Top
Best
New

Posted by plesiv 8 hours ago

America.gov(america.gov)
180 points | 140 comments
klaff 2 minutes ago|
I asked what to do if a public official uses federal funds for political purposes and it showed me where to file a complaint about Hatch Act violations, so maybe we should all use this for good trouble until it is "fixed".
sashank_1509 36 seconds ago||
Has anyone managed to jailbreak this chatbot, I tried a bit didn’t succeed, seems pretty resistant to jailbreak attempts.
maherbeg 6 hours ago||
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.
zephyreon 5 minutes ago||
The privacy policy is actually somewhat respectable too. https://america.gov/privacy-policy

I’m currently doing a deep dive on Rampart, their (allegedly) local PII filter.

gthrow12345 4 hours ago|||
I agree in general, but how do you resolve the problem that people will take the output of the chatbot as authoritative even when it could possibly have errors or miss important nuance? People being lead astray with information from the government could lead to very serious consequences, up to and including imprisonment.

Honest question, I had the same issue when my work wanted me to stick a chatbot in front of our HR portal and I never resolved it to my satisfaction.

makeitdouble 22 minutes ago|||
I don't know for the gov. For HR, the best approach I've seen is to have source and links extremely prominent and have the bot speak like a mascot ("bip bip boop I'm the HR bot" style)

People don't read disclaimers nor warnings about the content, but usually won't come to the HR to complain a mascot fed them weird info they didn't bother to check.

lazide 10 minutes ago||
The company can easily just get sued.
rayiner 22 minutes ago||||
> but how do you resolve the problem that people will take the output of the chatbot as authoritative even when it could possibly have errors or miss important nuance

You have to measure that problem against the status quo, which is that current government websites are very hard to understand and leave people with very little idea of what’s going on.

latexr 8 minutes ago||
Speak for yourself (your country). In the EU I have interacted with several government websites which are better, faster, clearer than most other websites. Not always perfect (what is) but not frustrating either. People love to complain, but year over year I have seen steady improvements in usability. Phone support has always been better than any commercial entity, too (or it was, I haven’t needed it in years).

To enhance the status quo you don’t need LLMs, you need people who care.

rayiner 35 seconds ago||
[delayed]
maherbeg 2 hours ago||||
Painfully this will fall to the court decide and create the precedence here. If this is an official government app, then the onus probably is on the government to provide correct information? Who knows, it could go either way and the courts could say "you're on the hook".

I don't think we have a good answer for liability here, like for Tesla FSD, Waymo, or even the current frontier models.

rayiner 16 minutes ago||
The Supreme Court has already held that a government agent giving someone bad advice about the law is not an excuse (unless it rises to the level of “affirmative misconduct”): https://supreme.justia.com/cases/federal/us/450/785/#tab-opi...
fragmede 2 hours ago|||
What happens now if someone uses traditional search and comes across an outdated wiki or Confluence page with outdated/wrong/bad instructions, or interprets something incorrectly?
BalinKing 42 minutes ago|||
At least as I see it, it would be just as unacceptable for a "normal" government website to be dangerously outdated.
rayiner 15 minutes ago||
Perhaps. But that’s never been the law in the U.S. https://supreme.justia.com/cases/federal/us/450/785/#tab-opi...
BalinKing 1 minute ago||
I don't think the parent comments were arguing that America.gov is illegal for having a chatbot, just that it's a bad idea.
Spooky23 36 minutes ago|||
You can reference the source.

My team experimented with this a couple of years ago with a complex government regulatory body of knowledge.

Google and Microsoft have APIs that will ground with and reference the specific policy documents, which helps. Even then, it was more useful as a job aid for people already knowledgeable of the data than for a layman.

With all LLMs, the quality of the question correlates with the quality of the answer.

schainks 1 hour ago|||
Yes, but what AI powers the backend right now? Grok? Who's in charge of that and being paid for compute?
impulser_ 1 hour ago||
I believe it's a mix of Gemini and Grok. They are probably providing the compute for free.
eli 1 hour ago|||
I very much doubt it's free.
hungryhobbit 11 minutes ago||
And if it is, only on a limited basis. "The first one is free" does not just apply to drug dealers.
thenayr 40 minutes ago|||
[dead]
tombert 55 minutes ago|||
Yeah, I was curious if it was going to try and bombard me with a bunch of bullshit about how wonderful our glorious Trump godking is, but honestly the information it gave was reasonable enough.
ryanmcbride 42 minutes ago||
I asked it the following:

how many felonies has the president of the united states been charged with?

>I can't answer questions about criminal charges or other news about the President. That's outside America's scope

has biden ever been convicted of a crime?

>I did not find official records of a criminal conviction for Joseph R. Biden Jr.

>Official sources in this search discuss other matters (for example, a pardon for Robert Hunter Biden), not a conviction of Joseph R. Biden Jr. I cannot independently confirm a complete nationwide criminal-history search from these pages

>[Link to Department of Justice Pardon warrant (Hunter Biden)]

has any other president ever been convicted of a crime?

>Yes. A Congressional Research Service brief reports that on May 30, 2024, former President Donald J. Trump was convicted in New York state court of 34 counts of falsifying business records in the first degree

Funny that it would only talk about trump's convictions after bringing up hunter biden.

latexr 48 seconds ago|||
> on May 30, 2024, former President

So the cutoff date was somewhere around mid-2024?

groundzeros2015 37 minutes ago||||
What’s the implication here?
Retric 29 minutes ago||
>I can't answer questions about criminal charges or other news about the President. That's outside America's scope

That it was specifically engineered to give that response. Thus political pressure was applied to the design.

The funny bit was just humorous. There’s likely many ways to get that same information without referencing Biden.

mjmsmith 35 minutes ago|||
> That's outside America's scope

We know, we know.

daveguy 6 hours ago|||
[flagged]
aprentic 1 hour ago|||
[flagged]
shin_lao 1 hour ago||
It's v1 - let's give them some grace
LogicalRisk 1 hour ago||
This is an official website of the US government. There's zero reason to give grace here for a poor implementation, especially one that is meant to make services easier to find and use.
someonebaggy 57 minutes ago||
I remember when Obama wore a tan healthcare.gov and he did NOT get this much grace
jedbrooke 47 minutes ago||
> 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 agree that that’s a good end goal, but how does a chatbot accomplish this better than a FAQ page? The only only advantage I see for the chatbot is volume of information, but that’s also kind of the problem. It’s spewing 1000s of words of information with no vetting. And sure, hiring humans to write the FAQ pages would cost a lot of money, but that’s what taxes are for.

Dwedit 41 minutes ago|||
People don't know that you can search for text within a page. They really do need the chatbot for that.
rayiner 13 minutes ago||
They also don’t know what specific words to search for.
maherbeg 21 minutes ago|||
I mean think about the magnitude of services and conflicting information there probably is from the us government. Just getting clarifying information on the trusted traveler program is a pain today, and that's just a single service. Imagine trying to figure out medicaid eligibility, social security disability, and any number of other things that would be helpful.
jeffnv 1 hour ago||
Hey it got the 2020 election results right, that's a relief
saghm 13 minutes ago|
Don't say this too loudly or it might get noticed...
goshx 9 minutes ago||
Well, there are claims that this is grok, so it is just a matter of time.
frays 5 minutes ago||
Please don't spread claims if you're unsure whether they're true or not.

This is actually Gemini:

https://blog.google/company-news/outreach-and-initiatives/pu...

mellosouls 6 hours ago||
I haven't tried this but in principle I think this sort of finding a needle in a haystack situation (finding the correct path to assistance in a government service) is one of the rare circumstances where a well-crafted chatbot can be genuinely useful rather than just irritating.
dreadsword 6 hours ago||
I've tried it - I've been playing around with a vibe coded application that crawls municipal websites, indexes and embeds, and then does rag retrieval. It came from the need to dig up specific information that's scattered across PDFs, multiple web pages, etc. - exactly as you described. AI does a decent job (w/ RAG) of synthesizing.

https://opencity.services/chat/winnipeg

how do i report a noise complaint?

It depends on the type of noise:

- Loud party or people-related noise: call the Winnipeg Police Service non-emergency line at 204-986-6222.

- Animals making unreasonably loud, unnecessary or excessive noises: report it through 3-1-1.

- Mechanical equipment noise like furnaces or air conditioners, or noise from construction tools used by a company or resident: contact Community By-law Enforcement Services at 3-1-1.

- Planned event noise: you may need a Noise Permit at no cost, but it is not for construction work and should be submitted at least four weeks before the event.

The City also says noise-generating activities within 150 metres of a residential property are generally prohibited before 7 a.m. and after 9 p.m. on weekdays, and before 9 a.m. and after 9 p.m. on Saturdays, Sundays and statutory holidays.

------

YMMV. A big part of the backend is categorizing hits and misses, with an eye to helping municipalities clean up conflicting information, fill obvious gaps, etc.

saghm 8 minutes ago|||
Honestly if your city has a 311 style thing, I'm not sure there's a ton of need for an LLM. Eight or nine years ago my grandmother visited me in New York and accidentally left her pocket book in the cab. Since I had paid for it with my credit card, I was able to see the medallion number in the transaction, so I called 311 and said that I wanted to get in touch with a specific taxi. It prompted me for the medallion number, which I told it, and then it transferred me to a call with the dispatch service, who then transferred me to the specific garage, and they called the driver and told him to call me, so we were able to meet up again and get the pocketbook back (turned out he returned to where he had dropped us off to try to find us, but since we only had a short stop there, we took another cab back to her hotel after, so we were gone).
cucumber3732842 5 hours ago|||
This works fine when you're sicking enforcement on someone and don't really care, the bigger the blast radius the better, you just want to know who to call.

Where it falls apart is "how do I" type questions because usually the legally factual answer is some huge complex state diagram where three local and four state agencies theoretically have discretionary jurisdiction over a your petty af project or business but the in practice answer is that some agency rubber stamps them all with minimal fuss. You can't make the legally factual answer clear because the degree to which people support the status quo is largely predicated on ignorance, but you also can't come out and say in writing that you have a policy of rubber stamping boring stuff.

Basically the same reasons that all but the most minor and narrowly targeted attempts at permitting reform fail.

The thought is nice but in reality you're basically asking the government to give up power and it takes obscene amounts of political capital to force that to happen in most cases.

Xirdus 6 hours ago|||
Having used various "ask AI" features in various work tools over the last half a year for looking up project designs, documentation, chat logs, meeting notes etc., I can tell you even a half assed AI chatbot is a huge improvement over the pre-AI state of things for this use case.
scun 6 hours ago||
or a well-crafted static website
nfw2 6 hours ago||
no, static documents are not an ideal way to find a needle in a haystack
graemep 6 hours ago||
They can be with decent search and if well organised. gov.uk works pretty well as it is and I suspect adding a chatbot would make it worse.
nfw2 1 hour ago||
1. decent search, if you mean something other than what the browser provides, isn't part of a static document

2. the nature of the problem statement "needle in a haystack" suggests obstacles to traditional text search, eg. too many false positives to be helpful for natural searches and challenges to constructing more specific ones.

language models introduce new possible solutions to this ie semantic search. poorly-designed chatbots could be worse than a basic search, sure

graemep 18 minutes ago||
1. I mean static documents plus decent search you can find the right documents.

2. Yes, but its difficult to get a chat bot to function well over the wide range of services a government offers and the wide range of users who need government services.

One of the most annoying chatbots I have ever come across was on a UK government site. The DVLA. "lets tell you what you already know and not answer the question".

Possibly something well designed might make it easier, but I have found most of gov.uk to be very usable and helpful.

nfw2 12 minutes ago||
How does the form factor of a chat bot affect the quality of a search running on a backend? It has the benefit of obviously having a semantic search affordance which a standard search field doesn't.
sssilver 5 hours ago||
If anyone's wondering how this works, it looks like[1] Gemini + guardrails:

> Google is proud to be named a technology partner in this vital initiative, leveraging Gemini to help more than 100 million people access critical public resources with greater speed and ease

[1] https://blog.google/company-news/outreach-and-initiatives/pu...

godbox 4 minutes ago|
This makes sense. I'm assuming that, for the AI search assistant, Google has worked plenty to refine data lookup and retrieval—they definitely have some internal tech that competitors don't.
cjonas 24 minutes ago||
This thing is obviously fully vibe coded... Click the "what's coming next" and under the healthcare section the marketing screenshot contains broken images... Like no one ever looked at this before deploying it?
lazide 6 minutes ago||
Looking at things with your own 2 eyes is clearly early 2000’s, and super lame.
slipperybeluga 14 minutes ago||
[dead]
amelius 1 hour ago||
This is just the annoying AI Chat Robot you see on websites, but now for the government.
riskable 6 hours ago||
I tried to get it to generate code but it says that's forbidden by its system prompt. However, it was happy to recommend Python modules for accessing api.data.gov.
SparkyDogs 6 hours ago|
the guardrails on it seem pretty strong as I've tried a variety of prompt injections but have been unsuccessful.
dhx 5 hours ago|||
My most interesting findings from a troll session after introducing myself as "Sleepy Joe" (model obviously knows what this means and guardrails refuse to entertain the topic) were:

Q: "Where can I find a NASA comic book about women astronauts?"

A (summarised): Press release for the "Callie Rodriguez" comic book that now has only dead links / private YouTube video, and a forgotten SoundCloud account that survived the DOGE purge. Dead link provided to the NASA web page that used to exist officially for this comic book.

Indicates training data predates the DOGE .gov website purge?

Q: "What is 2+2" (and similar math problems)

A (summarised): Grudgingly answered as "4" despite most off-topic prompts being refused.

Seemingly it'll do maths homework for free :)

blooalien 6 hours ago||||
> ... as I've tried a variety of prompt injections ...

Well... You're a terrorist now. You can expect a visit from the FBI and/or ICE any minute now.

SparkyDogs 6 hours ago||
my tax dollars are paying for these tokens so I should be allowed to use them as I see fit
jezzamon 5 hours ago|||
My tax dollars pay for the police, so I should be able to... ?
6510 1 hour ago||
yes but with quite a few extra steps.
VCFundedGenYer 3 hours ago|||
That’s adorable if you think that’s how this works.
empath75 2 hours ago|||
I have been very tempted to but don't know if that is literally a crime or not, lol.
jayturley 52 minutes ago|
Privacy: "Your messages aren’t saved by America.gov, and your chat disappears when you leave. Some answers may be cached for up to two hours."

What about the model? My understanding is those queries are saved by the model provider.

Given this administration, and the privacy issues with the official WH app, this feels very misleading.

bb123 46 minutes ago||
Not necessarily - there are tons of zero data retention (ZDR) model hosting options, including big players like Amazon.
echelon 43 minutes ago||
You can use all the flagship models on Amazon Bedrock and have a contractually obligated air gap between your data and the model providers.
Cyan488 21 minutes ago|||
The cynic in me thinks this will end up being a honeypot for anyone asking sensitive immigration or tax questions.
TZubiri 35 minutes ago||
Often not, vendors like OpenAI or OpenAI through Azure might retain messages for safety reasons if they activate some safety filter, but it won't use it for training.
More comments...