If you are interested in this sort of thing, you might want to take a look at a very simple neural network with two attention heads that runs right in the browser in pure Javascript, you can view source on this implementation:
Even after training for a hundred epochs it really doesn't work very well (you can test it in the Inference tab after training it), but it doesn't use any libraries, so you can see the math itself in action in the source code.
The screenshot is an example of someone making a reference to "removing fingers" (direct quote from the image) to make someone "care more" (direct quote from the image).
Regarding all the asterisks, your guess is as good as mine. Perhaps it is to make the text look scarier and more menacing, or more serious, or to make it seem like it was written by a madman who will actually do what is described. It's hard to know why they did that.
I'm interested in product type X when I'm ready to buy product type X and thus do my best googlefu search for product type X.
I don't need to see nor know about product type X 5 years before I'm ready to purchase it.
This is the fallacy of advertisers. They think that you as an individual are too stupid to do any legwork of your own. They treat consumers like toddlers and drooling idiots.
And said "new products and services" are /not interesting/ to me until I'm ready to engage with said product or service.
Actually new product categories are vanishingly rare, so it's a sure bet that when something new comes around, people will be talking about it and I'll hear about it at some point.
But you do have a point -- avoiding advertising does mean that there may be some specific new products and services that I won't hear about for a while (although I will eventually). Personally, I don't see that as a real problem that requires solving.
On the other hand, I think it is not reasonable to expect people to tell all their friends about all the goods and services that are available. For example, when was the last time you talked to your friends about what electricity provider you are with, about your telecom company, about where you buy your gas and for how much, about what brand of mobile phone you use, about where you shopped for clothing and what you bought, about your heating and ventilation system, your lawn care, your health insurance company, your tax preparation software, your dishwashing detergent, your banking services, your insurance, and so forth? These are tedious things that require consumer information and consumer choice.
Our vision is for AI to autonomously do all the work and provide every sector of goods and services. At the end of the day consumers need to do their part to be informed about it, for effective market mechanisms and competition to function.
What is the alternative? An AI robot is putting a giant TV up in your living room and doesn't even ask you if you want one, it just buys it, delivers it, installs it and tells you Enjoy your new TV?
This would remove all market feedback.
So how do you give a choice about whether you want resources directed at a larger TV or a faster and larger GPU for gaming, without consuming and potentially responding to advertising on these subjects?
Additionally, how would better choices be promoted that require consumer education? An example would be dietary choices. How would the benefits of farm fresh fruits and vegetables be promoted, to compete with processed food? Are your friends going to tell you to choose fruits and vegetables? How will they know about it?
Disseminating information and allowing consumer choice is vital to the vision of the State of Utopia, and there has to be a mechanism for consumer reading and information, even when it is a chore.
Let's take an extreme example of resource allocation questions that are in the hands of consumers: a base on Mars would cost hundreds of billions of dollars and support a few people who have nothing to do up there. When it comes time to vote on whether you want it, how will you be informed of the benefits, if any, if the AI did not pay for informative advertising to try to convince you? Voting on this question requires an informed consumer base, and getting informed is work in and of itself. If the space agency division of AI were forced to produce informative advertising for it, people would consume it and then could vote yes or no. This would be direct market feedback that informs the actions of AI.
Absent this, you would not have a basis for an informed vote on the most basic resource allocation questions.
Clearly, an informed consumer and an informed voter are both necessary for the functioning of a market economy and democratic government, and we need a mechanism to disseminate information about the available choices.
> On the other hand, I think it is not reasonable to expect people to tell all their friends about all the goods and services that are available.
I don't expect such a thing at all, but it's pretty natural for people to tell their friends when they're delighted (or dismayed) by things they use, or to talk about them if I ask them "hey, how to you do X?" They're not going to tell everyone about everything -- and people who do would become as annoying as ads are.
> Our vision is for AI to autonomously do all the work and provide every sector of goods and services.
I don't share that vision with you, that sounds horrible to me. But it's orthogonal to the question of advertising.
> So how do you give a choice about whether you want resources directed at a larger TV or a faster and larger GPU for gaming, without consuming and potentially responding to advertising on these subjects?
There are numerous ways. I'm of the opinion that advertising (at least, if what we mean by "advertising" is marketing messages from the manufacturers or sellers) is the worst option. It's always going to be heavily biased and incomplete, and is very likely to be manipulative.
> This would remove all market feedback.
Market feedback comes from what sells and how much, as well as direct feedback from customers, not from advertising.
> Clearly, an informed consumer and an informed voter are both necessary for the functioning of a market economy and democratic government, and we need a mechanism to disseminate information about the available choices.
I agree. Where we disagree is about whether or not advertising has a useful role in this. I think it does not. Advertising is, in my view, adversarial to that goal.
I learn of them through my friends, and when I'm actively looking for products or services to meet a need I have.
Ads aren't helpful, for the most part, even if they're about something that I might be interested in. They're very rarely actually informative enough for that.
Even if they were generally useful to me, I'd go out of my way to avoid them because of the data collection that comes with them. I certainly wouldn't be willing to take part in a system where I get paid to see them, because that automatically would involve identifying and tracking me.
By the way... this:
> If you email us and do not receive a reply within 24 hours, you are entitled to $10,000 in a class-action lawsuit against the U.S. federal government, the Department of Defense, telecommunications companies, and defense companies.
Makes whatever this effort is sound deeply unserious. Is that the intent?
1. I agree that ads aren't usually informative enough. Our ad policy could emphasize informative ads.
2. You say you learn of them through your friends. How do they learn of the products? (If you say, through their friends, then where do their friends learn about it?)
3. Regarding automatically identifying and tracking you, we have mocked up an interface in the Profile section of the mockup (upper-right) whereby you can control what information is collected:
> You say you learn of them through your friends. How do they learn of the products?
I don't know, but it's possible that they could do it the same way I do: as a result of actively researching solutions to a problem I'm facing. Or maybe they see ads. However it enters their awareness isn't important to me.
Ads aren't important to me for product discovery. When I need a product, I go looking for it. I don't need them to come looking for me.
> How do you feel about that as a middle ground?
In order to target ads to me, someone needs to be tracking me. Whether it's the advertisers directly or a middleman is not an important distinction for me. In order for me to get paid, someone will have to know my identity and how to pay me, and records will be kept about what ads I've paid attention to and what ads I haven't. I personally find that unacceptable. Is your system able to avoid that?
I think so, Facebook and Google ads are frequently how people first learn of products. However, we want to step away from the model of monetizing personal information, while retaining the benefits of informed consumers.
>I personally find [tracking] unacceptable. Is your system able to avoid that?
Yes, our system can do anything that is possible, and you can literally vote for the prompts that you want. If it's physically possible to do it, then we can do it. Please remember that State of Utopia is not a business. Instead of trying to make profit for its shareholders, it is a sovereign nation and democracy whereby citizens (anyone who wants to be one, no charge, no catch) vote for the prompts they want AI to autonomously execute using state-run and state-owned companies. If you don't want the State of Utopia's companies to build profiles on you, you just vote against it and AI (which by the way the State of Utopia develops as sovereign independent models that the state itself owns) will make sure to follow its prompt. There will be transparency over this.
We don't have voting set up yet and only have a rudimentary AI model that cannot autonomously complete tasks or run companies yet, so this is very much a work in progress.
Our vision is sending you a free bank card we put money on, and you can just spend free money on whatever you want, never having to work since AI does all the work. The issue we're currently tackling is how will you know what to spend your money on? How will you know about the goods and services we make available?
It sounds like your answer defers this to times that you are actively searching for products.
Since we intend to have state-run companies in every industry, that includes search as well and we can include ads there.
Are you all right with our ads showing up in your search results when using our state search engine to browse for products that meet your needs, and if you can see that these ads do not track you? (By literally seeing the prompt for it.)
> Please remember that State of Utopia is not a business.
I'm not talking about the "State of Utopia". I'm talking about advertising. Since I don't really have a clue what this "State of Utopia" is about, if that's the real topic of discussion, then I have nothing useful to add.
> It sounds like your answer defers this to times that you are actively searching for products.
Yes.
> Are you all right with our ads showing up in your search results when using our state search engine to browse for products that meet your needs, and if you can see that these ads do not track you?
I would accept that such ads wouldn't have serious ethical issues, if that's what you mean. But, honestly, I wouldn't pay attention to ads. When I'm looking for a solution to some problem I have, a search engine isn't the only (sometimes not even the primary) method I use and I ignore any obvious advertising that may turn up in the course of it. Sales pitches are not useful to me for this.
>Google is probably the most undervalued tech company there is currently, by far: [reasons]
The only thing you left out of this analysis is their valuation. The market values Google at $2.05T (just over $2,000,000,000,000) which is 21 times their earnings (net profit). They are valued at $250 per person on Earth while selling, annually, $43.75 per person on Earth (sales) of which $12 per person is their profit.
How much would you pay to own a golden goose laying $12 in gold per year? Like, $250? If so you are the proud buyer of Google right now. (There is a buyer on every sale of every stock and this is the price they are paying right now.)
An alternative viewpoint is the consideration of the P/E of all of the Mag 7. These numbers might be slightly off since there's been a lot of market movement lately, but...
Apple (AAPL): 34.07
Microsoft (MSFT): 35.07
Amazon (AMZN): 36.69
Alphabet (GOOGL): 21.82
Meta Platforms (META): 24.49
Nvidia (NVDA): 41.33
Tesla (TSLA): 87.87
from this perspective Google, and to a lesser extent Meta, stand out as being valued quite conservatively.
Do I think Microsoft is performing 50% better than Google? Not really, no.
If the goose is likely to live for significantly longer than 20 years and has potential to lay $15 or $20 in the future then yes I'd probably buy that goose for $250. Of course there's risk with it (eg. Google might significantly lose business to competitors) but that's why you diversify. A PE of 20 for a mature company like Google isn't crazy. Even Coca Cola has a high PE at 28.
While you're debugging using AI (specifically, ChatGPT o1), you can benefit from copying stack traces. It debugs better than if you just describe what's wrong.
Another tip: I have found that it is helpful to ask AI to "deeply analyze" (use those words) and think about the problem without providing a solution (say "don't reply with any code"). If you don't do that, it will take its first guess and then eagerly start outputing code that is still wrong and doesn't really identify or fix the issue. When you ask it to deeply analyze what's wrong and not reply with any code, it frequently finds the true underling problem, and then you can ask for how to solve it in the next step.
https://github.com/anthropic-experimental/agentic-misalignme...
reply