and selling it to far right politicians and scammers. I’m struggling to figure how AI does anything other than fuck that up.
I read recently (ish, couple months now?) that Meta was reshaping its engineering workforce toward having them train an AI to do their work. Big push to record all user activity with their machines and engineers re-tasked away from engineering to doing the crucial, all-important labeling of this recorded data, so it can be used to train a model explicitly intended to replace them.
So yeah I don’t think stuff is going great over there (thank fuck).
Meta is an advertising company. AI is the tool that they can use both to read every post and identify the categories each user will be most receptive to, and also to create customised advertising tailored to that user.
For example, the AI notices that your attention lingers a little longer on user posts about travel, and then inserts an advert into your stream which looks indistinguishable from those posts and which encourages you to take that overseas holiday you’ve been thinking about.
From there, it becomes even more incidious. Advertising is about promotion and persuasion, and word of mouth is a very effective persuasion tool, so imagine the AI becomes your ‘friend’, or at least one of the AIs many personalities does, because it can post from millions of accounts, each of which can slightly alter your worldview, biases, morals, and influence your needs, your desires, and so on. Now it’s not just inserting an advert into your post stream, it’s actively participating in the content. Each post is unique to you - while you see posts from your ‘friends’ who are thinking about about travel to Antarctica, someone else sees stories about replacing tiles on their roof.
AI can collate a sequence of posts about climate change and then need to move away from fossil fuels, in order to persuade someone who it already knows feels strongly about that topic, in order to encourage them to buy a new car, and not just any car, but a particular brand. It might even subtly offer a discount that the advertiser has authorised it to do, for example, mentioning that they saw a dealer just down the road from you has one car left in stock at a huge discount, but it will surely be gone soon.
This is why Google (for example) is pushing gemini hard - it wants it to become everyone’s friend and trusted source of information, because once they’ve achieved that, they can use it to subtly drop hints about products in its responses - nothing obvious, that would trigger the skip button, but just a careful comment here, another one there, a post about how a “user” had problems getting a particular brand of car repaired, and so on and so on.
It’s all about ad revenue, and you’re the product.
Gemini is a dark corner of the internet tbh. You can dig and dig and dig thinking youre getting accurate info. All it takes is one second to step back and be like, is this info accurate? Then you find it doesnt even remotely hold up. Their LLM is trash.
It’s been a couple years, but I used Gemini for coding and I found it to be at least better than ChatGPT at the time, but for everything else it was so much worse than it. But yes, even when coding and having it supposeldy dig through docs, it had quite the error rate.
Nvidia is also doing everything it can to cause artificial shortages in GPUs to raise prices.
The main product they’re pushing is Muse. It’s a glorified Bonzi Buddy. A chat bot with customizable avatars. That’s it. That’s the selling point. They’re screwed unless they have an enterprise LLM product. I don’t know if they do.
Facebook remade the world. It was a one in a generation thing. Zuck isn’t escaping the allegations that he was in the right place at the right time for that. He’s not going to catch lightning in a bottle twice.