OK, there's a lot of negative comments here, but this is a great idea at a high level. It really is hard to figure out where to do a thing and it's also very easy to get phished. If they can figure out how to help people get all the services they're eligible for that'll be an amazing improvement for the people.
I can see how LLMs might be useful for this kind of thing but I'm struggling to think of how this will be better than just asking ChatGPT the same questions.
The first suggestion I saw was "I just got married, how do I change my last name?"
It tells you to update your Social Security info first and gives you some helpful links.
ChatGPT tells me the same things with some of the same links. It didn't even have to ask if I was in America because it found my state based off my IP I assume.
Yeah that's a fair point. We shouldn't have to rely on private companies for this kind of thing.
On the other hand this is just going through one or more of the big AI companies anyway. So they still get data from users, now we're just funding them with our taxes instead of using one of many LLM interfaces available for free.
The privacy policy actually looks OK though, I'll be interested to see what the EFF has to say.
I’m a bit torn on that matter. I haven’t looked into it, but I suspect it’s not some kind of specialized foundational model trained to be a federal agent.
It’s likely just a mask of either ChatGPT and probably through some massively bloated, nepotistic federal contract. There may be some additional guidelines and instructions but it’s probably not RAG on anything that isn’t publicly slay accessible to any other model, even if it is hard limited to federal sites/domains.
In addition to what others have said, ChatGPT is also available free and without an account I believe. What are you more likely to think of when you have a question, your AI service you already use and have your information in and maybe even has access to your documents, or are you going to remember to go to America.gov for basic chat AI?
Frankly, it would be really interesting to benchmark it against the other models when it comes to accomplishing things related to the government, i.e. Which information is more accurate, useful, and actionable.
There’s also another matter I find a bit troubling depending on the nature of how this came about, it could also be something that creates an inextricable incumbency in whichever AI company is behind it, i.e., how Microsoft has become a kind of parasite on government and thereby on corporations for many decades now.
1. Americans pay for it with tax dollars and personal data.
2. US won't run their own local model they simply use one of the three who's the coziest with the government paying it $$$ and user data. And there was no request for proposals or public tender.
>Americans pay for it with tax dollars and personal data.
From the privacy policy[1], "Our AI providers operate under Zero Data Retention (ZDR) agreements and retain none of your prompts or the responses they generate. America.gov’s own response cache is separate and is described in Section 5.
Our AI providers are contractually prohibited from selling your information, building an advertising profile, or training their AI models with your data."
I know what you're saying, but OpenAI is an independent, profit driven company. They will make decisions that are in the best interests of the company, not the US. You really need something that is for the people (taxpayers), by the people (gov).
The whole industry is dealing with this now. You think current AI is not reliable enough for a task, so you augment it with know how to fix the problem, thoroughly test it, then decide to release it as a product. In the six months you took to do that, models improved enough that they work fine out of the box on your task.
In a sane world the US government’s chatbot would have the advantage that, as a representative of the government that they proactively decided to provide, it would at least be able to speak authoritatively on the government’s position on the law. Apparently this isn’t the case based on other comments, though, so it seems sort of pointless?
maherbeg · · focus · HN ↗
why_at · · focus · HN ↗
The first suggestion I saw was "I just got married, how do I change my last name?"
It tells you to update your Social Security info first and gives you some helpful links.
ChatGPT tells me the same things with some of the same links. It didn't even have to ask if I was in America because it found my state based off my IP I assume.
cousinbryce · · focus · HN ↗
nxobject · · focus · HN ↗
m348e912 · · focus · HN ↗
Levitating · · focus · HN ↗
why_at · · focus · HN ↗
On the other hand this is just going through one or more of the big AI companies anyway. So they still get data from users, now we're just funding them with our taxes instead of using one of many LLM interfaces available for free.
The privacy policy actually looks OK though, I'll be interested to see what the EFF has to say.
nightski · · focus · HN ↗
jonhohle · · focus · HN ↗
tyrabound · · focus · HN ↗
It’s likely just a mask of either ChatGPT and probably through some massively bloated, nepotistic federal contract. There may be some additional guidelines and instructions but it’s probably not RAG on anything that isn’t publicly slay accessible to any other model, even if it is hard limited to federal sites/domains.
In addition to what others have said, ChatGPT is also available free and without an account I believe. What are you more likely to think of when you have a question, your AI service you already use and have your information in and maybe even has access to your documents, or are you going to remember to go to America.gov for basic chat AI?
Frankly, it would be really interesting to benchmark it against the other models when it comes to accomplishing things related to the government, i.e. Which information is more accurate, useful, and actionable.
There’s also another matter I find a bit troubling depending on the nature of how this came about, it could also be something that creates an inextricable incumbency in whichever AI company is behind it, i.e., how Microsoft has become a kind of parasite on government and thereby on corporations for many decades now.
dzhiurgis · · focus · HN ↗
Good luck making it apolitical
doc_ick · · focus · HN ↗
ShinyLeftPad · · focus · HN ↗
2. US won't run their own local model they simply use one of the three who's the coziest with the government paying it $$$ and user data. And there was no request for proposals or public tender.
bhelkey · · focus · HN ↗
From the privacy policy[1], "Our AI providers operate under Zero Data Retention (ZDR) agreements and retain none of your prompts or the responses they generate. America.gov’s own response cache is separate and is described in Section 5.
Our AI providers are contractually prohibited from selling your information, building an advertising profile, or training their AI models with your data."
[1] <a href="https://america.gov/privacy-policy" rel="nofollow">https://america.gov/privacy-policy
ShinyLeftPad · · focus · HN ↗
gabbagool · · focus · HN ↗
Edit: Nevermind, someone made this point already.
shusaku · · focus · HN ↗
bee_rider · · focus · HN ↗