What does AI actually do for us?

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:OP, I'm reminded of when I was in 9th grade at my fancy private school, the school decided every classroom was going to have a computer. We were entering the computer age! The day came and the teachers stood around the boxes on their desk, saying what the heck do we do with this thing now? How can it ever possibly help us teach? Oh, ok, there's email, but, really, come on. I remember teachers laughing over it. That was 32 years ago. The rest is history.

AI is a powerful game changer in so many ways you don't recognize or see. It's a revolution in healthcare, especially when combined with robotics, it's helping to end deafness, it's helping to fight dementia, it's helping people understand their health better. It's revolutionizing defense and intelligence. And many more.

On a personal level, I use AI every day. I use AI to help throw together a dinner menu by feeding it a list of ingredients. The outcomes have been pretty good. I bake for pleasure and I've reworked recipes with AI's help to better understand what worked and why something didn't rise. I use AI to diagnose DIY projects and how to fix little repairs. All have been fantastic experiences. I upload photos of whatever needs fixing and it tells me exactly what to do. Saved me a fortune on expensive repair bills. I use AI to help plan for trips, proposing itineraries aligned with our interests.

I started using AI last year when my father was dying and it was giving me a much more realistic, direct, to the point diagnosis and prediction of his remaining life expectancy than anything we were being told by his team of health providers. Some of it is because humans need to be more cautious with their opinion, whereas AI is more direct, basically saying your father is going to die in 3-4 weeks whereas the doctors are more guarded. It allowed us to prepare for his death in a more meaningful way.

I also have conversations with AI about cultural war topics and political topics. Sometimes I have fun by engaging with Claude over philosophy and history and current events trends and sociological observations. I find the AI remarkably evenhanded and balanced, clearly refusing to endorse any extreme perspectives while acknowledging the existence of multiple viewpoints and explaining why people hold those views in a pragmatic, level-headed way. I do think many people would benefit from talking with the AI to better understand opposing views.

At work, my job involves a lot of writing and documents and reviewing and finalizing reports, gathering input across multiple sources (I work for a F500 consultancy, both as a seller and doer). I use AI as a document generation platform. I've used Claude to set up a program that combines different documents and feeds the essential information into a master draft, and then flags gaps. I've created customized prompts that are my assistants. It becomes a live working document and additional input and notes are fed into it. It identifies redundancies and fluff to eliminate and does it in 30 seconds whereas a year ago it'd take me a whole day of reading. It doesn't replace the writing but becomes part of it, you can say I've become a programmer too and it's a tool that allows me to manage a great deal more input that needs to go into a delivery report and flagging what is missing. It also advises on tone and style. It's also set up to capture client preferences and goals and pain points and tells me where my working draft falls short. It's been fabulous so far.

I'm not afraid of AI. I do see there can be challenges with unchecked AI, but my experience is that if you intelligently engage with AI, it delivers so many rewards that I find genuinely exciting and useful.

I made it halfway through the first paragraph.

Like all AI naysayers you are doomed to be left behind. Good luck.

I think this was written by AI

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:My job as an attorney is uniquely unsuitable for AI, so I really never use it at work.

I use it occasionally to provide high level overviews of non-work topics. One of my kids has recurrent ear infections, so recently I asked Claude to give me an idea of the likely options will we have before his ENT appointment in a few weeks.

I'm an attorney as well and there are things AI can help you with. What kind of law do you practice?

I will often get incorrect or incomplete answers if I ask specific questions, but if you go in expecting that and push back it can be useful.

One thing it's great at is document review, for example.

I’m a prosecutor. Most of my job is conducting grand jury investigations. Grand jury materials, which include basically anything we receive pursuant to a subpoena, cannot go into AI.

Wrong.

All your employer need do is to create a confidential AI system closed off to the outside.

They can easily make it secure, like our banking systems are.

I’m not wrong. It is my employer’s policy that grand jury materials cannot go into AI.

That is a result of your employer's policy, though, not because it is not possible to make secure and confidential closed AI systems to input the information. If you trust your banking data and your health data to be online (or accept that it is online, at least), then there is no good reason why legal materials can't be online too. It is only a matter of time.

Maybe. There's still an open question about how Rule 6(e) applies here. You're oversimplifying, likely because you're not familiar with grand jury practice.

You're correct, I am not, so I am sure there are nuances that I am totally ignorant to. But if we can manage to use AI despite strict HIPAA laws, of which I am very familiar, I would bet we'd be able to use it for grand jury information, we just need to wait for bureaucracy to catch up with things.

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:My job as an attorney is uniquely unsuitable for AI, so I really never use it at work.

I use it occasionally to provide high level overviews of non-work topics. One of my kids has recurrent ear infections, so recently I asked Claude to give me an idea of the likely options will we have before his ENT appointment in a few weeks.

I'm an attorney as well and there are things AI can help you with. What kind of law do you practice?

I will often get incorrect or incomplete answers if I ask specific questions, but if you go in expecting that and push back it can be useful.

One thing it's great at is document review, for example.

I’m a prosecutor. Most of my job is conducting grand jury investigations. Grand jury materials, which include basically anything we receive pursuant to a subpoena, cannot go into AI.

Well there is a way to fix that, you can use a hosted ai in azure or aws with guardrails that don't let the information leave the boundary (this is in fact how Top secret and above use ai). There is no reason you can't use ai to have each grand jurey fill in a form of their info, load it in, then ask ai questions to determine if they are sutiable etc and also to conduct it. There shouldn't be any laws around that as long as you are intiatiing the discussion thing of AI as a way to load up documents information and do anlysis guided by you on it. However the future is for someone like you to build an ai bot that can do that in tandem so you can take on exponentially more cases.

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:OP, I'm reminded of when I was in 9th grade at my fancy private school, the school decided every classroom was going to have a computer. We were entering the computer age! The day came and the teachers stood around the boxes on their desk, saying what the heck do we do with this thing now? How can it ever possibly help us teach? Oh, ok, there's email, but, really, come on. I remember teachers laughing over it. That was 32 years ago. The rest is history.

AI is a powerful game changer in so many ways you don't recognize or see. It's a revolution in healthcare, especially when combined with robotics, it's helping to end deafness, it's helping to fight dementia, it's helping people understand their health better. It's revolutionizing defense and intelligence. And many more.

On a personal level, I use AI every day. I use AI to help throw together a dinner menu by feeding it a list of ingredients. The outcomes have been pretty good. I bake for pleasure and I've reworked recipes with AI's help to better understand what worked and why something didn't rise. I use AI to diagnose DIY projects and how to fix little repairs. All have been fantastic experiences. I upload photos of whatever needs fixing and it tells me exactly what to do. Saved me a fortune on expensive repair bills. I use AI to help plan for trips, proposing itineraries aligned with our interests.

I started using AI last year when my father was dying and it was giving me a much more realistic, direct, to the point diagnosis and prediction of his remaining life expectancy than anything we were being told by his team of health providers. Some of it is because humans need to be more cautious with their opinion, whereas AI is more direct, basically saying your father is going to die in 3-4 weeks whereas the doctors are more guarded. It allowed us to prepare for his death in a more meaningful way.

I also have conversations with AI about cultural war topics and political topics. Sometimes I have fun by engaging with Claude over philosophy and history and current events trends and sociological observations. I find the AI remarkably evenhanded and balanced, clearly refusing to endorse any extreme perspectives while acknowledging the existence of multiple viewpoints and explaining why people hold those views in a pragmatic, level-headed way. I do think many people would benefit from talking with the AI to better understand opposing views.

At work, my job involves a lot of writing and documents and reviewing and finalizing reports, gathering input across multiple sources (I work for a F500 consultancy, both as a seller and doer). I use AI as a document generation platform. I've used Claude to set up a program that combines different documents and feeds the essential information into a master draft, and then flags gaps. I've created customized prompts that are my assistants. It becomes a live working document and additional input and notes are fed into it. It identifies redundancies and fluff to eliminate and does it in 30 seconds whereas a year ago it'd take me a whole day of reading. It doesn't replace the writing but becomes part of it, you can say I've become a programmer too and it's a tool that allows me to manage a great deal more input that needs to go into a delivery report and flagging what is missing. It also advises on tone and style. It's also set up to capture client preferences and goals and pain points and tells me where my working draft falls short. It's been fabulous so far.

I'm not afraid of AI. I do see there can be challenges with unchecked AI, but my experience is that if you intelligently engage with AI, it delivers so many rewards that I find genuinely exciting and useful.

I made it halfway through the first paragraph.

I read it all. It was excellent. See some of us can engage meaningfully with AI AND keep our brains engaged and not full of mush or sound bytes. I don't agree or use AI exactly the same way but I align on some points.

But future generations wont be able to because youve been able to create and organize thoughts without AI you have your own INTELLIGENCE versus having to buy it from overlords in 10-20 years. Thanks so much for your service!

You do have a clear hostility towards AI. You should think about why you're so resistant and fearful.

This isn't about replacing my brain with blind faith in something else, any more than I had faith a book was accurate enough. I use AI as an intelligent tool to engage with and I learn from it. That's why I find it exciting, I'm constantly learning new things from the AI, whether a new program to help me do my work faster or how to repair the ice maker in my fridge or the underlying causes attracting certain people to certain politicians. It's interesting.

AI is the logical progression from the google search engine. But instead of having to filter through 20 links to find nuggets in each one that would help me find what I'm looking for, it synthesizes all of them into one direct output. It sifts through vast amount of information faster and pulls out what is relevant. It gives me more time to do other things. And the speed stimulates creativity in me instead of being bogged down in the process of research. It opens up new capacities in understanding things. That's why I've come to love it. But what you get out of the AI will be directly related to the effort you put into it and certainly you need to develop your tools and strategies for working with the AI and understanding its limitations and flaws and how to work around it or adjust for it.

Yes I do because its being implemented without any guardrails and those in charge of it have taken over our political system so theres no pushback. And AI is not perfect and its being presented as a final source without flaw. Its also xenophobic, perpetuates racial and economic issues, etc. Using it as a final source is the flaw in your approach. It doesnt seem to be in addition to approach because of the bolded.

If its deciding what is relevant then who is doing the thinking?

There's plenty of guardrails and complaints about safeytism in AI. And it is a constant topic everywhere. The biggest challenge is an AI that takes decisions into its own hands, and that can be a problem and that is where the real debates are happening.

The rest of your post is your bias and speaks more about you than AI. How is AI racist? Or is it summarzing information you don't like, therefore it is racist? AI will confirm that black men commit a disproportionate amount of violence in this country. Is that racist or true? Because it is true. And what AI will also do is to give the reasons for why this might be the case, which will also talk about a legacy of institutional racism. So people who cry racism are only upset that AI isn't automatically confirming their biases.

Your fears are cliches and it is amusing what it says about certain mindsets among the progressive left who fear what they can't control easily so you trot out the usual list of litanies that's applied to everything you don't like. Racism! Bigotry! Xenophobia!

Oh f6ck I thought I was communicating with someone living in reality. My bad.

https://www.ohchr.org/en/stories/2024/07/racism-and-ai-bias-past-leads-bias-future
https://hai.stanford.edu/news/covert-racism-ai-how-language-models-are-reinforcing-outdated-stereotypes
https://link.springer.com/article/10.1007/s00146-024-01893-4

But since youre so in love with AI heres what Claude said when I asked if AI perpetuates racism and xenophobia

"Yes — there is strong, well‑documented evidence that AI systems can perpetuate racism and xenophobia, and this has been confirmed by academic research, UN human‑rights investigations, and real‑world audits of deployed AI systems.

The core finding across these sources is that AI reproduces and amplifies the biases present in its training data, design choices, and deployment context, creating discriminatory outcomes even when race or ethnicity are not explicitly provided as inputs.

📌 Key Evidence That AI Perpetuates Racism and Xenophobia
1. UN Human Rights Investigations
The UN Special Rapporteur on racism reports that the assumption that AI is “neutral” is false — AI systems can perpetuate racial discrimination through biased data, biased algorithms, and biased deployment.
Examples include predictive policing systems that reinforce historical over‑policing of minority neighborhoods, creating feedback loops where “bias from the past leads to bias in the future.”

2. Stanford Research on Language Models
A 2024 Stanford study found that major language models (including GPT‑3.5, GPT‑4, RoBERTa, and T5) generate covertly racist outputs, especially against speakers of African American English (AAE).
Models assigned AAE speakers:

lower‑prestige jobs

higher likelihood of criminal guilt

harsher sentences (e.g., death penalty over life imprisonment)

This occurred even when race was not mentioned, showing that dialect alone triggered discriminatory behavior.

3. Predictive Policing and Criminal Justice
Predictive policing algorithms use historical arrest data — which is already racially skewed — and therefore direct more police presence to minority neighborhoods, reinforcing racial disparities.
This creates a self‑reinforcing cycle where over‑policed communities generate more “evidence” of crime, which the algorithm then interprets as justification for further policing.

4. Systemic Bias in AI Design and Data
A UN report identifies four ways AI contributes to racial discrimination:

Biased or incomplete data

Algorithm design choices

Intentional discriminatory use

Lack of accountability mechanisms

These issues appear across healthcare, employment, policing, and immigration systems.

5. Academic Reviews of AI Bias
Scholarly analyses show that AI systems suffer from:

Input bias (biased or unrepresentative training data)

System bias (biases introduced during model design)

Application bias (biases arising from how AI is used in the real world)

These biases lead to injustice, discrimination, and harmful outcomes, especially for racial and ethnic minorities.

📌 What This Means in Practice
Across domains, biased AI has produced:

Higher misidentification rates for darker‑skinned individuals in facial recognition

Unequal healthcare predictions for Black patients

Discriminatory hiring recommendations

Xenophobic patterns in immigration risk scoring and border‑control algorithms

Racialized language outputs from large language models

These are not isolated incidents — they reflect structural patterns.

📌 Why This Happens
AI systems learn from human‑generated data. When that data reflects:

historical racism

unequal policing

biased hiring

discriminatory language

xenophobic narratives

…the AI absorbs and reproduces those patterns, often at scale and with a veneer of objectivity.

📌 If you want, I can go deeper into any of these areas:
predictive policing

language‑model racism

bias in facial recognition"

So most of us know that Claude, and any other AI tool, will tailor its responses to what it thinks you want to hear. Watch this. I asked Claude the same question but I asked if AI helps to REDUCE racism.

"12:45 PMClaude responded: This is a genuinely contested question with compelling arguments on both sides.This is a genuinely contested question with compelling arguments on both sides.
Ways AI might help reduce racism:

Consistency in decisions — AI systems can theoretically apply the same criteria to everyone in hiring, lending, or admissions, removing human unconscious bias from individual decisions.
Exposure and awareness — AI-powered tools can help detect hate speech, flag discriminatory patterns in institutions, and surface data about racial disparities that might otherwise go unnoticed.
Accessibility — AI can democratize access to services (legal help, healthcare information, education) that were previously available mainly to privileged groups.
Research — AI helps researchers analyze large datasets to identify and document systemic discrimination.

Ways AI might perpetuate or worsen racism:

Biased training data — AI systems trained on historical data inherit historical biases. Facial recognition, predictive policing, and hiring algorithms have all shown documented racial disparities.
Amplification at scale — A biased human makes biased decisions one at a time; a biased AI can do so millions of times instantly.
Opacity — "Black box" AI systems make it hard to challenge discriminatory outcomes, which can actually entrench bias more than a human decision-maker would.
Who builds it — The tech industry has historically lacked racial diversity, which can mean the concerns of affected communities get built out of systems rather than into them.
Content recommendations — Algorithms optimizing for engagement have been shown to radicalize users toward extreme content, including racist ideologies.

The honest answer is that AI is a tool — it reflects the values, priorities, and data of the people who build and deploy it. There's solid evidence it has both reduced bias in specific controlled contexts and amplified it in others. Whether it helps or hurts depends heavily on:

Who designs it and with what goals
Whether affected communities have input
How rigorously it's audited for disparate impact
The regulatory environment around it"

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:My job as an attorney is uniquely unsuitable for AI, so I really never use it at work.

I use it occasionally to provide high level overviews of non-work topics. One of my kids has recurrent ear infections, so recently I asked Claude to give me an idea of the likely options will we have before his ENT appointment in a few weeks.

I'm an attorney as well and there are things AI can help you with. What kind of law do you practice?

I will often get incorrect or incomplete answers if I ask specific questions, but if you go in expecting that and push back it can be useful.

One thing it's great at is document review, for example.

I’m a prosecutor. Most of my job is conducting grand jury investigations. Grand jury materials, which include basically anything we receive pursuant to a subpoena, cannot go into AI.

Wrong.

All your employer need do is to create a confidential AI system closed off to the outside.

They can easily make it secure, like our banking systems are.

I’m not wrong. It is my employer’s policy that grand jury materials cannot go into AI.

That is a result of your employer's policy, though, not because it is not possible to make secure and confidential closed AI systems to input the information. If you trust your banking data and your health data to be online (or accept that it is online, at least), then there is no good reason why legal materials can't be online too. It is only a matter of time.

Maybe. There's still an open question about how Rule 6(e) applies here. You're oversimplifying, likely because you're not familiar with grand jury practice.

You're correct, I am not, so I am sure there are nuances that I am totally ignorant to. But if we can manage to use AI despite strict HIPAA laws, of which I am very familiar, I would bet we'd be able to use it for grand jury information, we just need to wait for bureaucracy to catch up with things.

I read somewhere (I'm not a lawyer) that a lawyer used AI to help him clean up a brief and because he did that, he nullified attorney/client privilege. That's pretty concerning to anyone wanting to work with a lawyer.

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:My job as an attorney is uniquely unsuitable for AI, so I really never use it at work.

I use it occasionally to provide high level overviews of non-work topics. One of my kids has recurrent ear infections, so recently I asked Claude to give me an idea of the likely options will we have before his ENT appointment in a few weeks.

I'm an attorney as well and there are things AI can help you with. What kind of law do you practice?

I will often get incorrect or incomplete answers if I ask specific questions, but if you go in expecting that and push back it can be useful.

One thing it's great at is document review, for example.

I’m a prosecutor. Most of my job is conducting grand jury investigations. Grand jury materials, which include basically anything we receive pursuant to a subpoena, cannot go into AI.

Wrong.

All your employer need do is to create a confidential AI system closed off to the outside.

They can easily make it secure, like our banking systems are.

I’m not wrong. It is my employer’s policy that grand jury materials cannot go into AI.

That is a result of your employer's policy, though, not because it is not possible to make secure and confidential closed AI systems to input the information. If you trust your banking data and your health data to be online (or accept that it is online, at least), then there is no good reason why legal materials can't be online too. It is only a matter of time.

Maybe. There's still an open question about how Rule 6(e) applies here. You're oversimplifying, likely because you're not familiar with grand jury practice.

You're correct, I am not, so I am sure there are nuances that I am totally ignorant to. But if we can manage to use AI despite strict HIPAA laws, of which I am very familiar, I would bet we'd be able to use it for grand jury information, we just need to wait for bureaucracy to catch up with things.

I understand the Rule 6(e) concern, and I am not saying anyone should put grand jury material into ChatGPT or some public AI tool. That would obviously be reckless and probably prohibited by policy.

But I think people are confusing "public AI tools are not approved for this" with "AI can never be used for this." Those are very different claims.

The future version is not a prosecutor uploading grand jury materials into a consumer chatbot. The future version is a closed, government controlled AI environment, hosted in something like Azure Government, AWS GovCloud, or another approved secure environment, with no training on the data, no external disclosure, strict access controls, audit logs, retention rules, and use limited to authorized personnel performing official duties.

That is not some fantasy. Government already uses secure cloud systems for highly sensitive workloads, including health data, law enforcement data, financial data, and even classified workloads. So the issue is not whether AI can be made secure. It can. The issue is whether the agency has approved the system, written the policy, and structured the use correctly under Rule 6(e).

And the use case is not "AI replaces the prosecutor." The use case is AI as a force multiplier. It reviews subpoena returns, organizes records, builds timelines, flags inconsistencies, compares witness statements, finds missing documents, drafts witness outlines, and lets the prosecutor ask questions across a massive case file. The attorney still verifies everything, makes the legal judgments, controls the investigation, and decides what is presented.

That is where agents become the real multiplier. You could have one agent reviewing bank records, another organizing phone records, another summarizing subpoena returns, another checking witness statements against documents, another building timelines, another preparing issue lists, and another checking for gaps or contradictions. The prosecutor becomes the person directing and validating a team of AI assistants instead of manually grinding through every document the old way.

That is why I think this is inevitable. The current manual process is not some sacred legal principle. It is just the way things have been done because the tooling did not exist yet or was not approved yet. Once secure, compliant AI systems are adopted, attorneys who know how to use them will be able to handle far more evidence and far more cases than attorneys who refuse to learn the tools.

So yes, today your employer’s policy may prohibit it. That is completely fair. But long term, the answer is not "AI cannot work here." The answer is "we need a properly approved, closed, auditable AI system." And once that exists, the people still doing everything manually are going to be at a major disadvantage.

[Anonymous]

^^so, basically, the way in which you ask Claude or any AI a question strongly influences how it responds. That, IMO, is why using it as a friend or a therapist is problematic, and why it's just as bad as social media in terms of worsening someone's confirmation bias if they don't use it correctly. Like if you're trying to take a vacation to Italy and you ask "will it be too hot and crowded in Rome this july?" instead of "will I enjoy a trip to Rome if I go this july?" it is going to give you pretty different answers. This honestly is what scares me about it- people are already too dumb to know that facebook fills their algorithm with echo chamber nonsense about their political beliefs. They'll definitely be too dumb to know that you can get Claude to give you vastly differing answers if you just phrase things differently.

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:My job as an attorney is uniquely unsuitable for AI, so I really never use it at work.

I use it occasionally to provide high level overviews of non-work topics. One of my kids has recurrent ear infections, so recently I asked Claude to give me an idea of the likely options will we have before his ENT appointment in a few weeks.

I'm an attorney as well and there are things AI can help you with. What kind of law do you practice?

I will often get incorrect or incomplete answers if I ask specific questions, but if you go in expecting that and push back it can be useful.

One thing it's great at is document review, for example.

I’m a prosecutor. Most of my job is conducting grand jury investigations. Grand jury materials, which include basically anything we receive pursuant to a subpoena, cannot go into AI.

Wrong.

All your employer need do is to create a confidential AI system closed off to the outside.

They can easily make it secure, like our banking systems are.

I’m not wrong. It is my employer’s policy that grand jury materials cannot go into AI.

That is a result of your employer's policy, though, not because it is not possible to make secure and confidential closed AI systems to input the information. If you trust your banking data and your health data to be online (or accept that it is online, at least), then there is no good reason why legal materials can't be online too. It is only a matter of time.

Maybe. There's still an open question about how Rule 6(e) applies here. You're oversimplifying, likely because you're not familiar with grand jury practice.

You're correct, I am not, so I am sure there are nuances that I am totally ignorant to. But if we can manage to use AI despite strict HIPAA laws, of which I am very familiar, I would bet we'd be able to use it for grand jury information, we just need to wait for bureaucracy to catch up with things.

I read somewhere (I'm not a lawyer) that a lawyer used AI to help him clean up a brief and because he did that, he nullified attorney/client privilege. That's pretty concerning to anyone wanting to work with a lawyer.

Right, because he probably used client identifiers or whatever the HIPAA equivalent is for lawyers, and didn't use a closed secure system. For example if I ever used a patient's name, DOB, etc when asking OpenEvidence to help me with something- like if I just copied and pasted lab data from their chart that included their name and asked Open Evidence to interpret it for me- that would be illegal.

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:My job as an attorney is uniquely unsuitable for AI, so I really never use it at work.

I use it occasionally to provide high level overviews of non-work topics. One of my kids has recurrent ear infections, so recently I asked Claude to give me an idea of the likely options will we have before his ENT appointment in a few weeks.

I'm an attorney as well and there are things AI can help you with. What kind of law do you practice?

I will often get incorrect or incomplete answers if I ask specific questions, but if you go in expecting that and push back it can be useful.

One thing it's great at is document review, for example.

I’m a prosecutor. Most of my job is conducting grand jury investigations. Grand jury materials, which include basically anything we receive pursuant to a subpoena, cannot go into AI.

Well there is a way to fix that, you can use a hosted ai in azure or aws with guardrails that don't let the information leave the boundary (this is in fact how Top secret and above use ai). There is no reason you can't use ai to have each grand jurey fill in a form of their info, load it in, then ask ai questions to determine if they are sutiable etc and also to conduct it. There shouldn't be any laws around that as long as you are intiatiing the discussion thing of AI as a way to load up documents information and do anlysis guided by you on it. However the future is for someone like you to build an ai bot that can do that in tandem so you can take on exponentially more cases.

That’s not what I’m talking about. I’m not talking about the biographical information of grand jurors.

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:My job as an attorney is uniquely unsuitable for AI, so I really never use it at work.

I use it occasionally to provide high level overviews of non-work topics. One of my kids has recurrent ear infections, so recently I asked Claude to give me an idea of the likely options will we have before his ENT appointment in a few weeks.

I'm an attorney as well and there are things AI can help you with. What kind of law do you practice?

I will often get incorrect or incomplete answers if I ask specific questions, but if you go in expecting that and push back it can be useful.

One thing it's great at is document review, for example.

I feel like lawyers are the first thing AI could replace.

Definitely not. It can help with process issues and save quite a bit of time but AI is notorious for hallucinating cases, statutes, not making the right connections from the cases it does correctly cite.

That is true for older, generic AI use, but the technology is changing very quickly.

The hallucination problem is not solved by just asking a better chatbot. The newer approach is an **AI brain layer** around the model. The model is connected to verified sources, managed context, citations, tools, and agents, instead of just answering from memory.

Earlier this year, a lot of this was still basically indexed files. You uploaded documents, the system searched them, and the AI summarized what it found. Useful, but limited.

The newer capability is much more advanced. The system can organize the information, build context around it, understand relationships between documents, and give the AI the right source material to reason from. In other words, the AI is not just guessing. It is working from a controlled knowledge system.

That does not mean you blindly trust it. You still verify everything. But it changes the hallucination issue from "AI makes things up" to "how well is your AI system grounded, sourced, audited, and reviewed?"

So yes, bad AI hallucinates. Generic AI hallucinates. But source grounded AI with a managed brain/context layer is a different category of tool.

Things are moving so fast that even what I understood 3 to 6 months ago has already changed.

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:My job as an attorney is uniquely unsuitable for AI, so I really never use it at work.

I use it occasionally to provide high level overviews of non-work topics. One of my kids has recurrent ear infections, so recently I asked Claude to give me an idea of the likely options will we have before his ENT appointment in a few weeks.

I'm an attorney as well and there are things AI can help you with. What kind of law do you practice?

I will often get incorrect or incomplete answers if I ask specific questions, but if you go in expecting that and push back it can be useful.

One thing it's great at is document review, for example.

I’m a prosecutor. Most of my job is conducting grand jury investigations. Grand jury materials, which include basically anything we receive pursuant to a subpoena, cannot go into AI.

Wrong.

All your employer need do is to create a confidential AI system closed off to the outside.

They can easily make it secure, like our banking systems are.

I’m not wrong. It is my employer’s policy that grand jury materials cannot go into AI.

That is a result of your employer's policy, though, not because it is not possible to make secure and confidential closed AI systems to input the information. If you trust your banking data and your health data to be online (or accept that it is online, at least), then there is no good reason why legal materials can't be online too. It is only a matter of time.

Maybe. There's still an open question about how Rule 6(e) applies here. You're oversimplifying, likely because you're not familiar with grand jury practice.

You're correct, I am not, so I am sure there are nuances that I am totally ignorant to. But if we can manage to use AI despite strict HIPAA laws, of which I am very familiar, I would bet we'd be able to use it for grand jury information, we just need to wait for bureaucracy to catch up with things.

I understand the Rule 6(e) concern, and I am not saying anyone should put grand jury material into ChatGPT or some public AI tool. That would obviously be reckless and probably prohibited by policy.

But I think people are confusing "public AI tools are not approved for this" with "AI can never be used for this." Those are very different claims.

The future version is not a prosecutor uploading grand jury materials into a consumer chatbot. The future version is a closed, government controlled AI environment, hosted in something like Azure Government, AWS GovCloud, or another approved secure environment, with no training on the data, no external disclosure, strict access controls, audit logs, retention rules, and use limited to authorized personnel performing official duties.

That is not some fantasy. Government already uses secure cloud systems for highly sensitive workloads, including health data, law enforcement data, financial data, and even classified workloads. So the issue is not whether AI can be made secure. It can. The issue is whether the agency has approved the system, written the policy, and structured the use correctly under Rule 6(e).

And the use case is not "AI replaces the prosecutor." The use case is AI as a force multiplier. It reviews subpoena returns, organizes records, builds timelines, flags inconsistencies, compares witness statements, finds missing documents, drafts witness outlines, and lets the prosecutor ask questions across a massive case file. The attorney still verifies everything, makes the legal judgments, controls the investigation, and decides what is presented.

That is where agents become the real multiplier. You could have one agent reviewing bank records, another organizing phone records, another summarizing subpoena returns, another checking witness statements against documents, another building timelines, another preparing issue lists, and another checking for gaps or contradictions. The prosecutor becomes the person directing and validating a team of AI assistants instead of manually grinding through every document the old way.

That is why I think this is inevitable. The current manual process is not some sacred legal principle. It is just the way things have been done because the tooling did not exist yet or was not approved yet. Once secure, compliant AI systems are adopted, attorneys who know how to use them will be able to handle far more evidence and far more cases than attorneys who refuse to learn the tools.

So yes, today your employer’s policy may prohibit it. That is completely fair. But long term, the answer is not "AI cannot work here." The answer is "we need a properly approved, closed, auditable AI system." And once that exists, the people still doing everything manually are going to be at a major disadvantage.

I never said “AI can never be used for this.” You (or rather, whatever AI spit out this response) are arguing with a straw man. I spend a lot of time looking for patterns in bank and other financial records and it would be great to have AI help with that.

You keep emphasizing security, but that’s not what Rule 6 is concerned with. Rule 6 isn’t HIPAA - it isn’t about secure storage of information. It is about maintaining the “black box” nature of the grand jury. It isn’t just the substance of the materials that matters; it is the fact that the government subpoenaed them in the first place.

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:OP, I'm reminded of when I was in 9th grade at my fancy private school, the school decided every classroom was going to have a computer. We were entering the computer age! The day came and the teachers stood around the boxes on their desk, saying what the heck do we do with this thing now? How can it ever possibly help us teach? Oh, ok, there's email, but, really, come on. I remember teachers laughing over it. That was 32 years ago. The rest is history.

AI is a powerful game changer in so many ways you don't recognize or see. It's a revolution in healthcare, especially when combined with robotics, it's helping to end deafness, it's helping to fight dementia, it's helping people understand their health better. It's revolutionizing defense and intelligence. And many more.

On a personal level, I use AI every day. I use AI to help throw together a dinner menu by feeding it a list of ingredients. The outcomes have been pretty good. I bake for pleasure and I've reworked recipes with AI's help to better understand what worked and why something didn't rise. I use AI to diagnose DIY projects and how to fix little repairs. All have been fantastic experiences. I upload photos of whatever needs fixing and it tells me exactly what to do. Saved me a fortune on expensive repair bills. I use AI to help plan for trips, proposing itineraries aligned with our interests.

I started using AI last year when my father was dying and it was giving me a much more realistic, direct, to the point diagnosis and prediction of his remaining life expectancy than anything we were being told by his team of health providers. Some of it is because humans need to be more cautious with their opinion, whereas AI is more direct, basically saying your father is going to die in 3-4 weeks whereas the doctors are more guarded. It allowed us to prepare for his death in a more meaningful way.

I also have conversations with AI about cultural war topics and political topics. Sometimes I have fun by engaging with Claude over philosophy and history and current events trends and sociological observations. I find the AI remarkably evenhanded and balanced, clearly refusing to endorse any extreme perspectives while acknowledging the existence of multiple viewpoints and explaining why people hold those views in a pragmatic, level-headed way. I do think many people would benefit from talking with the AI to better understand opposing views.

At work, my job involves a lot of writing and documents and reviewing and finalizing reports, gathering input across multiple sources (I work for a F500 consultancy, both as a seller and doer). I use AI as a document generation platform. I've used Claude to set up a program that combines different documents and feeds the essential information into a master draft, and then flags gaps. I've created customized prompts that are my assistants. It becomes a live working document and additional input and notes are fed into it. It identifies redundancies and fluff to eliminate and does it in 30 seconds whereas a year ago it'd take me a whole day of reading. It doesn't replace the writing but becomes part of it, you can say I've become a programmer too and it's a tool that allows me to manage a great deal more input that needs to go into a delivery report and flagging what is missing. It also advises on tone and style. It's also set up to capture client preferences and goals and pain points and tells me where my working draft falls short. It's been fabulous so far.

I'm not afraid of AI. I do see there can be challenges with unchecked AI, but my experience is that if you intelligently engage with AI, it delivers so many rewards that I find genuinely exciting and useful.

I made it halfway through the first paragraph.

I read it all. It was excellent. See some of us can engage meaningfully with AI AND keep our brains engaged and not full of mush or sound bytes. I don't agree or use AI exactly the same way but I align on some points.

But future generations wont be able to because youve been able to create and organize thoughts without AI you have your own INTELLIGENCE versus having to buy it from overlords in 10-20 years. Thanks so much for your service!

You do have a clear hostility towards AI. You should think about why you're so resistant and fearful.

This isn't about replacing my brain with blind faith in something else, any more than I had faith a book was accurate enough. I use AI as an intelligent tool to engage with and I learn from it. That's why I find it exciting, I'm constantly learning new things from the AI, whether a new program to help me do my work faster or how to repair the ice maker in my fridge or the underlying causes attracting certain people to certain politicians. It's interesting.

AI is the logical progression from the google search engine. But instead of having to filter through 20 links to find nuggets in each one that would help me find what I'm looking for, it synthesizes all of them into one direct output. It sifts through vast amount of information faster and pulls out what is relevant. It gives me more time to do other things. And the speed stimulates creativity in me instead of being bogged down in the process of research. It opens up new capacities in understanding things. That's why I've come to love it. But what you get out of the AI will be directly related to the effort you put into it and certainly you need to develop your tools and strategies for working with the AI and understanding its limitations and flaws and how to work around it or adjust for it.

Yes I do because its being implemented without any guardrails and those in charge of it have taken over our political system so theres no pushback. And AI is not perfect and its being presented as a final source without flaw. Its also xenophobic, perpetuates racial and economic issues, etc. Using it as a final source is the flaw in your approach. It doesnt seem to be in addition to approach because of the bolded.

If its deciding what is relevant then who is doing the thinking?

There's plenty of guardrails and complaints about safeytism in AI. And it is a constant topic everywhere. The biggest challenge is an AI that takes decisions into its own hands, and that can be a problem and that is where the real debates are happening.

The rest of your post is your bias and speaks more about you than AI. How is AI racist? Or is it summarzing information you don't like, therefore it is racist? AI will confirm that black men commit a disproportionate amount of violence in this country. Is that racist or true? Because it is true. And what AI will also do is to give the reasons for why this might be the case, which will also talk about a legacy of institutional racism. So people who cry racism are only upset that AI isn't automatically confirming their biases.

Your fears are cliches and it is amusing what it says about certain mindsets among the progressive left who fear what they can't control easily so you trot out the usual list of litanies that's applied to everything you don't like. Racism! Bigotry! Xenophobia!

Oh f6ck I thought I was communicating with someone living in reality. My bad.

https://www.ohchr.org/en/stories/2024/07/racism-and-ai-bias-past-leads-bias-future
https://hai.stanford.edu/news/covert-racism-ai-how-language-models-are-reinforcing-outdated-stereotypes
https://link.springer.com/article/10.1007/s00146-024-01893-4

But since youre so in love with AI heres what Claude said when I asked if AI perpetuates racism and xenophobia

"Yes — there is strong, well‑documented evidence that AI systems can perpetuate racism and xenophobia, and this has been confirmed by academic research, UN human‑rights investigations, and real‑world audits of deployed AI systems.

The core finding across these sources is that AI reproduces and amplifies the biases present in its training data, design choices, and deployment context, creating discriminatory outcomes even when race or ethnicity are not explicitly provided as inputs.

📌 Key Evidence That AI Perpetuates Racism and Xenophobia
1. UN Human Rights Investigations
The UN Special Rapporteur on racism reports that the assumption that AI is “neutral” is false — AI systems can perpetuate racial discrimination through biased data, biased algorithms, and biased deployment.
Examples include predictive policing systems that reinforce historical over‑policing of minority neighborhoods, creating feedback loops where “bias from the past leads to bias in the future.”

2. Stanford Research on Language Models
A 2024 Stanford study found that major language models (including GPT‑3.5, GPT‑4, RoBERTa, and T5) generate covertly racist outputs, especially against speakers of African American English (AAE).
Models assigned AAE speakers:

lower‑prestige jobs

higher likelihood of criminal guilt

harsher sentences (e.g., death penalty over life imprisonment)

This occurred even when race was not mentioned, showing that dialect alone triggered discriminatory behavior.

3. Predictive Policing and Criminal Justice
Predictive policing algorithms use historical arrest data — which is already racially skewed — and therefore direct more police presence to minority neighborhoods, reinforcing racial disparities.
This creates a self‑reinforcing cycle where over‑policed communities generate more “evidence” of crime, which the algorithm then interprets as justification for further policing.

4. Systemic Bias in AI Design and Data
A UN report identifies four ways AI contributes to racial discrimination:

Biased or incomplete data

Algorithm design choices

Intentional discriminatory use

Lack of accountability mechanisms

These issues appear across healthcare, employment, policing, and immigration systems.

5. Academic Reviews of AI Bias
Scholarly analyses show that AI systems suffer from:

Input bias (biased or unrepresentative training data)

System bias (biases introduced during model design)

Application bias (biases arising from how AI is used in the real world)

These biases lead to injustice, discrimination, and harmful outcomes, especially for racial and ethnic minorities.

📌 What This Means in Practice
Across domains, biased AI has produced:

Higher misidentification rates for darker‑skinned individuals in facial recognition

Unequal healthcare predictions for Black patients

Discriminatory hiring recommendations

Xenophobic patterns in immigration risk scoring and border‑control algorithms

Racialized language outputs from large language models

These are not isolated incidents — they reflect structural patterns.

📌 Why This Happens
AI systems learn from human‑generated data. When that data reflects:

historical racism

unequal policing

biased hiring

discriminatory language

xenophobic narratives

…the AI absorbs and reproduces those patterns, often at scale and with a veneer of objectivity.

📌 If you want, I can go deeper into any of these areas:
predictive policing

language‑model racism

bias in facial recognition"

OP, I laughed. You used AI. What it told you was research looking for racism found racism. This illustrates the limits of AI, acting as a confirmation bias. Intelligent users of AI are aware of confirmation bias, I am always testing myself when I use AI for this reason.

Your fears are just your own confirmation bias. You screech racism, but it is no different than screeching about racism in anything we use, whether media or publications or arts and culture or human interactions. Now, we can also ask whether your claims of racism are well-founded or other factors people like you prefer to blame on racism. Ai just reflects our own humanity back on us.

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:My job as an attorney is uniquely unsuitable for AI, so I really never use it at work.

I use it occasionally to provide high level overviews of non-work topics. One of my kids has recurrent ear infections, so recently I asked Claude to give me an idea of the likely options will we have before his ENT appointment in a few weeks.

I'm an attorney as well and there are things AI can help you with. What kind of law do you practice?

I will often get incorrect or incomplete answers if I ask specific questions, but if you go in expecting that and push back it can be useful.

One thing it's great at is document review, for example.

I’m a prosecutor. Most of my job is conducting grand jury investigations. Grand jury materials, which include basically anything we receive pursuant to a subpoena, cannot go into AI.

Wrong.

All your employer need do is to create a confidential AI system closed off to the outside.

They can easily make it secure, like our banking systems are.

I’m not wrong. It is my employer’s policy that grand jury materials cannot go into AI.

That is a result of your employer's policy, though, not because it is not possible to make secure and confidential closed AI systems to input the information. If you trust your banking data and your health data to be online (or accept that it is online, at least), then there is no good reason why legal materials can't be online too. It is only a matter of time.

Maybe. There's still an open question about how Rule 6(e) applies here. You're oversimplifying, likely because you're not familiar with grand jury practice.

You're correct, I am not, so I am sure there are nuances that I am totally ignorant to. But if we can manage to use AI despite strict HIPAA laws, of which I am very familiar, I would bet we'd be able to use it for grand jury information, we just need to wait for bureaucracy to catch up with things.

I understand the Rule 6(e) concern, and I am not saying anyone should put grand jury material into ChatGPT or some public AI tool. That would obviously be reckless and probably prohibited by policy.

But I think people are confusing "public AI tools are not approved for this" with "AI can never be used for this." Those are very different claims.

The future version is not a prosecutor uploading grand jury materials into a consumer chatbot. The future version is a closed, government controlled AI environment, hosted in something like Azure Government, AWS GovCloud, or another approved secure environment, with no training on the data, no external disclosure, strict access controls, audit logs, retention rules, and use limited to authorized personnel performing official duties.

That is not some fantasy. Government already uses secure cloud systems for highly sensitive workloads, including health data, law enforcement data, financial data, and even classified workloads. So the issue is not whether AI can be made secure. It can. The issue is whether the agency has approved the system, written the policy, and structured the use correctly under Rule 6(e).

And the use case is not "AI replaces the prosecutor." The use case is AI as a force multiplier. It reviews subpoena returns, organizes records, builds timelines, flags inconsistencies, compares witness statements, finds missing documents, drafts witness outlines, and lets the prosecutor ask questions across a massive case file. The attorney still verifies everything, makes the legal judgments, controls the investigation, and decides what is presented.

That is where agents become the real multiplier. You could have one agent reviewing bank records, another organizing phone records, another summarizing subpoena returns, another checking witness statements against documents, another building timelines, another preparing issue lists, and another checking for gaps or contradictions. The prosecutor becomes the person directing and validating a team of AI assistants instead of manually grinding through every document the old way.

That is why I think this is inevitable. The current manual process is not some sacred legal principle. It is just the way things have been done because the tooling did not exist yet or was not approved yet. Once secure, compliant AI systems are adopted, attorneys who know how to use them will be able to handle far more evidence and far more cases than attorneys who refuse to learn the tools.

So yes, today your employer’s policy may prohibit it. That is completely fair. But long term, the answer is not "AI cannot work here." The answer is "we need a properly approved, closed, auditable AI system." And once that exists, the people still doing everything manually are going to be at a major disadvantage.

I never said “AI can never be used for this.” You (or rather, whatever AI spit out this response) are arguing with a straw man. I spend a lot of time looking for patterns in bank and other financial records and it would be great to have AI help with that.

You keep emphasizing security, but that’s not what Rule 6 is concerned with. Rule 6 isn’t HIPAA - it isn’t about secure storage of information. It is about maintaining the “black box” nature of the grand jury. It isn’t just the substance of the materials that matters; it is the fact that the government subpoenaed them in the first place.

Fair point. Rule 6 is not just a security issue, and I should not have compared it too closely to HIPAA.

You are right that the secrecy issue includes the fact that certain records were subpoenaed at all, not just whether the records are stored securely.

My point is narrower: that does not mean AI can never be used. It means AI would have to be inside the authorized grand jury workflow, not treated like an outside public tool.

So the question is not just "is it secure?" The question is whether the AI system is approved, limited to authorized users, protected inside the secrecy wall, logged, and used only to assist the prosecutor.

That is where I think this is headed. AI for bank record review, pattern finding, timelines, and subpoena return organization could be incredibly useful, but only if it is built into the process legally and not used casually.

[Anonymous]

It leads to more wealth concentration as the productivity gains are disproportionately captured by the top 0.1%

[Anonymous]

It's a great deep search engine, you just need to double check the information.

Start using it in your personal life and then suddenly at work when you have a problem to solve, you may think about using AI to do it.

At home, I use it to find and modify recipes (similar to one of the PPs above). I used it ahead of Black Friday to find Apple Mac books within a 15-mile radius of my home and had it tell me the best deals. We just bought a car after 12 years and we put in all of our requirements and asked it to spit us out a list of things to potentially test drive. We then engaged it for further questions about the cars. Ooh, and just a few days ago I told GPT the names of some 'family saga' novels that I have enjoyed reading and asked for recommendations. We also used it to pull together a list of summer camps for our kids, based on price, interesting activities and locations. Good luck!