What does AI actually do for us?

[Anonymous]

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.

Tell me you don't understand AI without telling you don't understand AI.

Maybe you could explain it since some of us are obviously so mentally deficient?

[Anonymous]

I had a presentation and it helped me improve my layout, focus on key messages etc

[Anonymous]

There should be a huge class action against the tech sector for calling LLMs AI. It makes the public think it's a sentient intelligence and ask it for therapy, business decisions, and even love. All this "AI" stuff does is boost productivity if you know how to use it just like when the typewriter was introduced. The tech sector simply made the outrageous claim that they can create a general intelligence so they could rake in investor money. China doesn't claim AI will do anything more than boost worker productivity or be a personal assistant for everyone.

[Anonymous]

Anonymous wrote:I had a presentation and it helped me improve my layout, focus on key messages etc

But why wouldn’t you want to improve your own communication skills? That’s such a major part of having something worth saying in the first place that I can’t believe you are ok with outsourcing it. Organizing your thoughts for a presentation is a key part of working through the idea in and of itself.

[Anonymous]

Anonymous wrote:There should be a huge class action against the tech sector for calling LLMs AI. It makes the public think it's a sentient intelligence and ask it for therapy, business decisions, and even love. All this "AI" stuff does is boost productivity if you know how to use it just like when the typewriter was introduced. The tech sector simply made the outrageous claim that they can create a general intelligence so they could rake in investor money. China doesn't claim AI will do anything more than boost worker productivity or be a personal assistant for everyone.

The L part of LLM is the AI

[Anonymous]

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.

Pardon me, but are you f-ing kidding me? You don't know how to throw together dinner? You don't know why something doesn't rise while baking? Did someone never tell you what the different ingredients actually do?!?

[Anonymous]

I think you have to think of LLMs as making processes more efficient. They are not intelligent in the way we traditionally use that word (we never should have let the "AI" label happen.)

For example, we have to review a high volume of documents from multiple sources that don't use the same format. We trained our system to reformat everything so all the information is presented in the same way and then we can build rules that put the documents into different tiers.

We basically taught the system to triage and then allow people to make the judgements about the documents more efficiently.

You can't rely on "AI" to make the judgement - it can't do that unless your judgement is completely based on definable criteria. If there is anything subjective in your process, you need to keep humans involved.

[Anonymous]

LLMs can hold intelligent and genuinely insightful conversations at a high level, including about niche topics. This makes them genuinely useful for some things, especially if you could use someone to bounce ideas around with.

But these conversations give a false sense of confidence in their overall abilities. They are performing “reasoning,” and making predictions about what reasoning looks like, which is different than actual reasoning.

On the subject of LLMs for legal work — they are pretty good at analyzing the language that is in a contract. But in my experience they are horrid at understanding what’s missing from a contract, which is just as important. And critically, they will not tell you about this glaring blind spot, even as they’re speaking about legal issues with a high degree of fluency. It would be very easy to feel like you’re getting good legal representation, without getting essential parts of legal representation.

So be careful out there, friends.

[Anonymous]

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.

Pardon me, but are you f-ing kidding me? You don't know how to throw together dinner? You don't know why something doesn't rise while baking? Did someone never tell you what the different ingredients actually do?!?

You sound angry. Looking at the fridge and seeing what ingredients you have and typing them into the AI to come up with ideas for dinner is called being smart.

Baking is both easy and a complicated science involving chemistry. Small amounts of flour or temperature changes or even humidity on a given day can affect baking outcomes despite following a recipe perfectly. AI is useful in helping to explain what may have gone wrong.

--experienced cook and baker who uses AI to be even better at cooking and baking, learning more from simple AI chats than a thousand cookbooks.

[Anonymous]

Anonymous wrote:

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 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.

Sigh. You don't understand how any of that works. Please don't repeat that story.

You'd have a lot more credibility if you yourself were able to understand and articulate what it was that I didn't understand.

Ok, let me break it down for you. A brief is a legal document that you file with a court, generally electronically, and they are generally available for public inspection (sealed cases are an exception).

Attorney-client privilege protects information obtained by an attorney during their representation of their client. It allows the attorney to not have to provide that information to third parties BUT if they do so, then they have broken the privilege and can no longer assert it.

Putting a document that you are able to publicly file into AI would break attorney-client privilege how? You're not putting anything into the brief that wouldn't be about to become public knowledge, and even in a sealed case you would be sharing it with the judge and the opposing counsel, hence breaking the privilege.

So the point is that if the lawyer put a brief they wrote, that was going to be filed, into an AI source for review, there wasn't anything in there that would have have been protected by attorney-client privilege and therefore putting it into AI couldn't have nullified the privilege.

Get it?

[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.

Tell me you don't understand AI without telling you don't understand AI.

Maybe you could explain it since some of us are obviously so mentally deficient?

Maybe you could Google it?

I don't think all of you are using the term AI correctly or understanding what it is. There are open-source AI options like Claude, where whatever you put into it can be used by Claude to further its education/understanding. Then there are closed-source AI options like Co-Pilot that your company can house internally that can further the education/understanding of your software, but isn't being used by Co-Pilot as a whole. We have a ChatGPT version called, for example, ChatApple, that is a close-sourced AI option for those of us who work at Apple (I do not, that's an example). I am allowed to input my client's information into ChatApple because it's not going anywhere off my system. I cannot input my client's information into ChatGPT because that would be a violation of multiple things.

If you don't think there are lawyers at firms and with the government who are using closed-source AI for client information, I don't know what to tell you. I understand how grand juries work, and I understand the secrecy behind them and the need to keep that information out of the public domain. But storing those documents on your hard drive is no different than running them through a proprietary closed-source AI system on your computer.

[Anonymous]

Anonymous wrote:I think you have to think of LLMs as making processes more efficient. They are not intelligent in the way we traditionally use that word (we never should have let the "AI" label happen.)

For example, we have to review a high volume of documents from multiple sources that don't use the same format. We trained our system to reformat everything so all the information is presented in the same way and then we can build rules that put the documents into different tiers.

We basically taught the system to triage and then allow people to make the judgements about the documents more efficiently.

You can't rely on "AI" to make the judgement - it can't do that unless your judgement is completely based on definable criteria. If there is anything subjective in your process, you need to keep humans involved.

THIS

That's not what AI is for and it's not what it's good at

[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.

Pardon me, but are you f-ing kidding me? You don't know how to throw together dinner? You don't know why something doesn't rise while baking? Did someone never tell you what the different ingredients actually do?!?

You sound angry. Looking at the fridge and seeing what ingredients you have and typing them into the AI to come up with ideas for dinner is called being smart.

Baking is both easy and a complicated science involving chemistry. Small amounts of flour or temperature changes or even humidity on a given day can affect baking outcomes despite following a recipe perfectly. AI is useful in helping to explain what may have gone wrong.

--experienced cook and baker who uses AI to be even better at cooking and baking, learning more from simple AI chats than a thousand cookbooks.

No, I was laughing at them. Because it's pathetic.

[Anonymous]

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.

This entire response is why I am so scared about AI, and I say this as someone who uses AI extensively on a daily basis and who has been immersed in LLMs for years now. I have agents running nearly 24 hours a day and a generous token budget.

The bolded in particular is alarming.

You do not seem to understand the bias to user affirmation that is a core issue with today’s major LLM providers, and something they’ve been trying (and failing) to fix for years. Your conversations aren’t actually teaching you what opposing views are. They are reflecting back to you what Claude believes you want to hear what the opposing views are, combined with a layer of Claude’s own biases. Your beliefs are being continually affirmed, not challenged, and it is subtle enough that you don’t realize that. The sycophancy is a core issue with today’s LLMs and it shows up across the board.

I look at a post like yours and it is truly terrifying in many respects. We are handing off our abilities to engage in independent thought to a few massive corporations. This is not going to end well.

[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.

Tell me you don't understand AI without telling you don't understand AI.

Maybe you could explain it since some of us are obviously so mentally deficient?

Maybe you could Google it?

I don't think all of you are using the term AI correctly or understanding what it is. There are open-source AI options like Claude, where whatever you put into it can be used by Claude to further its education/understanding. Then there are closed-source AI options like Co-Pilot that your company can house internally that can further the education/understanding of your software, but isn't being used by Co-Pilot as a whole. We have a ChatGPT version called, for example, ChatApple, that is a close-sourced AI option for those of us who work at Apple (I do not, that's an example). I am allowed to input my client's information into ChatApple because it's not going anywhere off my system. I cannot input my client's information into ChatGPT because that would be a violation of multiple things.

If you don't think there are lawyers at firms and with the government who are using closed-source AI for client information, I don't know what to tell you. I understand how grand juries work, and I understand the secrecy behind them and the need to keep that information out of the public domain. But storing those documents on your hard drive is no different than running them through a proprietary closed-source AI system on your computer.

NP. You do not understand at all what the term “open source” means. You are using it completely incorrectly.